Meta-analyses, first author, year | Analysis included only RCTs in which randomized groups differ by DPP-4 inhibitor treatment to avoid the confounding effect of other medications | Statistically compared HF outcomes between different DPP-4 inhibitors | Additional ≥ 24-week follow-up DPP-4 inhibitor v. placebo RCTs/enrolled patients with HF events included in the current meta-analysis | Avoidance of inadvertent double counting of some included RCTs | Inclusion of all* HF outcomes for EXAMINE Trial7 | Inclusion of most recently published HF results for VIVIDD Trial113 | Main conclusions |
---|---|---|---|---|---|---|---|
Including TECOS | |||||||
Current meta-analysis | Yes | Yes | (reference) | Yes | Yes | Yes | •13% increase in HF risk only statistically significant (p = 0.03) if results of smaller RCTs added to large cardiovascular safety RCTs •differences between agents not statistically significant (interaction p = 0.07-0.12) |
Li, 2016151 | No | No | 10 RCTs/ 5541 patients | Yes | No | Yes | •12% increase in HF risks¶ (p = 0.05) pooling HF hospitalization outcomes from 5 RCTs only; no significant increase in HF for the remaining RCTs v. all comparators |
Abbas, 2016149 | Yes† | No | 29 RCTs/ 18 097 patients† | Yes | No | -‡ | •non-significant 11% increase in HF risk (p = 0.19) pooling only the 3 large cardiovascular safety RCTs but not including all HF outcomes for EXAMINE |
Kongwat-charapong, 2016150 | No | No | 4 RCTs/ 1639 patients | Yes | Yes | No | •non-significant 11% increase in HF risk (p = 0.06) v. all comparators •highlighted increase in HF for saxagliptin but differences not statistically compared with other agents |
Pre-TECOS | |||||||
Monami, 2014152 | No | No | 8 RCTs/ 17 463 patients | No (double counted NCT0102839176 which was an extension of NCT0039763177) | No | -‡ | •19% increase in HF odds (p = 0.015) v. all comparators •highlighted increase in HF for saxagliptin but differences not statistically compared with other agents |
Wu, 2014154 | No§ | No | 15 RCTs/ 19 339 patients | No (double counted 2 publications for NCT0032701555,56) | No | No | •16% increase in HF risk (p = 0.04) v. all comparators and 17% increase (p = 0.03) v. only placebo comparators |
Savarse, 2015153 | No | No | 9 RCTs/ 18 055 patients | No (double counted NCT0091577237 which was an extension of NCT0079816136) | No | No | •16% increase in HF risk (p = 0.03) v. all comparators pooling long-term follow up RCTs but no increase pooling short-term follow up RCTs |
Note: DPP-4 = dipeptidyl peptidase-4, HF = heart failure, RCT = randomized controlled trial.
*Some of the previously published meta-analyses149,151-154 included only heart failure hospital admissions that were counted in the analysis of the composite end point reported in the abstract of the follow-up publication for EXAMINE focusing on heart failure outcomes,7 instead of all admissions for heart failure reported in the main body of this publication. Including all hospital events results in a higher risk for alogliptin for this RCT (RR 1.18 v. RR 1.07).
†Abbas 2016149 only included the 3 large cardiovascular RCTs (SAVOR-TIMI 53,4,5 EXAMINE6,7 and TECOS8).
‡VIVIDD113 was not included in Monami 2014152and Abbas 2016.149
§For Wu 2014,154 comparison to placebo trials was included as a secondary analysis.
¶Li 2016151 actually reported 13% increase in odds using Peto odds ratios (p = 0.05), which corresponds to a 12% increase in risk using risk ratios (p = 0.05).