In this pooled analysis of 9156 patients with T2DM from 20 phase 2 and 3 clinical trials, treatment with saxagliptin was not associated with an increased risk of CV events and heart failure compared with placebo or active comparator. These results expand on previous findings on the CV safety of saxagliptin reported in a meta-analysis of 8 phase 2 and 3 trials . In that analysis, a total of 40 MACE events in 4607 patients were reported. The relative risk (95% CI) for saxagliptin versus comparator for a composite endpoint of adjudicated CV death, MI, and stroke was 0.43 (0.23, 0.80), which suggested possible CV protection with saxagliptin. The present analysis expanded on the previous study and included 9156 patients who experienced 74 MACE events. In this larger population, which should provide a more precise risk estimate, the relative risk (95% CI) for a composite endpoint of adjudicated CV death, MI, and stroke was 0.75 (0.46, 1.21), suggesting no increased risk of MACE in this 20-study pool. Incidence rates for CV events for saxagliptin were not different from those for placebo or comparator in most analyses, with the exception of the lower IR for MACE in the saxagliptin 2.5-mg group in the subanalysis of the 20-study pool. However, it should be noted that only 7 of the 20 studies included patients who had received the 2.5-mg saxagliptin dose.
The present findings are also consistent with previously published meta-analyses of CV events from clinical trial programs for other DPP-4 inhibitors, including vildagliptin, sitagliptin, linagliptin, and alogliptin. In a pooled analysis of 25 clinical trials, the relative risk (95% CI) for cardiocerebrovascular events for vildagliptin was 0.88 (0.37, 2.11) for 50 mg once daily and 0.84 (0.62, 1.14) for 50 mg twice daily . In other meta-analyses, the IRR or HR (95% CI) for CV-related events versus comparators was 0.83 (0.53–1.30) for sitagliptin , 0.34 (0.16, 0.70) for linagliptin , and 0.64 (1-sided 97.5% CI, 0.0, 1.406) for alogliptin . In addition, a meta-analysis of 70 trials of DPP-4 inhibitors enrolling 41,959 patients reported a reduction in MACE (n = 495 total events of CV death, nonfatal MI, and stroke and acute coronary syndromes and/or heart failure; odds ratio, 0.71 [95% CI, 0.59, 0.86]) . Although these studies are not directly comparable because of different CV endpoints, study designs, adjudication procedures, patient populations and background medication, all supported the hypothesis that DPP-4 inhibitors do not increase CV risk and may possibly have CV benefits in patients with T2DM.
Results from the large outcome trial of saxagliptin in patients with prior CV disease or multiple CV risk factors (SAVOR) have recently been reported . Results generally consistent with those were also reported from the alogliptin trial (EXAMINE) in patients after acute coronary syndrome . SAVOR demonstrated neutrality on the composite primary endpoint of CV death, MI, or ischemic stroke (HR, 1.00 [95% CI, 0.89, 1.12]). The MACE results reported here in a much lower-risk population with an event rate approximately a third of that observed in SAVOR are consistent with SAVOR in demonstrating a safe profile of saxagliptin with respect to MACE events. The fact that SAVOR did not demonstrate superiority compared with placebo raises at least two alternative, though not mutually exclusive, interpretations: (1) evidence suggesting benefit from meta-analysis and preclinical evidence [16, 54] was due to chance or (2) saxagliptin and likely other DPP-4 inhibitors are safe in all populations and trends to benefit occur only in the lower-risk general population studied in the phase 3 clinical development program. The latter hypothesis has been previously suggested based on the only positive interaction of subgroups in a patient level meta-analysis of UKPDS, ACCORD, ADVANCE, and VADT . Owing to the marked difference in population characteristics (eg, age, CV history and risk factors, duration of diabetes, background diabetes and CV medications, proportion of patients with baseline glycated hemoglobin <7%) and population risk (3- to 6-fold higher event rate) between SAVOR and EXAMINE and the meta-analyses of phase 3 programs of saxagliptin and alogliptin, it is difficult to support or dismiss either interpretation for the lack of benefit observed in SAVOR and EXAMINE.
SAVOR also demonstrated neutrality on the broader composite endpoint of CV death, MI, stroke, or hospitalization for unstable angina, heart failure, or coronary revascularization (HR, 1.02 [95% CI, 0.94, 1.11]). One component of this broader endpoint, hospitalization for heart failure, did have an HR with 95% CI which did not include 1 (HR, 1.27 [95% CI, 1.07, 1.51]). As reported here, heart failure in the 20-study pool had an HR (95% CI) of 0.55 (0.27, 1.12). Again, differences in the patient population, background medications, and/or chance may be involved in the relative inconsistency of these results. Moreover, SAVOR was an event-driven trial in a highly defined population (prior CV disease or multiple CV risk factors), whereas the 20 clinical trials analyzed in this study had defined treatment periods ranging from 4 to 206 weeks and included diverse patient populations with T2DM (eg, patients who were treatment naïve, receiving varying background antihyperglycemic medications, or with renal impairment). The phase 3 data presented in this manuscript suggest that the observation of hospitalization for heart failure could not have been anticipated based on the phase 3 development program. It may be that further analysis of SAVOR results or the other prospective CV outcome trials with DPP-4 inhibitors [56, 57] will give further clarity to the two issues raised here.
Certain limitations of this analysis should be recognized and considered when interpreting the results. To handle missing data as the result of premature discontinuation, analysis methods assumed similar event rates had the patient completed the study. However, patients treated with saxagliptin tended to be followed longer and had a lower rate of discontinuation compared with those who received control treatment. Results using this assumption should be interpreted with caution.
Both groups received a range of background medications, including metformin, sulfonylureas, and thiazolidinediones and the control group received both active medications and placebo. In several studies, a titration of background medication [22, 41, 46–48] or a titration of double-blind saxagliptin [19, 29, 30, 39, 40] was permitted. In addition, in the majority of studies, rescue medication was permitted [19–23, 32, 36, 38–43]. These factors complicate interpretation of the findings.
The saxagliptin group was also heterogeneous and included patients treated with doses higher than the approved 2.5- and 5-mg once-daily doses. Further, the analyses of the 2.5- and the 5-mg doses used distinct study pools because not all studies included 2.5- and 5-mg arms, which precludes direct comparison of results for the 2 doses. It is important to recognize that the pooled patient population in these clinical trials was highly selected, which may have resulted in a lower event rate compared with that observed in clinical practice. Finally, there was relatively limited experience beyond 18 months.