The placebo response of injectable GLP-1 receptor agonists vs. oral DPP-4 inhibitors and SGLT-2 inhibitors: a systematic review and meta-analysis

Abstract

Aims: The size of the placebo response in type 2 diabetes (T2DM) treatment and its relation to the route of drug administration have not been systematically reviewed. We aimed to determine weight loss, change in HbA1c and incidence of adverse events after treatment with injectable placebo GLP-1 receptor agonist (GLP-1ra), compared with oral placebo DPP-4 inhibitor (DPP-4i) and placebo SGLT-2 inhibitor (SGLT-2i).

Methods: PubMed, EMBASE and Central were searched up to September 2014 for randomized placebo controlled trials investigating GLP-1ra, DPP-4i or SGLT2-i. Data on placebo groups were extracted and pooled using a generic inverse variance random effects model.

Results: Sixty-seven trials were included, involving 2522, 5290 and 2028 patients randomized to placebo GLP-1ra, placebo DPP-4i and placebo SGLT-2i, respectively. Body weight decreased by -0.67 kg (95% CI -1.03, -0.31) after treatment with placebo GLP-1ra (-0.76 kg [95% CI -1.10, -0.43] with placebo short acting GLP-1ra and -0.32 kg [95% CI -1.75, 1.10] with placebo long acting GLP-1ra) and by -0.31 kg (95% CI -0.64, 0.01) with placebo DPP-4i (P = 0.06 for difference with placebo short acting GLP-1ra). Placebo SGLT-2i resulted in an intermediate -0.48 kg (95% CI -0.81, -0.15) weight loss. Weight loss with placebo showed a strong correlation with the active comparator drug (r(2) = 0.40-0.78). HbA1c changed little with placebo treatment (-0.23%, 0.10% and -0.13% for placebo GLP-1ra, DPP-4i and SGLT-2i). Adverse events occurred frequently with placebo, were often similar to the active comparator drug and led to drop-out in 2.0-2.7% of cases.

Conclusions: The response to placebo treatment was related to its active comparator, with injectable placebo GLP-1ra showing a relevant response on weight, whereas oral placebo DPP4i showed no significant response. These findings may suggest that subjective expectations influence T2DM treatment efficacy, which can possibly be employed therapeutically.

Keywords: meta-analysis; placebo effect; placebo response; type 2 diabetes.

Figure 1

Flow diagram of the study selection process. OAD oral antidiabetic agent, TZD thiazolidinedione, SU sulphonylurea, RCT randomized controlled trial

Figure 2

Pooled‐effects summary of mean ± SEM change in body weight (A) and HbA1c (B) with placebo. GLP‐1ra GLP‐1 receptor agonist, DPP‐4i DPP‐4 inhibitor, SGLT2‐i SGLT‐2 inhibitor. () placebo long acting GLP‐1ra; () placebo short acting GLP‐1ra; () placebo GLP‐1ra; () placebo DPP‐4i; () placebo SGLT‐2i

Figure 3

Forrest plot of change in body weight (A) and HbA1c (B) with placebo. Forrest plot representing pooled‐analyses of change in (A) body weight (kg) and (B) HbA1c (%), using a generic inverse variance random effects model

Figure 4

Scatterplot representing the relationship between the response to placebo and its active comparator drug on body weight (A) and HbA1c (B). GLP‐1ra GLP‐1 receptor agonist, DPP‐4i DPP‐4 inhibitor, SGLT2‐i SGLT‐2 inhibitor. A) () GLP‐1ra (r ² = 0.40); () DPP‐4i (r ² = 0.78); () SGLT‐2i (r ² = 0.61); B) () GLP‐1ra (r ² = 0.42); () DPP‐4i (r ² = 0.50); () SGLT‐2i (r ² = 0.07)