Sotagliflozin against placebo across its pivotal cardiovascular outcome programme: three separate estimands, never merged -- two pooled within themselves, one reported from a single trial

In both phase 3 trials that supported sotagliflozin's approval, what is the effect of sotagliflozin compared with placebo on cardiovascular death and worsening heart failure -- counted as total events, counted as time to a first event, and on the three-component atherothrombotic endpoint -- with each of those questions answered separately?

In this section: Registration and administrative information (PROSPERO field set) · What this object claims to contain

Registration and administrative information (PROSPERO field set)

Table 1. Registration and administrative information (PROSPERO field set)
Review title	Sotagliflozin against placebo across its pivotal cardiovascular outcome programme: three separate estimands, never merged -- two pooled within themselves, one reported from a single trial
Review question (PICO)	In both phase 3 trials that supported sotagliflozin's approval, what is the effect of sotagliflozin compared with placebo on cardiovascular death and worsening heart failure -- counted as total events, counted as time to a first event, and on the three-component atherothrombotic endpoint -- with each of those questions answered separately?
Population	Not shown — sglt2i_hf_nma is a network protocol whose intervention is the SGLT2 network, not sotagliflozin alone. State: PROTOCOL_IS_A_NETWORK.
Intervention	Not shown — sglt2i_hf_nma is a network protocol whose intervention is the SGLT2 network, not sotagliflozin alone. State: PROTOCOL_IS_A_NETWORK.
Comparator	Not shown — sglt2i_hf_nma is a network protocol whose intervention is the SGLT2 network, not sotagliflozin alone. State: PROTOCOL_IS_A_NETWORK.
Background / rationale	Not recorded — this object holds no background field, and an introduction generated without one would be argument that no source in this review supports
Eligibility criteria	Eligibility turns on population, intervention and comparator: a phase 3 randomised trial of sotagliflozin against placebo in the pivotal cardiovascular outcome programme. It does not turn on which analysis a trial reported -- section 3.2.4 cautions that eligibility resting on reported outcomes invites outcome reporting bias, and this object carries three estimands precisely so that a trial is not admitted or refused according to which one it happens to print. What the estimand governs is which pool a value may enter, under section 10.9.
Information sources	10 source layers, listed on the Extraction tab
Search strategy	The executed strings, datetimes, filters and hit counts are on the Search tab
Study selection process	Not recorded — this object holds no screening record, so the selection process cannot be described
Number of screeners	Not recorded — this object records no screener count
Data extraction items	Extracted per trial on this object: registry id; primary publication; year; design; arms; per-arm counts and the published effect; the randomised total, separately from the analysed denominator. Fields not listed here were not extracted.
Outcomes and prioritisation	Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio; Time to the first occurrence of that same composite, as a hazard ratio; Time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio
Risk of bias method	RoB 2, per result, five domains — see the risk-of-bias table below
Synthesis methods	Pooled under random-effects with the REML estimator. Whether they were prespecified is not recorded on this object.
Subgroup analyses	Not recorded — this object records no subgroup-analysis field at all, so whether any were prespecified is UNKNOWN -- which is not the same as none having been
Meta-bias assessment	Not recorded — no funnel, Egger or Peters value is held on this object, so no meta-bias assessment is claimed
Certainty assessment	GRADE, recorded for 3 outcome(s) in the results block. The ratings are on the Certainty tab.
Confidence level	95%
Funding	Not recorded — no funding statement is recorded for this review
Competing interests	Not recorded — no declaration is recorded
Built	2026-08-10
Schema	2

What this object claims to contain

Full build over a named programme, with the boundary stated. Both phase 3 cardiovascular outcome trials of this drug are held and each contributes to every estimand it reports. There is exactly one source list, `screening.records`; `screening.excluded` is a view computed from it at build time, row for row, so it cannot say anything the records do not. What is not claimed is a systematic database search: the scope is the pivotal programme as the regulator's own review defines it, and the prior-synthesis search that WAS run is recorded with its terms, its results and its date in a staged file.

Not held in this object. No search record is held in this object. The included set was reconciled against published syntheses rather than produced by a database search, so no query, date or yield can be shown. Treat the included set as a convenience sample, not a systematic one.

In this section: SOTA-P-CARDIA NCT05562063

SOTA-P-CARDIA NCT05562063

This review's decision: excluded.

A sotagliflozin trial, and it is in this log rather than absent because a reader looking for the drug's trials will look for it. It is not part of the pivotal programme this object is scoped to: its registry record describes an investigator-sponsored later-phase mechanistic study in heart failure with preserved ejection fraction without diabetes, with imaging and functional endpoints and an enrolment two orders of magnitude below either pivotal trial. An earlier version of this row called it part of the same programme, which a gate leg correctly read as false of the record. Its endpoints are cardiac and metabolic measures rather than the event composites this object holds, so it contributes to none of the three estimands. It is not excluded for reporting the wrong outcome in the sense section 3.2.4 warns against -- it reports no estimate of any of these three quantities at all, which is a different thing from reporting them and being ignored.

outcome not this review's

What it actually reports: cardiac and metabolic endpoints; no estimate of any composite this object holds

Why that is not this review's measure: It is not stored anywhere in this object in any form. This row exists so that the absence is visible and dated rather than silent.

registry: https://clinicaltrials.gov/study/NCT05562063

In this section: Extracted values, and where each came from · Correction — what this page said about the trials’ endpoint history · Risk of bias (RoB-2) · Overall judgement per trial · Inter-assessor agreement · References · Included studies · Regulatory documents · Methods and guidance · Knowingly not included · Sources · Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio · Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio · What was measured · Pooled result · What this pool holds constant · Contributing trials · Endpoint definitions, read from the registry · The methods rule governing this decision · Leave-one-out sensitivity · Does the answer depend on the pooling method? · Time to the first occurrence of that same composite, as a hazard ratio · Time to the first occurrence of that same composite, as a hazard ratio · What was measured · Pooled result · What this pool holds constant · Contributing trials · Endpoint definitions, read from the registry · The methods rule governing this decision · Leave-one-out sensitivity · Does the answer depend on the pooling method? · Time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio · Time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio · What was measured · Single cohort — no synthesis · Contributing trials · Endpoint definitions, read from the registry · The methods rule governing this decision · Leave-one-out sensitivity

Extracted values, and where each came from

One row per extracted value. Every row carries the value, the verbatim sentence it was read from, a resolvable link to the source, and whether the number was read or derived. Where any of those is absent the row says so rather than omitting the value.

Table 2. Extracted values, and where each came from
Trial / outcome	Value as extracted	Verbatim source sentence	Source links	Read or derived
SOLOIST-WHF hfcv_total	HR 0.67 (95% CI 0.52 to 0.85)	“The primary end point was the total number of deaths from cardiovascular causes and hospitalizations and urgent visits for heart failure (first and subsequent events).” “hazard ratio, 0.67; 95% confidence interval [CI], 0.52 to 0.85”	NCT03521934 · PMID 33200892 · source · primary_oa · registry · regulator_fda	READ from the source as printed The point estimate and interval are stored as the source prints them. The log point and log standard error are DERIVED here, the standard error from the width of the printed interval on the log scale, and are not claimed to appear in any source.
SOLOIST-WHF hfcv_first	HR 0.69 (95% CI 0.56 to 0.85)	“An analysis was performed on the primary endpoint using time to first event analysis” “0.69 [0.56, 0.85] and 0.74 [0.57, 0.95] for investigator-reported and adjudicated events, respectively SOLOIST”	NCT03521934 · PMID 33200892 · source · regulator_fda	READ from the source as printed The point estimate and interval are stored as the source prints them. The log point and log standard error are DERIVED here, the standard error from the width of the printed interval on the log scale, and are not claimed to appear in any source.
SCORED hfcv_total	HR 0.74 (95% CI 0.63 to 0.88)	“The primary end point was changed during the trial to the composite of the total number of deaths from cardiovascular causes, hospitalizations for heart failure, and urgent visits for heart failure.” “hazard ratio, 0.74; 95% confidence interval [CI], 0.63 to 0.88”	NCT03315143 · PMID 33200891 · source · primary_oa · registry · regulator_fda	READ from the source as printed The point estimate and interval are stored as the source prints them. The log point and log standard error are DERIVED here, the standard error from the width of the printed interval on the log scale, and are not claimed to appear in any source.
SCORED hfcv_first	HR 0.78 (95% CI 0.67 to 0.9)	“An analysis was performed on the primary endpoint using time to first event analysis” “0.78 [0.67, 0.90] and 0.72 [0.59, 0.86] for investigator-reported and adjudicated”	NCT03315143 · PMID 33200891 · source · regulator_fda	READ from the source as printed The point estimate and interval are stored as the source prints them. The log point and log standard error are DERIVED here, the standard error from the width of the printed interval on the log scale, and are not claimed to appear in any source.
SCORED mace3_first	HR 0.84 (95% CI 0.72 to 0.99)	“original coprimary end point of the first occurrence of death from cardiovascular causes, nonfatal myocardial infarction, or nonfatal stroke, the hazard ratio was 0.84 (95% CI, 0.72 to 0.99)”	NCT03315143 · PMID 33200891 · source · primary_oa · registry · regulator_fda	READ from the source as printed The point estimate and interval are stored as the source prints them. The log point and log standard error are DERIVED here, the standard error from the width of the printed interval on the log scale, and are not claimed to appear in any source.

Correction — what this page said about the trials’ endpoint history

What this review said. That SCORED changed its primary endpoint during the trial, with no corresponding statement about SOLOIST-WHF. Stated in that one-sided form in at least eight fields, including the GRADE risk-of-bias basis on three outcomes.

What is true. Both programmes changed their primary objectives during the trials. The shift was from time-to-event analyses of adjudicated events to total (first and potentially subsequent) INVESTIGATOR-REPORTED events, and the FDA reviewer records it as pre-specified.

Source: Regulator: FDA integrated review — FDA_216203_IntegratedR.extract.txt

While there was a major change to the primary objectives for SOLOIST and SCORED, the Applicant's pre-specified plan to shift from time-to-event analyses of adjudicated events to total (first and potentially subsequent) investigator-reported events was reasonable and is sufficient to establish substantial evidence of effectiveness.
SOLOIST was originally planned to assess a time-to-event analysis for first and recurrent CV death, HHF and UHFV events.

What this does not change. Two things. First, no rating moves: the risk-of-bias assessment on this object is unadjudicated and every certainty rating is PENDING, and a corrected trial history is an input to that adjudication rather than a reason to re-rate tonight. Second, the claim that neither trial registered a time-to-first-event analysis remains true and is not withdrawn -- both registry records post the amended total-occurrence primary. Registered is not the same as planned.

Bearing on the pending adjudication. The FDA reviewer describes the change as PRE-SPECIFIED and notes a sensitivity analysis of the original time-to-event strategy. A change made before database lock and unblinding is not automatically high risk of bias under RoB 2 domain 5. Both assessors judged selection of the reported result without this fact in front of them, and their recorded reasoning is left untouched here so the adjudicator can see what each reader actually had.

Unresolved: The external review states soloist's original primary was time to first CV death or HF hospitalisation with the recurrent three-component composite secondary. The source held here says the original plan was a time-to-event analysis for first and recurrent CV death, HHF and uhfv. This object holds neither the original protocol nor the final SAP, so the two accounts are recorded side by side rather than one being chosen.

Risk of bias (RoB-2)

Variant:

Unit assessed: A result, not a study -- Handbook 8.2: risk of bias is assessed for a specific result. The same trial is assessed twice here and lands differently: SOLOIST-WHF's total-event result is its registered primary and its first-event result was never registered at all. A study-level rating would have averaged those two into one number and lost the only distinction this assessment actually makes.

Assessor 1: (anthropic family). Assessor 2: (openai family).

Table 3. Risk of bias (RoB-2)
Trial	Result assessed	Domain	Assessor 1 (anthropic)	Assessor 2 (openai)	Agreed
SOLOIST-WHF	Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio	D1 D1_randomisation	no information	no information	yes
SOLOIST-WHF	Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio	D2 D2_deviations	no information	no information	yes
SOLOIST-WHF	Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio	D3 D3_missing_outcome_data	no information	no information	yes
SOLOIST-WHF	Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio	D4 D4_measurement	no information	LOW	NO
SOLOIST-WHF	Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio	D5 D5_selection_of_reported_result	LOW	LOW	yes
SCORED	Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio	D1 D1_randomisation	no information	no information	yes
SCORED	Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio	D2 D2_deviations	no information	no information	yes
SCORED	Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio	D3 D3_missing_outcome_data	no information	no information	yes
SCORED	Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio	D4 D4_measurement	no information	LOW	NO
SCORED	Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio	D5 D5_selection_of_reported_result	HIGH	LOW	NO
SOLOIST-WHF	Time to the first occurrence of that same composite, as a hazard ratio	D1 D1_randomisation	no information	no information	yes
SOLOIST-WHF	Time to the first occurrence of that same composite, as a hazard ratio	D2 D2_deviations	no information	no information	yes
SOLOIST-WHF	Time to the first occurrence of that same composite, as a hazard ratio	D3 D3_missing_outcome_data	no information	no information	yes
SOLOIST-WHF	Time to the first occurrence of that same composite, as a hazard ratio	D4 D4_measurement	some concerns	LOW	NO
SOLOIST-WHF	Time to the first occurrence of that same composite, as a hazard ratio	D5 D5_selection_of_reported_result	HIGH	some concerns	NO
SCORED	Time to the first occurrence of that same composite, as a hazard ratio	D1 D1_randomisation	no information	no information	yes
SCORED	Time to the first occurrence of that same composite, as a hazard ratio	D2 D2_deviations	no information	no information	yes
SCORED	Time to the first occurrence of that same composite, as a hazard ratio	D3 D3_missing_outcome_data	no information	no information	yes
SCORED	Time to the first occurrence of that same composite, as a hazard ratio	D4 D4_measurement	no information	LOW	NO
SCORED	Time to the first occurrence of that same composite, as a hazard ratio	D5 D5_selection_of_reported_result	HIGH	some concerns	NO

Overall judgement per trial

Table 4. Overall judgement per trial
Trial	Result assessed	Assessor 1 (anthropic)	Assessor 2 (openai)	Agreed
SOLOIST-WHF	Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio	some concerns	no overall judgement reached — this assessor answered “no information”, which RoB 2 defines as a response to a signalling question and not as one of the three overall categories (low, some concerns, high)	NO
SCORED	Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio	HIGH	no overall judgement reached — this assessor answered “no information”, which RoB 2 defines as a response to a signalling question and not as one of the three overall categories (low, some concerns, high)	NO
SOLOIST-WHF	Time to the first occurrence of that same composite, as a hazard ratio	HIGH	some concerns	NO
SCORED	Time to the first occurrence of that same composite, as a hazard ratio	HIGH	some concerns	NO

Inter-assessor agreement

This assessment is DUAL — two assessors from different model families, the second asked blind. The agreement rate is withheld pending adjudication. Both readers were given this review’s decision rule; an earlier version of this sentence said the blinding had withheld it, and that was wrong. What the 20 domain comparison(s) show is a reading difference that runs in OPPOSITE directions by domain — on domains 1 to 3 the second assessor more often declines to judge at all, and on domains 4 and 5 it more often judges low. A single agreement rate averages those two behaviours into one number that describes neither. Both assessors’ per-result judgements are shown in the table above, unadjudicated, and neither is this review’s finding. No adjudication has been performed, so this review holds no final risk-of-bias judgement for these results.

References

Included studies

Bhatt DL, et al. Sotagliflozin in Patients with Diabetes and Recent Worsening Heart Failure. N Engl J Med 2021;384:117-128. — https://pubmed.ncbi.nlm.nih.gov/33200892/
Bhatt DL, et al. Sotagliflozin in Patients with Diabetes and Chronic Kidney Disease. N Engl J Med 2021;384:129-139. — https://pubmed.ncbi.nlm.nih.gov/33200891/

Regulatory documents

FDA integrated review, INPEFA (sotagliflozin) NDA 216203 — https://www.accessdata.fda.gov/drugsatfda_docs/nda/2023/216203Orig1s000IntegratedR.pdf
FDA prescribing information, INPEFA — https://www.accessdata.fda.gov/drugsatfda_docs/label/2023/216203s000lbl.pdf
EMA withdrawal assessment, Zynquista — https://www.ema.europa.eu/en/medicines/human/withdrawn-applications/zynquista

Methods and guidance

Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, Welch VA (eds). Cochrane Handbook for Systematic Reviews of Interventions version 6.5. — https://training.cochrane.org/handbook
Sterne JAC, et al. RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ 2019;366:l4898. — https://www.bmj.com/content/366/bmj.l4898
GRADE: Guyatt GH, et al. GRADE guidelines: a new series of articles in the Journal of Clinical Epidemiology. J Clin Epidemiol 2011;64:380-2.

Knowingly not included

NOT RETRIEVED, NOT ASSESSED, NOT POOLED A 2025 analysis of SCORED reporting recurrent major adverse cardiovascular events.
It is a different estimand from anything on this page: recurrent MACE is neither of the two composites pooled here, nor the superseded first-event three-component co-primary carried at k=1. Adding it to an existing pool would merge estimands, which is the single thing this review's title says it does not do.

Sources

Table 5. Sources
Layer	Source	How it was obtained
Regulator: FDA integrated review	Inpefa (sotagliflozin) NDA 216203 integrated review, including the statistical review's sensitivity and supportive analyses https://www.accessdata.fda.gov/drugsatfda_docs/nda/2023/216203Orig1s000IntegratedR.pdf	Staged as text and read. It is the only layer holding the time-to-first-event analysis of either trial's primary composite, which is the single most load-bearing fact in this object's sourcing.
Regulator: FDA approved label	Inpefa (sotagliflozin) US prescribing information, clinical studies https://www.accessdata.fda.gov/drugsatfda_docs/label/2023/216203s000lbl.pdf	Discovered from the product's Drugs@FDA overview page and staged as text. Its efficacy tables are one side of the divergences this object records on the total-occurrence cells.
Regulator: EMA, authorisation for a different indication, withdrawn	Zynquista (sotagliflozin) epar medicine overview and the European Commission's public statement of withdrawal https://www.ema.europa.eu/en/medicines/human/EPAR/zynquista	Both documents staged and read. They are the evidence for the European layer being recorded as not available on every cell: the EU authorisation was for insulin-dependent diabetes, never heart failure, and was withdrawn in 2022.
Regulator: EMA medicine overview	Zynquista (sotagliflozin) epar medicine overview https://www.ema.europa.eu/en/medicines/human/EPAR/zynquista	Staged so the claim that the European indication never mentioned heart failure rests on the document rather than on this object's word.
Prior synthesis (abstract, open)	Vaduganathan et al. SGLT-2 inhibitors in patients with heart failure: a comprehensive meta-analysis of five randomised controlled trials. Lancet 2022. https://pubmed.ncbi.nlm.nih.gov/36041474/	Staged for the reconciliation step. Its abstract names its five trials, one of which is in this object, so the included-study-list check is possible. It reports a pooled value on a harmonized first-event composite and no per-trial row, so it is a scope check and never a value check.
Prior synthesis (abstract, open)	Bantounou et al. Meta-analysis of sotagliflozin, a dual sodium-glucose-cotransporter 1/2 inhibitor, for heart failure in type 2 diabetes. ESC Heart Fail 2025. https://pubmed.ncbi.nlm.nih.gov/39257196/	The drug-specific prior synthesis, staged after a gate leg found it in one step against an earlier claim that none could be located. It pools relative risks across nine studies of broader heart-failure and atherothrombotic outcomes and restates neither trial's own primary as a separate row, so it neither checks nor contradicts the cells here -- which is recorded rather than treated as agreement. The reason is one of quantity, and an earlier version of this note gave a chronology instead: it said the synthesis searched to a date before the regulator's review. That was false in both direction and relevance. Its search ran to early June 2023 -- after the review's stated completion date, after the statistical review filed in March, and after the label was issued -- and a regulatory review is not indexed in the bibliographic databases it searched, so no search date could have reached it.
Record of what the literature does not hold	Prior-synthesis search record for these two trials, 2026-08-10 https://pubmed.ncbi.nlm.nih.gov/	Staged because every cell records the prior-synthesis layer as not available, and an absence claim needs the queries behind it. It lists the four terms run, the two that returned nothing, and the two candidates fetched and rejected with the reason.
Primary paper (abstract, open)	Bhatt et al. Sotagliflozin in Patients with Diabetes and Recent Worsening Heart Failure. N Engl J Med 2021. https://pubmed.ncbi.nlm.nih.gov/33200892/	Abstract staged from PubMed. It prints the total-occurrence primary with its interval and states in its own words that the endpoint counts first and subsequent events.
Primary paper (abstract, open)	Bhatt et al. Sotagliflozin in Patients with Diabetes and Chronic Kidney Disease. N Engl J Med 2021. https://pubmed.ncbi.nlm.nih.gov/33200891/	Abstract staged from PubMed. It prints the changed primary, both original co-primaries, and the fact that the primary endpoint was changed during the trial.
Registry with posted results	ClinicalTrials.gov results records for both trials https://clinicaltrials.gov/study/NCT03521934	Both records are staged under their own accession names; this entry names the first and each cell's cascade links its own record.

Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio

What was measured

Table 6. What was measured
Comparator	placebo on top of background therapy
Estimand	total_occurrence — proportional-rates model over first and subsequent events, reported by every layer as a hazard ratio
Unit of analysis	event, with each participant contributing every qualifying event they have rather than only their first
Case definition	every occurrence of cardiovascular death, hospitalization for heart failure or an urgent heart failure visit -- first and subsequent. The contributing trials are not equally explicit about that and this field no longer says they are. The worsening-heart-failure trial's own abstract names the recurrent counting in those words, in a parenthesis on its primary endpoint. The larger trial's abstract says its primary endpoint is the composite of the total number of the three events and stops there; the recurrent counting is established for that trial by its registry description, which says first and potentially subsequent, and by the regulator's total-occurrence tables. An earlier version of this sentence credited both abstracts with the explicit phrase, which the object's own note on that trial's cell already contradicted. Every layer nonetheless reports both results as a hazard ratio. This field exists so the quantity is identified by what was analysed rather than by what it was called.
Window	from randomisation to the end of the trial's own follow-up
Analysis population	all randomised patients in both trials
Effect scale	combined on the log scale; ESTIMAND UNIFORMITY NOT ESTABLISHED -- the contributing trials have not been shown to measure the same quantity

Pooled result

HR 0.7171 (0.6246 to 0.8234), 95% interval

Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio. random, estimator REML, k = 2.

τ² 0 · I² 0% · Q 0.4296 on 1 df

Favours: the intervention (lower is better)

Computed from Q and its single degree of freedom. With two studies this statistic has almost no power to detect heterogeneity, so a low value is consistent both with agreement and with not being able to tell, and it is reported with that caveat rather than as reassurance. The difference these two trials genuinely have is in their populations, and no statistic computed from two effect estimates can see it -- the uniformity table is the only place it is visible.

One effect measure, one comparator, one estimand and one drug, across trials that share no participant and no control group.

What this pool holds constant

Table 7. What this pool holds constant
Dimension	Across the pooled cohorts	If it differs, why it is crossed anyway
effect measure	identical	Both trials report this quantity as a hazard ratio and no conversion is performed.
comparator	identical	Both randomised against a matching placebo on top of background therapy.
drug and dose	identical	Sotagliflozin, 200 mg once daily titrated to 400 mg as tolerated, in both.
population	differs	This is the axis the pool crosses and it is the largest thing it crosses. One trial enrolled patients hospitalised for worsening heart failure and started treatment around discharge. The other enrolled patients with chronic kidney disease and cardiovascular risk in whom heart failure was not an entry requirement, and most did not have it. The pooled value is therefore a DRUG-WIDE estimate across the pivotal programme, which is the question this object asks, and it is not an estimate for a heart-failure population. The heart-failure-only result is the first trial's own row and a reader who wants it should read that row rather than the pool.
why stopped	identical	Both trials were terminated early because the sponsor withdrew funding. That is not a data-dependent stopping rule and it is not a boundary crossing; it is a commercial event, and it applies to both. What it cost was information -- power and follow-up -- so it is carried into the GRADE imprecision domain, where an unmet optimal information size is the ground for that downgrade. It is not carried into risk of bias: lost information widens an interval rather than biasing it. An earlier version of this row said the opposite and contradicted the risk-of-bias domain it was describing.
follow up length	differs	Medians of about nine and about sixteen months. Each trial's own median is stored on its cell.
analysis population	identical	All randomised patients in both, with the registry posting the same two denominators on every outcome.
endpoint rank within its own trial	identical	This composite is the primary endpoint of both trials as they were finally analysed. In one of them it became the primary by a change made during the trial, which the trial's own abstract states and which this object records on the cell and in the risk-of-bias domain rather than in a footnote.
ascertainment	identical	Both trials' primary analyses rest on investigator-reported events, which the regulator's review states in those words. Centrally adjudicated versions of this endpoint DO exist for both trials in that review, and this outcome does not pool them: the parallel-ascertainment pool lives on the first-event estimand only, where both bases are printed in one regulator sentence. Here the contributing cells come from the publication and the registry on an investigator-reported basis, and building a second pool from a layer the cells are not drawn from would be a different synthesis wearing this one's name. An earlier version of this row said the adjudicated versions were reported in this outcome's sensitivity block; they are not, and a review leg was right that the prose claimed a feature the structure denies.

Contributing trials

Table 8. Contributing trials
Trial	Analysed treatment / control	HR (95% CI), or events	Source of this cell
SOLOIST-WHF NCT03521934	608 / 614 All randomised patients. The registry posts these denominators on every outcome of this trial and they sum to its randomised total, so there is no gap between randomised and analysed to reconcile.	0.67 (0.52 to 0.85) on the log scale: -0.4005 (standard error 0.1254)	Primary paper (abstract, open): The primary end point was the total number of deaths from cardiovascular causes and hospitalizations and urgent visits for heart failure (first and subsequent events). hazard ratio, 0.67; 95% confidence interval [CI], 0.52 to 0.85 Randomised, as posted by the registry, and equal to the arms below. Arm-level event counts are deliberately absent: this trial's primary endpoint counts first and subsequent events, so a per-arm participant count of events would not be the numerator of the stored ratio and printing one would invite exactly the misreading this object exists to prevent. The sources print event rates per 100 patient-years instead, and those are quoted on the cell.
SCORED NCT03315143	5292 / 5292 All randomised patients. The registry posts these denominators on every outcome of this trial and they sum to its randomised total, so there is no gap between randomised and analysed to reconcile.	0.74 (0.63 to 0.88) on the log scale: -0.3011 (standard error 0.08526)	Primary paper (abstract, open): The primary end point was changed during the trial to the composite of the total number of deaths from cardiovascular causes, hospitalizations for heart failure, and urgent visits for heart failure. hazard ratio, 0.74; 95% confidence interval [CI], 0.63 to 0.88 Randomised, as posted by the registry, and equal to the arms below. Arm-level event counts are deliberately absent: this trial's primary endpoint counts first and subsequent events, so a per-arm participant count of events would not be the numerator of the stored ratio and printing one would invite exactly the misreading this object exists to prevent. The sources print event rates per 100 patient-years instead, and those are quoted on the cell.

The contributing trials share no participant and no control group.

Endpoint definitions, read from the registry

What each trial COUNTED, in its own registry record's words, before anything here was pooled. A sentence saying what HAPPENED is not a sentence saying what was COUNTED, and only the second one licenses a pool. Read from the registry on 2026-08-17. Follow the registration link to disagree with any row.

Table 9. Endpoint definitions, read from the registry
Trial	Rank in its own trial	Registered outcome measure	Description, verbatim	Analysis set / window
SOLOIST-WHF NCT03521934	—	Number of Total Occurrences of Cardiovascular (CV) Death, Hospitalizations for Heart Failure (HHF) and Urgent Visits for Heart Failure (HF)	Combined endpoint of the total number of occurrences (first and potentially subsequent) of CV death, HHF, and urgent HF visits after randomization. Events that occurred during the study were calculated as the total number of events per 100 person-years of follow-up.	Up to 21.9 months
SCORED NCT03315143	—	Number of Total Occurrences of Cardiovascular (CV) Death, Hospitalizations for Heart Failure (HHF) and Urgent Visits for Heart Failure (HF)	Combined endpoint of the total number of occurrences (first and potentially subsequent) of CV death, HHF, and urgent HF visits after randomization. Events that occurred during the study were calculated as the total number of events per 100 person-years of follow-up.	Up to 30 months

The methods rule governing this decision

Decision: Pool the two pivotal trials on this estimand alone, producing a drug-wide average across two documented population differences, and combine it with NO other estimand in this object

Cochrane Handbook sections: 3.2.1, 3.2.3, 3.2.4, 10.3, 10.8, 10.10.1, 10.10.2, 10.10.4, 10.10.4.4, 10.10.4.5, 10.14

Sections 3.2.1 and 3.2.3 fix the axes on which eligibility turns -- population and comparator. Section 3.2.4 is cited for the caution it gives against making eligibility depend on which outcomes a study reported, which is why this object carries three outcomes rather than admitting only the trials that report one of them. Section 10.3 gives the method -- log ratios combined by generic inverse variance -- and is the authority for combining within each estimand. What licenses the refusal to combine across them is that the estimands fall under different sections: a composite counting first and subsequent events is governed by 10.8, on counts and rates, while a time-to-first analysis is governed by 10.9. An earlier version cited 10.9 for both and for the method as well, which made the object commit in its own citations the conflation it exists to refuse. Section 10.10.1 is what distinguishes clinical and methodological diversity from statistical heterogeneity, which is why the uniformity table sits beside the statistic rather than being replaced by it -- with two trials the statistic is nearly powerless and the population difference it cannot see is the largest thing about this pool. Section 10.10.4 governs the model and the between-study variance estimator, 10.10.4.4 records restricted maximum likelihood as the current default, and 10.10.4.5 is why a more conservative interval is reported beside the ordinary one at this number of studies. Section 10.14 is the authority for the sensitivity analyses.

Leave-one-out sensitivity

Whether the pooled estimate depends on either single trial, on the choice of between-study variance estimator and interval method, and -- where two ascertainment bases exist -- on which of them is used.

Table 10. Leave-one-out sensitivity
Cohort omitted	k	HR (95% CI)
soloist-whf	1	0.74 (0.6261 to 0.8746)
scored	1	0.67 (0.524 to 0.8566)

With two trials a leave-one-out fit IS the other trial, and this table says so rather than presenting single-trial rows as though they were pooled sensitivity results. Read it as what it is: the two trials' own estimates, side by side, with the interval each carries alone.

Cochrane Handbook version 6.5.1, section 10.14 for sensitivity analyses generally. Section 10.10.4.4 governs the estimator half: it records that methods other than DerSimonian and Laird estimate the between-study variance better and that restricted maximum likelihood is the current default. Section 10.10.4.5 is cited for the interval comparison and not for the estimator comparison; it concerns interpretation with few studies and the method it compares is the confidence interval.

Does the answer depend on the pooling method?

Section 10.10.4.4 records restricted maximum likelihood as the current default and that estimators other than DerSimonian and Laird perform better; that is the whole of the authority for running this comparison. Nothing in the Handbook instructs a review to run it at a particular number of studies, and running it here is a choice made under 10.14 because with two trials the choice is plausibly influential.

Table 11. Does the answer depend on the pooling method?
Between-study variance	Interval	Summary (95% CI)
DerSimonian-Laird (DL)	ordinary (Wald)	0.7171 (0.6246 to 0.8234)
Paule-Mandel (PM)	ordinary (Wald)	0.7171 (0.6246 to 0.8234)
restricted maximum likelihood (REML)	ordinary (Wald)	0.7171 (0.6246 to 0.8234)
restricted maximum likelihood (REML)	Hartung-Knapp-Sidik-Jonkman (HKSJ)	0.7171 (0.2928 to 1.756)

Reported as computed rather than summarised in advance. Where the estimators agree it is because there is little between-study variance for them to disagree about, which at two studies is close to uninformative; where the conservative interval is much wider than the ordinary one, that width is the honest expression of how little two trials constrain a random-effects average, and section 10.10.4.5 says to expect it.

Restricted maximum likelihood with the ordinary interval, because section 10.10.4.4 records it as the current default and as a better estimator than DerSimonian and Laird. The Hartung-Knapp-Sidik-Jonkman interval is reported beside it with the floor on its scale factor applied, so it can widen and can never come out narrower than the ordinary one. It is not presented as the interval to prefer: section 10.10.4.5 says the opposite is possible, that with only two or three studies such intervals can be excessively wide and an ordinary interval may be preferable, and it asks for the comparison rather than for a winner. An earlier version of this field called it the interval to weigh most heavily, which attributed to that section a preference it does not express.

Time to the first occurrence of that same composite, as a hazard ratio

What was measured

Table 12. What was measured
Comparator	placebo on top of background therapy
Estimand	time_to_first — Cox proportional hazards on the time to a first event
Unit of analysis	participant, with the time to the first component event
Case definition	the first occurrence, per participant, of that same composite. The two contributing trials do not stand in the same relation to it, and an earlier version of this field said neither trial designated it at all. For the larger trial the regulator's review lists this first-occurrence composite among that trial's Exploratory Endpoints, so it is a designated endpoint there, at exploratory rank. For the worsening-heart-failure trial it appears in no such list and is a regulator analysis only. In both trials the values stored here are the regulator's own time-to-first analysis of the trial's primary composite, reported on an investigator-reported and a centrally adjudicated basis, and both bases are stored. The asymmetry is declared on the uniformity row rather than smoothed away.
Window	from randomisation to the end of the trial's own follow-up
Analysis population	all randomised patients in both trials
Effect scale	combined on the log scale; ESTIMAND UNIFORMITY NOT ESTABLISHED -- the contributing trials have not been shown to measure the same quantity

Pooled result

HR 0.7488 (0.6638 to 0.8446), 95% interval

Time to the first occurrence of that same composite, as a hazard ratio. random, estimator REML, k = 2.

τ² 0 · I² 0% · Q 0.8841 on 1 df

Favours: the intervention (lower is better)

One effect measure, one comparator, one estimand and one drug, across trials that share no participant and no control group.

What this pool holds constant

Table 13. What this pool holds constant
Dimension	Across the pooled cohorts	If it differs, why it is crossed anyway
effect measure	identical	Both trials report this quantity as a hazard ratio and no conversion is performed.
comparator	identical	Both randomised against a matching placebo on top of background therapy.
drug and dose	identical	Sotagliflozin, 200 mg once daily titrated to 400 mg as tolerated, in both.
population	differs	This is the axis the pool crosses and it is the largest thing it crosses. One trial enrolled patients hospitalised for worsening heart failure and started treatment around discharge. The other enrolled patients with chronic kidney disease and cardiovascular risk in whom heart failure was not an entry requirement, and most did not have it. The pooled value is therefore a DRUG-WIDE estimate across the pivotal programme, which is the question this object asks, and it is not an estimate for a heart-failure population. The heart-failure-only result is the first trial's own row and a reader who wants it should read that row rather than the pool.
why stopped	identical	Both trials were terminated early because the sponsor withdrew funding. That is not a data-dependent stopping rule and it is not a boundary crossing; it is a commercial event, and it applies to both. What it cost was information -- power and follow-up -- so it is carried into the GRADE imprecision domain, where an unmet optimal information size is the ground for that downgrade. It is not carried into risk of bias: lost information widens an interval rather than biasing it. An earlier version of this row said the opposite and contradicted the risk-of-bias domain it was describing.
follow up length	differs	Medians of about nine and about sixteen months. Each trial's own median is stored on its cell.
analysis population	identical	All randomised patients in both, with the registry posting the same two denominators on every outcome.
endpoint rank within its own trial	differs	The rows are not alike on this axis and this object no longer says they are. The regulator's review lists, under Exploratory Endpoints for the larger trial, the first occurrence of the composite of cardiovascular death, hospitalization for heart failure or an urgent heart failure visit -- so for that trial this analysis IS a designated endpoint of the trial, at exploratory rank. The same list for the worsening-heart-failure trial does not contain it: there the quantity is a regulator analysis and nothing more. This row has now been wrong in both directions -- an earlier version marked it differing with prose beside it denying any difference, that was corrected to identical on the premise that neither trial designated it, and the premise was false of the larger trial. It is recorded as differing with the asymmetry named, which is what the source supports. What remains true of both rows equally: this is a sensitivity analysis of each trial's primary composite, performed and reported by the regulator, and in one of the trials the composite it re-analyses had itself been changed mid-trial. A pool of regulator sensitivity analyses is a legitimate quantity and it is not the same kind of evidence as a pool of prespecified primary endpoints. That does not show up as a different certainty rating here -- this outcome and the total-occurrence one carry the same rating from the same two downgrades -- and an earlier version of this row claimed it did.
ascertainment	identical	Both stored cells use investigator-reported events, so the contributing rows do not differ on this axis and an earlier version of this row was wrong to say they did -- both review families caught it, and the companion outcome had always recorded the same situation as identical. What differs is not between the rows but between two analyses of each row: centrally adjudicated estimates exist for both trials, are stored beside each cell, and pool to a visibly different summary reported in the sensitivity block. Those are a sensitivity analysis, not contributing rows. They are worth reading because the two bases move the two trials in opposite directions, which is the clearest signal in this object that ascertainment is not a rounding detail.

Contributing trials

Table 14. Contributing trials
Trial	Analysed treatment / control	HR (95% CI), or events	Source of this cell
SOLOIST-WHF NCT03521934	not stated / not stated	0.69 (0.56 to 0.85) on the log scale: -0.3711 (standard error 0.1065)	Regulator: FDA integrated review: An analysis was performed on the primary endpoint using time to first event analysis 0.69 [0.56, 0.85] and 0.74 [0.57, 0.95] for investigator-reported and adjudicated events, respectively SOLOIST Randomised, as posted by the registry, and equal to the arms below. Arm-level event counts are deliberately absent: this trial's primary endpoint counts first and subsequent events, so a per-arm participant count of events would not be the numerator of the stored ratio and printing one would invite exactly the misreading this object exists to prevent. The sources print event rates per 100 patient-years instead, and those are quoted on the cell.
SCORED NCT03315143	not stated / not stated	0.78 (0.67 to 0.9) on the log scale: -0.2485 (standard error 0.07529)	Regulator: FDA integrated review: An analysis was performed on the primary endpoint using time to first event analysis 0.78 [0.67, 0.90] and 0.72 [0.59, 0.86] for investigator-reported and adjudicated Randomised, as posted by the registry, and equal to the arms below. Arm-level event counts are deliberately absent: this trial's primary endpoint counts first and subsequent events, so a per-arm participant count of events would not be the numerator of the stored ratio and printing one would invite exactly the misreading this object exists to prevent. The sources print event rates per 100 patient-years instead, and those are quoted on the cell.

The contributing trials share no participant and no control group.

Endpoint definitions, read from the registry

Table 15. Endpoint definitions, read from the registry
Trial	Rank in its own trial	Registered outcome measure	Description, verbatim	Analysis set / window
SOLOIST-WHF NCT03521934	—	Cardiovascular (CV) Death, Hospitalizations for Heart Failure (HHF) and Urgent Visits for Heart Failure (HF) -- the same event class as this trial's registered primary endpoint, analysed as time to first event instead of as total occurrences	no description is recorded in this registry field	—
SCORED NCT03315143	—	Cardiovascular (CV) Death, Hospitalizations for Heart Failure (HHF) and Urgent Visits for Heart Failure (HF) -- the same event class as this trial's registered primary endpoint, analysed as time to first event instead of as total occurrences	no description is recorded in this registry field	—

The methods rule governing this decision

Decision: Pool the two pivotal trials on this estimand alone, producing a drug-wide average across two documented population differences, and combine it with NO other estimand in this object

Cochrane Handbook sections: 3.2.1, 3.2.3, 3.2.4, 10.3, 10.9, 10.10.1, 10.10.2, 10.10.4, 10.10.4.4, 10.10.4.5, 10.14

Leave-one-out sensitivity

Table 16. Leave-one-out sensitivity
Cohort omitted	k	HR (95% CI)
soloist-whf	1	0.78 (0.673 to 0.904)
scored	1	0.69 (0.5601 to 0.8501)

Does the answer depend on the pooling method?

Table 17. Does the answer depend on the pooling method?
Between-study variance	Interval	Summary (95% CI)
DerSimonian-Laird (DL)	ordinary (Wald)	0.7488 (0.6638 to 0.8446)
Paule-Mandel (PM)	ordinary (Wald)	0.7488 (0.6638 to 0.8446)
restricted maximum likelihood (REML)	ordinary (Wald)	0.7488 (0.6638 to 0.8446)
restricted maximum likelihood (REML)	Hartung-Knapp-Sidik-Jonkman (HKSJ)	0.7488 (0.3429 to 1.635)

Time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio

What was measured

Table 18. What was measured
Comparator	placebo on top of background therapy
Estimand	time_to_first — Cox proportional hazards on the time to a first event
Unit of analysis	participant, with the time to the first component event
Case definition	the first occurrence, per participant, of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke. It was an original co-primary of the trial that reports it, before that trial changed its primary endpoint.
Window	from randomisation to the end of the trial's own follow-up
Analysis population	all randomised patients in both trials
Effect scale	reported on the log scale; nothing is pooled on this outcome

Single cohort — no synthesis

HR 0.84 (0.72 to 0.99)

Favours: the intervention (lower is better)

Only one contributing trial reports this estimand. The companion trial posts no three-component analysis at all: its nearest registry row is a total-occurrence composite of four components, adding hospitalization for heart failure to these three. An earlier version of this field called it the same three components, which was false of the record, and a gate leg was right to read the registry rather than this sentence.

Contributing trials

Table 19. Contributing trials
Trial	Analysed treatment / control	HR (95% CI), or events	Source of this cell
SCORED NCT03315143	not stated / not stated	0.84 (0.72 to 0.99) on the log scale: -0.1744 (standard error 0.08124)	Primary paper (abstract, open): original coprimary end point of the first occurrence of death from cardiovascular causes, nonfatal myocardial infarction, or nonfatal stroke, the hazard ratio was 0.84 (95% CI, 0.72 to 0.99) Randomised, as posted by the registry, and equal to the arms below. Arm-level event counts are deliberately absent: this trial's primary endpoint counts first and subsequent events, so a per-arm participant count of events would not be the numerator of the stored ratio and printing one would invite exactly the misreading this object exists to prevent. The sources print event rates per 100 patient-years instead, and those are quoted on the cell.

The contributing trials share no participant and no control group.

Endpoint definitions, read from the registry

Table 20. Endpoint definitions, read from the registry
Trial	Rank in its own trial	Registered outcome measure	Description, verbatim	Analysis set / window
SCORED NCT03315143	—	the first occurrence of death from cardiovascular causes, nonfatal myocardial infarction, or nonfatal stroke	no description is recorded in this registry field	—

The methods rule governing this decision

Decision: Do not pool. One trial reports this endpoint in this form, so the result is that trial's own estimate, presented as such

Cochrane Handbook sections: 3.2.4, 10.8, 10.9

There is no meta-analysis here and none is implied. The companion trial posts a total-occurrence composite of four components -- these three plus hospitalization for heart failure -- so it differs from this cell in both its event set and its unit of analysis, and the two differences are governed by different sections: the unit of analysis by 10.8 read against 10.9, and the event set by 3.2.4, under which a composite's components are part of how the review defines its outcome -- so two composites differing in a component are two outcomes rather than two names for one. That is a limit on what may be pooled, not a criterion for which trials are eligible: 3.2.4 cautions against letting eligibility depend on which outcomes a study reported, and this review's eligibility does not. An earlier version of this field read it the other way round and called the outcome a defining criterion of the review, which is the over-strict reading the object disowns elsewhere. An earlier version credited both to 10.9 alone, which is a time-to-event section and has nothing to say about which events a composite contains; an earlier version still called this the same three components, which was false of the registry record. Section 3.2.4 is also why this outcome is carried at all rather than dropped for being reported by only one trial: dropping it would make the review's content depend on what the studies chose to report.

Leave-one-out sensitivity

none; there is one study

Not run. With one study there is nothing to leave out, and an object that printed a table here would be manufacturing reassurance.

Cochrane Handbook version 6.5.1, section 10.14. A leave-one-out analysis of one study leaves nothing, and an estimator comparison with no between-study variance to estimate is not a sensitivity analysis. Neither is run, and this field says so rather than presenting an empty table.

In this section: Summary of findings · Reproducibility artifact · Other recorded qualifications · Comparison with published syntheses · The syntheses reconciled against · Where the numbers differ, and why · Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio · Certainty of the evidence (GRADE) · Time to the first occurrence of that same composite, as a hazard ratio · Certainty of the evidence (GRADE) · Time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio · Certainty of the evidence (GRADE)

Summary of findings

Table 21. Summary of findings
Outcome	k	Pooled effect	Certainty
Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio	2	HR 0.717 (0.625 to 0.823)	Pending¹²
Time to the first occurrence of that same composite, as a hazard ratio	2	HR 0.749 (0.664 to 0.845)	Pending³²
Time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio	1	not pooled	Pending⁴²

This certainty rating is PENDING, not rated. It rests on a risk-of-bias judgement that this review has not finalised: two assessors read every contributing result independently, they disagree, and no adjudication has been performed, so this review holds no final risk-of-bias judgement for these results. Cochrane rates certainty per outcome by aggregating the risk-of-bias judgements of the results contributing to it, so a certainty rating cannot be more final than the risk-of-bias assessment it reads. The dependency is not inferred: this object records the risk-of-bias domain as supported by “A RoB 2 assessment of every result contributing to this pool, added 2026-08-20. 1 of 2 results are HIGH overall, all of them on domain 5, selection of the reported result”, which is one assessor's count of results at HIGH risk of bias -- a count the second assessor's judgements do not contain, on any result. The rating this object currently records, and the steps behind it, are shown below as the work so far rather than as this review's answer. No level is published for this outcome -- not the recorded one, not a more cautious one, and not a midpoint -- because choosing one would be the judgement that has not been made.
Read this beside `whether the estimand was established`. That flag records that every contributing trial measures the same quantity. It records nothing about what that quantity was measured against -- whether the comparators are of one kind, whether each comparison was randomised, whether it was concurrent. The two questions were never separated. Named as P48 in PAGE-STANDARD.md v1.21.0 and swept by scripts/audit_mixed_contrast_pools.py.
This certainty rating is PENDING, not rated. It rests on a risk-of-bias judgement that this review has not finalised: two assessors read every contributing result independently, they disagree, and no adjudication has been performed, so this review holds no final risk-of-bias judgement for these results. Cochrane rates certainty per outcome by aggregating the risk-of-bias judgements of the results contributing to it, so a certainty rating cannot be more final than the risk-of-bias assessment it reads. The dependency is not inferred: this object records the risk-of-bias domain as supported by “A RoB 2 assessment of every result contributing to this pool, added 2026-08-20. 2 of 2 results are HIGH overall, all of them on domain 5, selection of the reported result”, which is one assessor's count of results at HIGH risk of bias -- a count the second assessor's judgements do not contain, on any result. The rating this object currently records, and the steps behind it, are shown below as the work so far rather than as this review's answer. No level is published for this outcome -- not the recorded one, not a more cautious one, and not a midpoint -- because choosing one would be the judgement that has not been made.
This certainty rating is PENDING, not rated. It rests on a risk-of-bias judgement that this review has not finalised: this review's RoB 2 assessment does not cover this outcome at all -- it assesses 4 result(s) across 2 other outcome(s) -- so the risk-of-bias domain here rests on no assessment recorded on this object. Cochrane rates certainty per outcome by aggregating the risk-of-bias judgements of the results contributing to it, so a certainty rating cannot be more final than the risk-of-bias assessment it reads. The rating this object currently records, and the steps behind it, are shown below as the work so far rather than as this review's answer. No level is published for this outcome -- not the recorded one, not a more cautious one, and not a midpoint -- because choosing one would be the judgement that has not been made.

Reproducibility artifact

Table 22. Reproducibility artifact
Canonical object	`—`
Registered at	`—` —
Schema	2
Object last updated	2026-08-10
Source object SHA-256, first 16 hex characters	`3d67ee0a6edb3491`
Page generated	2026-08-28 19:54 UTC
Generator build	`a441e4187`. The standard this page was built to is NOT ESTABLISHED: this object carries no `build_stamp`, so the version above is the standard that exists now, not the one this page was built against.
Statistical engine	metafor 5.0.1 under R 4.6.0, generic inverse variance on the same log point and log standard error this object stores

This page CANNOT be rebuilt from what is recorded here: no canonical object is recorded and no registered commit is recorded. Each figure on the Analysis tab downloads as an SVG carrying exactly the values shown.

Other recorded qualifications

Statements this review records about the limits of what it establishes, which are not shown elsewhere on this page. They are reproduced from the object verbatim and are not summarised.

Table 23. Other recorded qualifications
Analysed scope	All randomised patients. The registry posts these denominators on every outcome of this trial and they sum to its randomised total, so there is no gap between randomised and analysed to reconcile.
Analysed absent because	This trial has one analysis set and this object records it once, on the row for the trial's own primary endpoint. The denominators are not unknown and they are not different here -- every outcome of this trial was analysed on all randomised patients, which the registry posts on every one of its outcome records. They are omitted from this row on purpose: repeating them under a different effect estimate would assert three independently-scoped analysis populations where the trial has one, and that is the shape a structural rule in this repository blocks, for a good reason learned from a trial that really did borrow a wider population for a narrower contrast.
Registry does not post	The registry posts this trial's composite only as total occurrences. The time-to-first analysis of the same composite exists in the regulator's review and nowhere in the registry, so the event class is quoted from the registry field above and the analysis is named from the review. Those are two different facts and they are recorded separately on purpose.
Endpoint was changed during the trial	The publication states the primary end point was changed during the trial. The registry record read here is the amended one, and it is the endpoint the reported hazard ratio belongs . The superseded original coprimary is carried separately in this object as mace3 first, from a single trial, and is not pooled.
Registry does not post	As for SOLOIST-WHF: the registry posts the composite only as total occurrences. Note the publication also reports a first-occurrence composite for this trial, but of CV death OR hospitalisation for heart failure without urgent visits -- a different event class, which is why it is held out rather than used here. That hold-out is recorded in this object's scope decisions.
Registry does not post	The current registry record does not post a FIRST-occurrence version of this composite. It posts a TOTAL-occurrence one ('Total Number of Occurrences of CV Death, Non-fatal Myocardial Infarction and Non-fatal Stroke') as a secondary outcome, which is a different quantity from the one stored here. The first-occurrence composite is the superseded original coprimary and survives only in the publication. This is exactly the case the registry read cannot settle alone, and it is why this outcome is reported from a single trial and pooled with nothing.
Not stated here because no source is held	The proportion of SCORED participants with heart failure at baseline is not stated on this page. The external review gives approximately 31%, and that figure is not in any source this object holds -- not the staged abstract, not the registry record, not the FDA integrated review extract. It is listed as extraction work rather than repeated here.
Why that quantity is never stored as a hazard ratio	It is not stored anywhere in this object in any form. This row exists so that the absence is visible and dated rather than silent.
Record limitation	This row is written from the registry record and from the trial's place in the programme. Its results were not read for this object and no value from it is claimed, which is stated here rather than left for a reader to infer from the empty quantity field.
Access limitation	The synthesis reconciled against is staged as its abstract; its full text is not open, which was established by querying the full-text archive for its identifier rather than inferred from a paywall notice. Its abstract names all five of its trials, so the included-study list -- the thing this step needs -- is readable in full. Its per-trial rows are not, which is why no per-trial value is attributed to it anywhere in this object.
Review scope	SGLT2 inhibitors in heart failure across the ejection fraction spectrum, five randomised trials, fixed-effect, with endpoint definitions harmonized by the authors
Scope	sotagliflozin in type 2 diabetes, any randomised design reporting heart failure events
Scope	narrative systematic review comparing sotagliflozin and dapagliflozin
What this assessment could actually decide	One domain of five, and it is worth saying plainly rather than burying in the table. Domain 5 is judged from evidence -- the full registered outcome list for both trials, read from ClinicalTrials.gov on 2026-08-18, which Handbook 8.7 names as the method for this domain. Domains 1 to 3 are no information on all four results. Domain 4 is judged on one result of four, and only because this object happens to hold the same estimand measured a second way.
Why stated rather than left implicit	A ceiling that is not stated looks like a verdict. A reader who sees three of five domains unjudged and no explanation will read the overall rating as a criticism of the trials, and it is not one.
This does not replace the stored judgement	This is the judgement the published algorithm proposes from the signalling responses named here. It does not replace the stored `judgement` beside it, and nothing reads it yet. The tool itself says its algorithms "provide proposed judgements, but users should verify these and change them if they feel this is appropriate" -- so this is an input to a person, not an output to a reader.
Registry facts backfilled	Read from ClinicalTrials.gov API v2 on 2026-08-21. The judgements above were not revisited -- this adds the registry evidence beside a verdict that was made from it and did not record it, so a blind second assessor can be shown the same facts. A second assessment built on fewer facts is not a comparable assessment.
What this does not establish	That the pooled estimate is wrong, or that either trial is unreliable. Domain 5 being HIGH on three of four results means the results we use were selected from among more analyses than were pre-specified -- on SCORED because its primary was changed mid-trial, and on both first-event values because no first-event analysis was ever registered. It is a reason to weigh the total-event pool, which rests on soloist's registered primary, above the first-event pool, which rests on regulator-reported analyses on both trials.
The allow list the rate is conditional on	A disagreement rate is meaningless without the facts both assessors saw, so the list is stated here the way a denominator is stated. Fields emitted: trial, cohort, registered enrolment, registered masking, registered sites, registered comparator, number of registered arms, number of registered primary outcomes, registered primary outcome, registered analysis population, the rank the pooled result holds, the result being assessed, participants this review pools, registered enrolment minus participants pooled. Established by test, not argument: widening this list from 9 fields to 14 halved gepotidacin's disagreement -- its entire selection of the reported result divergence was a fact the second assessor had never been shown -- and left ceftaroline and tigecycline unmoved.
What this review does not publish	What this review does not publish. This review also examined one outcome for which no pooled estimate is published: Time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio (k=1), which is not pooled. Each of those outcomes is recorded on this review with the reason it was not pooled, and none of them contributes to the estimate reported above (0.7171 at k=2, 0.7488 at k=2). A reader meeting the pooled result should not take it as covering every outcome, or every trial, this review examined. This sentence exists because a second, independent reading -- by a different model family, given the object and the abstract and none of our conclusions -- found that a reader would reasonably conclude the review pools all relevant trials, and that the object does not support that. The withdrawals were already recorded beside the outcomes they concern; they were invisible from the abstract, which is where the impression forms.
Which risk of bias domains drove the rating	No domain is named as driving the rating, because this review holds no final risk-of-bias judgement to be driven: two assessors read every contributing result independently and no adjudication has been performed. Both readings are given rather than one: Claude Opus 5 (anthropic family) records selection of the reported result on 3 result(s); GPT-5 Codex (openai family), via `codex exec` records selection of the reported result on 2 result(s) (and reaches no overall judgement at all on 2 result(s), for want of documents this review did not retrieve). Naming a single domain here would publish one assessor's count as the review's finding, which the risk-of-bias section of this same page explicitly declines to do.
What this does not change	Two things. First, no rating moves: the risk-of-bias assessment on this object is unadjudicated and every certainty rating is PENDING, and a corrected trial history is an input to that adjudication rather than a reason to re-rate tonight. Second, the claim that neither trial registered a time-to-first-event analysis remains true and is not withdrawn -- both registry records post the amended total-occurrence primary. Registered is not the same as planned.

Comparison with published syntheses

This review pools two trials on a total-occurrence composite reported as a hazard ratio. The question a reader needs answered is whether the published literature reports that quantity, and whether anyone -- them or us -- got it wrong. Both halves are reported below, with the number of records screened.

7 checks were applied and 4 came back clean, 0 found an error, 1 found something absent and 2 could not be settled at the layer available. The denominator is stated because a list of only the failures is not a finding, it is a selection.

Confirmations are listed in the same table, in the same detail, as errors. Across the cardiology topics reconciled so far the published literature has been implicated in none of them: where a defect existed it was ours. A comparison with room only for their errors could not have found that out.

Table 24. Comparison with published syntheses
Check	Verdict	What was found
Our SOLOIST-WHF input against the registry's posted result soloist-hr-vs-registry-results	CONFIRMED checked and clean	The registry's results section posts, on the primary outcome whose definition we read, a hazard ratio of 0.67 with a 95% interval of 0.52 to 0.85. This object stores 0.67 (0.52 to 0.85). Identical, and read from the registry's own analysis record rather than from the publication that reports it -- two independent surfaces agreeing. Quoted: “Hazard Ratio (HR) 0.67, CI 0.52 to 0.85” — ClinicalTrials.gov NCT03521934, resultsSection, primary outcome analysis
Our SCORED input against the registry's posted result scored-hr-vs-registry-results	CONFIRMED checked and clean	The registry posts a hazard ratio of 0.74 with a 95% interval of 0.63 to 0.88 on the same primary outcome. This object stores 0.74 (0.63 to 0.88). Identical. Quoted: “Hazard Ratio (HR) 0.74, CI 0.63 to 0.88” — ClinicalTrials.gov NCT03315143, resultsSection, primary outcome analysis
Whether the two trials counted the same events endpoint-definitions-identical	CONFIRMED checked and clean	Both registry records give the primary outcome measure in identical words, and both descriptions specify the total number of occurrences, first and potentially subsequent. This is the check SGLT2_HF failed -- there, two trials counted urgent heart-failure visits and two did not -- and here it comes back clean, on text read from the registry rather than inferred from a result sentence. Quoted: “Number of Total Occurrences of Cardiovascular (CV) Death, Hospitalizations for Heart Failure (HHF) and Urgent Visits for Heart Failure (HF)” — ClinicalTrials.gov NCT03521934 and NCT03315143, primary outcome measure
Whether the amended SCORED endpoint is the one our estimate belongs to scored-endpoint-was-changed	CONFIRMED checked and clean	SCORED's primary end point was changed during the trial. The registry record read here is the amended one, and the posted hazard ratio 0.74 sits on that amended outcome, which is the one we pool. The superseded original coprimary -- first occurrence of CV death, non-fatal myocardial infarction or non-fatal stroke -- is carried separately in this object at k=1 and is pooled with nothing. A reader who worried that an amended endpoint had been quietly substituted can see that both halves are recorded. This check is about SCORED because SCORED is the trial whose amended endpoint our estimate belongs ; it is not a statement that only SCORED changed. Both programmes changed their primary objectives -- see endpoint history correction 2026 08 26. Quoted: “The primary end point was changed during the trial to the composite of the total number of deaths from cardiovascular causes, hospitalizations for heart failure, and urgent visits for heart failure.” — SCORED publication; and the amended measure on the registry record
Whether any published synthesis reports the quantity we pool no-synthesis-reports-this-estimand	ABSENT the thing checked for is not there	112 records matched the query and all 112 were read at abstract level. One names both trials we pool and reports no pooled estimate of its own. The drug-specific meta-analysis pools relative risks over nine mixed-design studies. No located synthesis reports a hazard ratio for the total-occurrence composite, so this pooled value cannot be checked against an external number. That is a statement about the literature, not a claim of novelty. the screen of published syntheses.json, 112 of 112 records read
Whether the 107 syntheses that name no trial included ours what-the-abstract-layer-cannot-settle	UNRESOLVED could not be settled at the layer available	107 of 112 records do not name any trial we pool, and four have no indexed abstract. Abstract-level screening cannot establish what a synthesis pooled. Reporting '0 syntheses affected' from this would be absence of evidence dressed as evidence of absence -- the same error the first version of this topic's search record made, when it concluded no synthesis existed and one was found in a single step the same day. Settling it needs full text. the screen of published syntheses.json, tally
An unexplained precision gap in the drug-specific meta-analysis bantounou-precision-query	UNRESOLVED could not be settled at the layer available	Its printed interval 0.64 to 0.69 implies a standard error on the log scale of about 0.0192; our two-trial pool over 11,806 participants implies 0.0705. With 8,040 and 7,280 participants, an event in 10% of each arm gives 0.0485, 20% gives 0.0324, and even 30% in both arms gives 0.0247 -- all wider than printed. The strongest innocent explanation is plausible and is stated: if the pool counts total recurrent events rather than participants, counts can exceed the participant number and that precision is attainable. The abstract does not state the unit of analysis, so this is recorded for a reader and not asserted as an error. Full text or the PROSPERO record would settle it. Quoted: “sotagliflozin significantly reduced the risk of HF [n = 8 studies; RR = 0.66 (0.64, 0.69)]” — PMID 39257196, abstract

The syntheses reconciled against

Table 25. The syntheses reconciled against
39257196	Bantounou MA, et al. Meta-analysis of sotagliflozin, a dual sodium-glucose-cotransporter 1/2 inhibitor, for heart failure in type 2 diabetes. ESC Heart Fail 2025;12(2):968-979. It pools relative risks across nine studies of mixed design and reports neither trial's own primary -- a total-occurrence composite as a hazard ratio -- as a separate row. Its abstract does not name its included studies, so a trial-by-trial reconciliation against it cannot be done at this layer.
37868384	Sotagliflozin vs Dapagliflozin: A Systematic Review Comparing Cardiovascular Mortality. Cureus 2024. It names both trials we pool -- the only record in 112 that does -- but reports no pooled estimate of its own. It therefore supplies no number to and is counted as located rather than as agreeing.

Where the numbers differ, and why

Table 26. Where the numbers differ, and why
This review	HR 0.7171 (0.6246 to 0.8234) for the total number of occurrences of CV death, HF hospitalisation and urgent HF visit, k=2, both pivotal trials.
The published synthesis	RR 0.66 (0.64 to 0.69) for heart failure over 8 studies, and RR 0.73 (0.66 to 0.81) for MACE over 8 studies.
Why they differ	A different quantity (a relative risk of experiencing an event is not a hazard ratio for the total number of events), a different eligible set (nine mixed-design studies against the two pivotal trials), and a different question. Same direction, but direction agreement is worth little across estimands: a close answer to a different question is the most persuasive kind of wrong. Neither side is shown to be in error here, and that is the finding.

How they were identified: PubMed E-utilities, query and counts recorded in ssot/sotagliflozin-hf/appraisal/the screen of published syntheses.json, which carries a decision for every record. This layer is machine-screened at abstract level and the limitation is stated there and in the checks below.

Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio

Certainty of the evidence (GRADE)

Certainty: Pending

Started at high — both contributing studies are randomised controlled trials.

The stored derivation of this rating is not shown, because it ends in the level this review is withholding. It is held with the assessment and will be published with the adjudicated rating.

Table 27. Certainty of the evidence (GRADE)
Domain	Rating	Why it was rated that way
Risk of bias	pending	This domain is not final: two assessors read every contributing result independently, they disagree, and no adjudication has been performed, so this review holds no final risk-of-bias judgement for these results. Its recorded reasoning is kept with the assessment rather than shown here as a settled judgement.
Inconsistency	not serious	—
Indirectness	not serious	Not downgraded, and the reasoning is worth stating because the opposite call is tempting. The frozen question is the drug across its pivotal programme, and the evidence is exactly that programme: the population, the intervention, the comparator and the outcomes are the ones the question names, measured in the trials themselves. The population spread between the trials is real and large, but it is diversity within the question rather than a mismatch between the question and the evidence, and section 10.10.1 is where it belongs -- the uniformity table -- not here. An earlier version of this domain downgraded on the ground that a reader wanting a heart-failure answer would misread the pooled value. That is a risk of misreading, answered by labelling and by the scope block, and downgrading for it would be stricter than the Handbook while claiming the Handbook's support -- the same defect a gate leg caught on the companion review in this batch.
Imprecision	serious	—
Publication bias	not assessable	—

Time to the first occurrence of that same composite, as a hazard ratio

Certainty of the evidence (GRADE)

Certainty: Pending

Started at high — both contributing studies are randomised controlled trials.

The stored derivation of this rating is not shown, because it ends in the level this review is withholding. It is held with the assessment and will be published with the adjudicated rating.

Table 28. Certainty of the evidence (GRADE)
Domain	Rating	Why it was rated that way
Risk of bias	pending	This domain is not final: two assessors read every contributing result independently, they disagree, and no adjudication has been performed, so this review holds no final risk-of-bias judgement for these results. Its recorded reasoning is kept with the assessment rather than shown here as a settled judgement.
Inconsistency	not serious	—
Indirectness	not serious	Not downgraded, and the reasoning is worth stating because the opposite call is tempting. The frozen question is the drug across its pivotal programme, and the evidence is exactly that programme: the population, the intervention, the comparator and the outcomes are the ones the question names, measured in the trials themselves. The population spread between the trials is real and large, but it is diversity within the question rather than a mismatch between the question and the evidence, and section 10.10.1 is where it belongs -- the uniformity table -- not here. An earlier version of this domain downgraded on the ground that a reader wanting a heart-failure answer would misread the pooled value. That is a risk of misreading, answered by labelling and by the scope block, and downgrading for it would be stricter than the Handbook while claiming the Handbook's support -- the same defect a gate leg caught on the companion review in this batch.
Imprecision	serious	—
Publication bias	not assessable	—

Time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio

Certainty of the evidence (GRADE)

Certainty: Pending

Started at high — the contributing study is a randomised controlled trial.

The stored derivation of this rating is not shown, because it ends in the level this review is withholding. It is held with the assessment and will be published with the adjudicated rating.

Table 29. Certainty of the evidence (GRADE)
Domain	Rating	Why it was rated that way
Risk of bias	pending	This domain is not final: this review's RoB 2 assessment does not cover this outcome at all -- it assesses 4 result(s) across 2 other outcome(s) -- so the risk-of-bias domain here rests on no assessment recorded on this object. Its recorded reasoning is kept with the assessment rather than shown here as a settled judgement.
Inconsistency	not assessable	—
Indirectness	not serious	Not downgraded, and the reasoning is worth stating because the opposite call is tempting. The frozen question is the drug across its pivotal programme, and the evidence is exactly that programme: the population, the intervention, the comparator and the outcomes are the ones the question names, measured in the trials themselves. There is no population spread to weigh on this outcome: it holds one trial, and that trial's population -- diabetes with chronic kidney disease, heart failure not an entry requirement -- is stated on its row and is not averaged with anything. A reader must not read this rating as covering a heart-failure population, and that is answered by saying so here and on the row rather than by a downgrade.
Imprecision	serious	—
Publication bias	not assessable	—

In this section: Paper · Title · Abstract · Introduction · Methods — eligibility criteria · Methods — outcomes sought at every registered rank · Methods — synthesis · Results · Risk of bias in the included results · Certainty of the evidence · Comparison with published syntheses · Limitations · Data availability · Software availability · Note on registration · References · Figures · Trial characteristics · Not reported in this record

Paper

Each statement below cites the record it came from; those sources are listed at the end of every section. Where the record holds nothing, the section says so by name rather than being left out, so a procedure that was not carried out can be told from one that simply was never mentioned.

Title

Sotagliflozin against placebo across its pivotal cardiovascular outcome programme: three separate estimands, never merged -- two pooled within themselves, one reported from a single trial¹

Sources for this section (1)

title

Abstract

Question. In both phase 3 trials that supported sotagliflozin's approval, what is the effect of sotagliflozin compared with placebo on cardiovascular death and worsening heart failure -- counted as total events, counted as time to a first event, and on the three-component atherothrombotic endpoint -- with each of those questions answered separately?¹

Methods. estimates were pooled under a random-effects model with the REML estimator; and risk of bias was assessed with RoB 2 (Cochrane risk-of-bias tool for randomized trials).²

Results. Time to the first occurrence of that same composite, as a hazard ratio: 0.749 (0.664 to 0.845) across 2 trials; Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio: 0.717 (0.625 to 0.823) across 2 trials; and no statistical heterogeneity was detected (I-squared 0%), though heterogeneity cannot be reliably assessed with two trials.³

Sources for this section (3)

question
screening.eligibility, results.by_outcome, risk_of_bias.by_outcome
results.by_outcome

Introduction

Background. This review asks: In both phase 3 trials that supported sotagliflozin's approval, what is the effect of sotagliflozin compared with placebo on cardiovascular death and worsening heart failure -- counted as total events, counted as time to a first event, and on the three-component atherothrombotic endpoint -- with each of those questions answered separately? It pools 2 randomised trials, against placebo on top of background therapy. The outcomes pooled are Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio; Time to the first occurrence of that same composite, as a hazard ratio. One further outcome is reported without pooling: Time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio, and the reason is recorded with it.¹

Sources for this section (1)

manuscript.introduction

Methods — eligibility criteria

Eligibility turns on population, intervention and comparator: a phase 3 randomised trial of sotagliflozin against placebo in the pivotal cardiovascular outcome programme. It does not turn on which analysis a trial reported -- section 3.2.4 cautions that eligibility resting on reported outcomes invites outcome reporting bias, and this object carries three estimands precisely so that a trial is not admitted or refused according to which one it happens to print. What the estimand governs is which pool a value may enter, under section 10.9.¹

Sources for this section (1)

screening.eligibility

Methods — outcomes sought at every registered rank

Every registered rank -- primary, secondary and other -- was read for 2 of 2 contributing trials before any decision about which outcomes could be combined; each records the time it was read.¹

Sources for this section (1)

inputs.trials[].all_ranks_read_utc

Methods — synthesis

For total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio, a random model was fitted with the REML estimator.¹

For time to the first occurrence of that same composite, as a hazard ratio, a random model was fitted with the REML estimator.²

Intervals are 95% confidence intervals.³

Methodological decisions follow Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al (editors). Cochrane Handbook for Systematic Reviews of Interventions version 6.5.1. Cochrane, 2025., version 6.5.1, and the sections relied on are listed in the object rather than cited generically.⁴

Risk of bias was assessed with RoB 2 (Cochrane risk-of-bias tool for randomized trials) (22 August 2019 version, as reproduced in the Cochrane Handbook), following Higgins JPT, Savovic J, Page MJ, Elbers RG, Sterne JAC. Chapter 8: Assessing risk of bias in a randomized trial. Cochrane Handbook for Systematic Reviews of Interventions version 6.5, 2024.. A result, not a study -- Handbook 8.2: risk of bias is assessed for a specific result. The same trial is assessed twice here and lands differently: SOLOIST-WHF's total-event result is its registered primary and its first-event result was never registered at all. A study-level rating would have averaged those two into one number and lost the only distinction this assessment actually makes. 4 result-level assessments were made. A domain that cannot be judged from the registration and the published report is no information, never some concerns. A rating of some concerns with no explanation reads as a judgement against the trial, and neither of these trials has done anything to earn one.⁵

A ceiling applies to every risk-of-bias judgement in this review; it is stated with the assessment itself rather than summarised here.⁶

Certainty of evidence is rated per outcome rather than in a single review-level block. 3 outcome(s) are pending and carry no published rating: Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio; Time to the first occurrence of that same composite, as a hazard ratio; Time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio (not pooled). This certainty rating is pending, not rated. It rests on a risk-of-bias judgement that this review has not finalised: two assessors read every contributing result independently, they disagree, and no adjudication has been performed, so this review holds no final risk-of-bias judgement for these results. Cochrane rates certainty per outcome by aggregating the risk-of-bias judgements of the results contributing to it, so a certainty rating cannot be more final than the risk-of-bias assessment it reads. The dependency is not inferred: this object records the risk-of-bias domain as supported by “A RoB 2 assessment of every result contributing to this pool, added 2026-08-20. 1 of 2 results are HIGH overall, all of them on domain 5, selection of the reported result”, which is one assessor's count of results at HIGH risk of bias -- a count the second assessor's judgements do not contain, on any result. The rating this object currently records, and the steps behind it, are shown below as the work so far rather than as this review's answer. No level is published for this outcome -- not the recorded one, not a more cautious one, and not a midpoint -- because choosing one would be the judgement that has not been made.⁷

Sources for this section (7)

results.by_outcome.hfcv_total.model, results.by_outcome.hfcv_total.estimator_used
results.by_outcome.hfcv_first.model, results.by_outcome.hfcv_first.estimator_used
config.confidence_level
methodological_authority.reference
risk_of_bias.tool, risk_of_bias.by_outcome
risk_of_bias.ceiling
results.by_outcome.hfcv_total.grade, results.by_outcome.hfcv_first.grade, results.by_outcome.mace3_first.grade, risk_of_bias.by_outcome

Results

The pooled hazard ratio for total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit was 0.717 (0.625 to 0.823, 95% interval) across 2 trials, against placebo on top of background therapy.¹

The trials' results were closely consistent with one another: I-squared, the share of the variation between them that is more than chance alone would produce, was 0%; the estimated variance of the true effects between trials (tau-squared) was 0, the heterogeneity test statistic Q was 0.43, on degrees of freedom 1. Computed from Q and its single degree of freedom. With two studies this statistic has almost no power to detect heterogeneity, so a low value is consistent both with agreement and with not being able to tell, and it is reported with that caveat rather than as reassurance. The difference these two trials genuinely have is in their populations, and no statistic computed from two effect estimates can see it -- the uniformity table is the only place it is visible.²

The grounds for pooling are recorded as: One effect measure, one comparator, one estimand and one drug, across trials that share no participant and no control group. Where a later pool on this page rests on the same grounds, they are not restated.³

The pooled hazard ratio for time to the first occurrence of that same composite was 0.749 (0.664 to 0.845, 95% interval) across 2 trials, against placebo on top of background therapy.⁴

Time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio. Only one trial reports this outcome, so nothing is pooled, and the reason is stated rather than the outcome being quietly omitted: Only one contributing trial reports this estimand. The companion trial posts no three-component analysis at all: its nearest registry row is a total-occurrence composite of four components, adding hospitalization for heart failure to these three. An earlier version of this field called it the same three components, which was false of the record, and a gate leg was right to read the registry rather than this sentence.⁶

The trial that did report it found hazard ratio 0.84 (0.72 to 0.99) in NCT03315143.⁷

The full model output for every pooled outcome -- the call, the estimator, the heterogeneity statistics and the back-transformed interval -- is stored verbatim on this review object at `results.by_outcome.<outcome>.r_output.verbatim`. It is not reproduced as an Extended data section in this article; an earlier version of this sentence said it was, and no such section exists.⁸

Sources for this section (8)

outcomes[id=hfcv_total].name, results.by_outcome.hfcv_total.pooled, config.confidence_level, outcomes[id=hfcv_total].comparator
results.by_outcome.hfcv_total.heterogeneity.i2, results.by_outcome.hfcv_total.heterogeneity.tau2, results.by_outcome.hfcv_total.heterogeneity.q, results.by_outcome.hfcv_total.heterogeneity.df, results.by_outcome.hfcv_total.heterogeneity_status
results.by_outcome.hfcv_total.poolable_reason
outcomes[id=hfcv_first].name, results.by_outcome.hfcv_first.pooled, config.confidence_level, outcomes[id=hfcv_first].comparator
results.by_outcome.hfcv_first.heterogeneity.i2, results.by_outcome.hfcv_first.heterogeneity.tau2, results.by_outcome.hfcv_first.heterogeneity.q, results.by_outcome.hfcv_first.heterogeneity.df, results.by_outcome.hfcv_first.heterogeneity_status
outcomes[id=mace3_first].name, results.by_outcome.mace3_first.poolable_reason
results.by_outcome.mace3_first.per_trial
results.by_outcome

Risk of bias in the included results

Risk of bias was assessed with RoB 2 at the level of each reported result, by two assessors reading independently. Their judgements differ on 4 of the 4 results and no adjudication has been performed, so this review holds no final risk-of-bias judgement for these results. Assessor 1 recorded: 3 at high risk of bias, 1 at some concerns. Assessor 2 recorded: 2 at some concerns, 2 at no information. Both are shown in the risk-of-bias table; neither is the review's finding. The per-domain judgements behind them are recorded with the extracted data.¹

Sources for this section (1)

risk_of_bias.SECOND_ASSESSOR_2026_08_21.DISAGREEMENT_RATE, risk_of_bias.SECOND_ASSESSOR_2026_08_21.PER_DOMAIN, risk_of_bias.SECOND_ASSESSOR_2026_08_21.assessor_1, risk_of_bias.SECOND_ASSESSOR_2026_08_21.each_disagreement, risk_of_bias.SECOND_ASSESSOR_2026_08_21.how_it_was_asked, risk_of_bias.ceiling.statement, risk_of_bias.default_rule, risk_of_bias.tool

Certainty of the evidence

This certainty rating is pending, not rated. It rests on a risk-of-bias judgement that this review has not finalised: two assessors read every contributing result independently, they disagree, and no adjudication has been performed, so this review holds no final risk-of-bias judgement for these results. Cochrane rates certainty per outcome by aggregating the risk-of-bias judgements of the results contributing to it, so a certainty rating cannot be more final than the risk-of-bias assessment it reads. The dependency is not inferred: this object records the risk-of-bias domain as supported by “A RoB 2 assessment of every result contributing to this pool, added 2026-08-20. 2 of 2 results are HIGH overall, all of them on domain 5, selection of the reported result”, which is one assessor's count of results at HIGH risk of bias -- a count the second assessor's judgements do not contain, on any result. The rating this object currently records, and the steps behind it, are shown below as the work so far rather than as this review's answer. No level is published for this outcome -- not the recorded one, not a more cautious one, and not a midpoint -- because choosing one would be the judgement that has not been made.¹

Certainty of the evidence, by outcome, with every rating step²
Outcome	Certainty	k	Started at	Rating steps
time to the first occurrence of that same composite, as a hazard ratio	Pending	2	high	the steps taken so far, not a final rating: rated down for risk of bias and imprecision
total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio	Pending	2	high	the steps taken so far, not a final rating: rated down for risk of bias and imprecision
time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio	Pending	1	high	the steps taken so far, not a final rating: rated down for risk of bias and imprecision

Sources for this section (2)

risk_of_bias.by_outcome
results.by_outcome.hfcv_first.grade, results.by_outcome.hfcv_total.grade, results.by_outcome.mace3_first.grade

Comparison with published syntheses

Published syntheses were identified as follows. PubMed E-utilities, query and counts recorded in ssot/sotagliflozin-hf/appraisal/the screen of published syntheses.json, which carries a decision for every record. This layer is machine-screened at abstract level and the limitation is stated there and in the checks below.¹

Sources for this section (1)

published_comparison._how_identified

Limitations

Known limitation of the screen: No systematic database search was run and nothing here should be read as though one had been. The scope is a named two-trial programme, which is what the frozen question asks about, so the risk a search would address -- an eligible trial nobody named -- is small but not zero, and it is stated rather than assumed away. A separate and dated record of the prior-synthesis search that WAS run is staged beside this object.¹

Eligible trials that do not contribute: No trial is in this category. Both pivotal trials contribute to every estimand they report, and the one further programme trial screened is in the screening log because it reports none of them. This field is present and empty rather than absent, so a reader can tell the category was considered.²

This review does not claim no internal check is described as independent verification. A script that reads this object and reports it consistent has confirmed consistency and nothing else. Note also what this object could not get: no published synthesis reports either trial's own primary as a separate row, so no independent re-analysis checks these values.³

4 claim(s) previously made about this review were corrected and the corrections are retained on the object rather than overwritten.⁴

What this review does not give a clinician: the number of events in each arm, any absolute effect or number needed to treat and any measure of harm. A relative effect cannot be turned into the benefit or the risk facing one patient without them, so this review can support a judgement about what these trials found and not a judgement about what to do.⁵

Sources for this section (5)

screening.known_limitation
eligible_but_not_contributing.note
verification_basis.what_is_not_claimed
claims_corrected
results.by_outcome

Data availability

The underlying records are the 10 source(s) listed under References; each names the layer it was read at, so a reader can tell a registry record from a published report.¹

Sources for this section (1)

sources

Software availability

Analyses were computed under R version 4.6.0 (2026-04-24 ucrt); metafor 5.0.1.¹

Intervals are reported at the 95% level.²

Sources for this section (2)

results.by_outcome
config.confidence_level

Note on registration

No PROSPERO registration or protocol record is held on this object. That is not the same as knowing this review was not registered, and neither claim is made. It is recorded here as an absence so that a reader does not have to infer it from silence.¹

Sources for this section (1)

bookkeeping_2026_08_21.whether_this_review_was_prospectively_registered

References

Included studies, by registration¹
Trial	Registration	Registry	Read
SOLOIST-WHF	NCT03521934	ClinicalTrials.gov
SCORED	NCT03315143	ClinicalTrials.gov

Methods guidance and software, with what each is cited for²
Cited for	Reference
risk of bias in the included results	Sterne JAC, Savovic J, Page MJ, et al. RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ. 2019;366:l4898. doi:10.1136/bmj.l4898
the model fitted, as R version 4.6.0 (2026-04-24 ucrt); metafor 5.0.1	Viechtbauer W. Conducting meta-analyses in R with the metafor package. J Stat Softw. 2010;36(3):1-48. doi:10.18637/jss.v036.i03
the reporting guideline this manuscript is written against	Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. doi:10.1136/bmj.n71

Sources this review reads, with the layer each was read at³
Id	Layer	Source	Location
EMA_ZYNQUISTA	Regulator: EMA, authorisation for a different indication, withdrawn	Zynquista (sotagliflozin) epar medicine overview and the European Commission's public statement of withdrawal	https://www.ema.europa.eu/en/medicines/human/EPAR/zynquista
EMA_ZYNQUISTA_OVERVIEW	Regulator: EMA medicine overview	Zynquista (sotagliflozin) EPAR medicine overview	https://www.ema.europa.eu/en/medicines/human/EPAR/zynquista
FDA_LABEL_INPEFA	Regulator: FDA approved label	Inpefa (sotagliflozin) US prescribing information, clinical studies	https://www.accessdata.fda.gov/drugsatfda_docs/label/2023/216203s000lbl.pdf
FDA_REVIEW_216203	Regulator: FDA integrated review	Inpefa (sotagliflozin) NDA 216203 integrated review, including the statistical review's sensitivity and supportive analyses	https://www.accessdata.fda.gov/drugsatfda_docs/nda/2023/216203Orig1s000IntegratedR.pdf
OA_SCORED2021	Primary paper (abstract, open)	Bhatt et al. Sotagliflozin in Patients with Diabetes and Chronic Kidney Disease. N Engl J Med 2021.	https://pubmed.ncbi.nlm.nih.gov/33200891/
OA_SOLOIST2021	Primary paper (abstract, open)	Bhatt et al. Sotagliflozin in Patients with Diabetes and Recent Worsening Heart Failure. N Engl J Med 2021.	https://pubmed.ncbi.nlm.nih.gov/33200892/
PM_BANTOUNOU2025	Prior synthesis (abstract, open)	Bantounou et al. Meta-analysis of sotagliflozin, a dual sodium-glucose-cotransporter 1/2 inhibitor, for heart failure in type 2 diabetes. ESC Heart Fa	https://pubmed.ncbi.nlm.nih.gov/39257196/
PM_VADUGANATHAN2022	Prior synthesis (abstract, open)	Vaduganathan et al. SGLT-2 inhibitors in patients with heart failure: a comprehensive meta-analysis of five randomised controlled trials. Lancet 2022.	https://pubmed.ncbi.nlm.nih.gov/36041474/
REG_CTGOV	Registry with posted results	ClinicalTrials.gov results records for both trials	https://clinicaltrials.gov/study/NCT03521934
SEARCH_RECORD	Record of what the literature does not hold	Prior-synthesis search record for these two trials, 2026-08-10	https://pubmed.ncbi.nlm.nih.gov/

Sources for this section (3)

manuscript.references
manuscript.references
sources

Figures

Figure 1. Forest plot -- time to the FIRST occurrence of that same composite, as a hazard ratio. k = 2.¹

Figure 2 not drawn. k = 2. A funnel plot and its asymmetry tests have almost no power below about ten trials (Cochrane Handbook 13.3.5.4), so a funnel drawn here would invite a reading of asymmetry this evidence cannot support. It is declined rather than drawn, and this slot says so where the plot would have been.

Figure 2. Funnel plot -- time to the FIRST occurrence of that same composite, as a hazard ratio. k = 2.²

Figure 3. Forest plot -- total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio. k = 2.³

Figure 4 not drawn. k = 2. A funnel plot and its asymmetry tests have almost no power below about ten trials (Cochrane Handbook 13.3.5.4), so a funnel drawn here would invite a reading of asymmetry this evidence cannot support. It is declined rather than drawn, and this slot says so where the plot would have been.

Figure 4. Funnel plot -- total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio. k = 2.⁴

Figure 5. Forest plot -- time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio. k = 1.⁵

Figure 6 not drawn. k = 1. A funnel plot and its asymmetry tests have almost no power below about ten trials (Cochrane Handbook 13.3.5.4), so a funnel drawn here would invite a reading of asymmetry this evidence cannot support. It is declined rather than drawn, and this slot says so where the plot would have been.

Figure 6. Funnel plot -- time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio. k = 1.⁶

Sources for this section (6)

results.by_outcome.hfcv_first.per_trial, results.by_outcome.hfcv_first.pooled
results.by_outcome.hfcv_first.per_trial, results.by_outcome.hfcv_first.pooled
results.by_outcome.hfcv_total.per_trial, results.by_outcome.hfcv_total.pooled
results.by_outcome.hfcv_total.per_trial, results.by_outcome.hfcv_total.pooled
results.by_outcome.mace3_first.per_trial
results.by_outcome.mace3_first.per_trial

Trial characteristics

Characteristics of every trial contributing to this review¹
Registration	Trial	Participants	Randomised	Primary outcome measured over	Design
NCT03315143	SCORED	adults with type 2 diabetes, chronic kidney disease with an estimated glomerular filtration rate of 25 to 60, and cardiovascular risk; heart failure was not an entry requirement	10584	Up to 30 months	phase 3, double-blind, randomised one to one against placebo, stopped early because the sponsor withdrew funding, and with its primary endpoint changed during the trial.
NCT03521934	SOLOIST-WHF	adults with type 2 diabetes recently hospitalised for worsening heart failure, with the first dose given before discharge in about half and a median of two days after it in the rest	1222	Up to 21.9 months	phase 3, double-blind, randomised one to one against placebo, stopped early because the sponsor withdrew funding.

Sources for this section (1)

inputs.trials

Not reported in this record

The items below were not written because the record does not hold what they would be composed from. They are named here so that an absence is not mistaken for an omission.

Items not reported, and why
Article area	Item not reported	Reason
Abstract	The Background sentence of the abstract	Background is argument -- why this question matters -- and no arrangement of the quantities this object holds yields it. It is emitted only from an authored field and there is no composition path for it here
Abstract	The Conclusions sentence of the abstract	A conclusion is argument -- what a reader should now believe -- and it is emitted only from an authored field. Composing one from the estimate would be this renderer telling the reader what to conclude, which no field supports
Methods — search	A database search for primary trials, which was not carried out for this review; the included trials were identified from named registrations	No further reason is recorded.
Methods — synthesis	The model/estimator sentence for time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio	No pooled estimate was produced for this outcome, so no model was fitted and none is described
Methods — synthesis	The claim that records were screened in duplicate by two independent reviewers	No further reason is recorded.
Methods — synthesis	The claim that the review methods were prespecified before data collection	No further reason is recorded.
Note on registration	The registration note	No further reason is recorded.

Not held in this object. No statistical panel set is held in this object.

Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio

What population this estimate belongs to

This is a programme-level estimate in type 2 diabetes with cardiovascular or renal risk. It is not a general heart-failure estimate, and it is neither an HFrEF nor an HFpEF estimate. Both trials required type 2 diabetes. SOLOIST-WHF enrolled patients hospitalised for worsening heart failure; SCORED enrolled patients with chronic kidney disease and cardiovascular risk, in whom heart failure was not an entry requirement.

The two trials contribute very unequally, and the larger contributor is the one that did not require heart failure. Inverse-variance weights, derived from this object's own stored log standard errors: None 32%, None 68%.

The weights above are computed from `per trial[].se_log` on this outcome. No weight is quoted from any external summary.

The proportion of SCORED participants with heart failure at baseline is not stated on this page. The external review gives approximately 31%, and that figure is not in any source this object holds -- not the staged abstract, not the registry record, not the FDA integrated review extract. It is listed as extraction work rather than repeated here.

Time to the first occurrence of that same composite, as a hazard ratio

What population this estimate belongs to

The weights above are computed from `per trial[].se_log` on this outcome. No weight is quoted from any external summary.

Every number on this page is projected from a single canonical object, and each measured cell names the source analysis it was read from. That is what is machine-checked. It does not establish that the underlying sources are right, only that this page faithfully reports them.

Submission readiness: NOT READY

Qualifications that travel with the headline

Sotagliflozin against placebo across its pivotal cardiovascular outcome programme: three separate estimands, never merged -- two pooled within themselves, one reported from a single trial

Registration and administrative information (PROSPERO field set)

What this object claims to contain

SOTA-P-CARDIA NCT05562063

Extracted values, and where each came from

Correction — what this page said about the trials’ endpoint history

Risk of bias (RoB-2)

Overall judgement per trial

Inter-assessor agreement

References

Included studies

Regulatory documents

Methods and guidance

Knowingly not included

Sources

Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio

Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio

What was measured

Pooled result

What this pool holds constant

Contributing trials

Endpoint definitions, read from the registry

The methods rule governing this decision

Leave-one-out sensitivity

Does the answer depend on the pooling method?

Time to the first occurrence of that same composite, as a hazard ratio

Time to the first occurrence of that same composite, as a hazard ratio

What was measured

Pooled result

What this pool holds constant

Contributing trials

Endpoint definitions, read from the registry

The methods rule governing this decision

Leave-one-out sensitivity

Does the answer depend on the pooling method?

Time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio

Time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio

What was measured

Single cohort — no synthesis

Contributing trials

Endpoint definitions, read from the registry

The methods rule governing this decision

Leave-one-out sensitivity

Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio

Forest plot

Visual abstract

Risk of bias, both assessors

GOSH

Trial-sequential analysis

Meta-regression

Time to the first occurrence of that same composite, as a hazard ratio

Forest plot

Visual abstract

Risk of bias, both assessors

GOSH

Trial-sequential analysis

Meta-regression

Time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio

Forest plot

Risk of bias, both assessors

GOSH

Trial-sequential analysis

Meta-regression

Summary of findings

Reproducibility artifact

Other recorded qualifications

Comparison with published syntheses

The syntheses reconciled against

Where the numbers differ, and why

Total occurrences of cardiovascular death, hospitalization for heart failure and urgent heart failure visit, as a hazard ratio

Certainty of the evidence (GRADE)

Time to the first occurrence of that same composite, as a hazard ratio

Certainty of the evidence (GRADE)

Time to the first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke, as a hazard ratio

Certainty of the evidence (GRADE)

Paper

Title

Abstract