FormalPara Key Points

A total 26 randomized controlled trials with 16,090 patients and 161 tumour cases were included in our research.

Our study showed that once-weekly GLP-1RAs did not increase tumour risk compared with other therapies.

1 Introduction

Glucagon-like peptide-1 receptor agonists (GLP-1RAs) are a novel class of injectable antidiabetic drugs with multiple glucoregulatory effects, including enhancement of glucose-dependent insulin secretion, suppression of inappropriately elevated postprandial glucagon secretion and slowing of gastric emptying [1–3]. They have been the focus of much attention during the last years because of their unique mechanisms of action [4–6]. They are beneficial for blood glucose control; their other potential benefits include preservation of beta-cell function and improvements in other diabetes-related co-morbid conditions, such as hypertension, hyperlipidaemia and obesity [7–9]. The potentiation of insulin and glucagon release is glucose-dependent and is therefore associated with a low risk of hypoglycaemia [10–12].

Currently, several once-weekly GLP-1RAs have been approved by the US Food and Drug Administration (FDA) as adjunctive therapy for the management of type 2 diabetes [13–15]. Besides consistent glycaemic control, once-weekly GLP-1RAs offer some advantages over exenatide and liraglutide, including less frequent injections and improved treatment satisfaction. A few of studies indicate that the GLP-1RAs exenatide and liraglude might increase the occurrence of some specific cancers (e.g. pancreatic or thyroid cancer). Compared with liraglutide and exenatide, once-weekly GLP-1RAs have a longer half-life and continual action. Therefore it might be easier to obtain a definite result if we perform a study focused on once-weekly GLP-1RAs. Thus far, no study has evaluated the clinical safety of once-weekly GLP-1RAs with respect to tumour risk. Many randomized controlled trials (RCTs) have been published in PubMed, EMBASE and the Cochrane Library; however, most of these studies did not report findings related to tumour occurrence. Subsequently, we searched RCTs registered in ClinicalTrials.gov. Here, we report the findings of our informal meta-analysis on once-weekly GLP-1RAs and the risk of occurrence of tumours.

2 Methods

2.1 Search Strategy and Selection Criteria

We selected relevant studies registered in ClinicalTrials.gov up to 25 October 2015 using the following keywords: glucagon like peptide 1 receptor agonist OR exenatide OR albiglutide OR taspoglutide OR dulaglutide OR lixisenatide OR semaglutide OR CJC-1131 OR LY315902 OR CJC-1134-PC.

Two independent authors screened trials that could potentially be included in our study one by one. An RCT was considered eligible if the following criteria were met: (1) adult patients with type 2 diabetes were studied; (2) once-weekly GLP-1RAs and other treatments were compared; (3) the incidences of neoplasms (benign, malignant and unspecified) were reported as serious adverse events; (4) the duration of intervention was at least 12 weeks; and (5) there were more than 60 samples in each arm. Exclusion criteria were as follows: (1) observational and retrospective studies; (2) non-clinical studies; and (3) lack of information about the outcome that we analysed in this study.

2.2 Data Extraction

Two independent authors extracted the following data from each selected study: NCT number, study duration, trial sponsors, intention-to-treat (ITT) population, interventions and the number of individuals in the population without tumours.

2.3 Bias Assessment

Most of the RCTs included in our study were well designed and no obvious bias was founded in relevant published papers. However, all the data about tumour occurrence were extracted from ClinicalTrials.gov directly, rather than published papers, and it might be not appropriate for us to assess the risk of bias according to the information in ClinicalTrials.gov. Therefore, we performed this informal meta-analysis without assessment of risk of bias.

2.4 Statistical Analysis

We combined groups to create a single pairwise comparison when a trial contained multiple intervention groups. Outcomes were calculated as risk ratios (RRs) with 95 % confidence intervals (CIs) using the Mantel–Haenszel method and fixed-effects model. I 2 testing was performed to assess the magnitude of the heterogeneity between studies, with values greater than 50 % considered indicative of moderate-to-high heterogeneity [16]. To evaluate the influence of each study on the overall effect size, sensitivity analyses were conducted using the leave-one-out method, i.e. by removing one study at a time and then repeating the analysis. An inverse variance random-effects model was also used to further prove the robustness of the analysis results. Subgroup analyses were also performed according to study duration, tumour location and type of control groups.

We assessed funnel plot asymmetry using Egger tests and defined significant publication bias as p < 0.1. The trim-and-fill computation method was used to estimate the effect of publication bias on the interpretation of the results when publication bias was significant [17]. All these analyses were performed using RevMan 5.1 (Nordic Cochrane Centre) and Stata version 11 software (StataCorp LP, College Station, TX, USA).

3 Results

We identified 429 studies from the ClinicalTrials.gov registry, and 26 RCTs with 16,090 patients were included in our analysis [18–43]. The study selection process is shown in Fig. 1. Nine of them evaluated albiglutide [18–26], six evaluated exenatide extended-release (ER) [27–32] and 11 evaluated dulaglutide [33–43]. According to the types of control groups, eight trials compared once-weekly GLP-1RAs with a placebo [18, 20, 23, 24, 26, 28, 40, 41], seven with exenatide and liraglutide [20, 21, 27, 30, 33, 39, 42], seven with insulins [22, 25, 31, 34, 36, 37, 43] and eight with oral drugs [19, 23, 24, 28, 29, 32, 35, 38]. The mean duration of these trials was 64 weeks (range 16–156 weeks). All of these trials were supported by companies; other characteristics of these trials are shown in Table 1.

Fig. 1
figure 1

Trial selection process. GLP-1RAs glucagon-like peptide-1 receptor agonists

Full size image
Table 1 Characteristics of the randomized controlled trials included in the study
Full size table

In a pooled analysis of 26 RCTs, use of once-weekly GLP-1RAs did not result in an increase in tumour risk (RR 1.02; 95 % CI 0.74–1.41; p = 0.91) (Fig. 2) when compared with the use of other therapies [18–43]. Similarly, in the subgroup analysis according to the type of once-weekly GLP-1RAs, use of albiglutide [17–26], exenatide-ER [27–32] and dulaglutide [33–43] did not increase the risk for tumours (RR 0.81; 95 % CI 0.51–1.27; p = 0.63 for albiglutide; RR 1.54; 95 % CI 0.54–4.40; p = 0.72 for exenatide-LAR; RR 1.26; 95 % CI 0.75–2.12; p = 0.83 for dulaglutide). As the duration of trials might influence the pooled results, we performed a subgroup analysis for trials with a duration of ≥52 weeks. Seven trials that evaluated albiglutide and another seven trials that evaluated dulaglutide were included in this analysis. Once-weekly GLP-1RAs did not increase tumour risk in 52 weeks (RR 0.93; 95 % CI 0.65–1.33; p = 0.81) [18, 19, 22–26, 35–39, 42, 43]. No heterogeneity (I 2 = 0 %) or publication biases were noted (p = 0.110 for total, p = 0.169 for 52 weeks, p = 0.484 for albiglutide, p = 0.635 for exenatide-ER and p = 0.538 for dulaglutide) in these analyses. All of these results were robust in the sensitivity analysis and were not affected by any single study. A funnel plot for the total analysis is shown in Fig. 3.

Fig. 2
figure 2

Forest plot of risk ratio in total analysis. Exenatide-ER exenatide extended-release, CI confidence interval, M–H Mantel–Haenszel

Full size image
Fig. 3
figure 3

Funnel plot for the total analysis. Exenatide-ER exenatide extended-release, RR risk ratio, logRR natural logarithm of RR, SE standard error

Full size image

We also performed subgroup analyses according to the type of control groups and tumour location. Once-weekly GLP-1RAs did not increase the tumour risk compared with placebo (RR 0.72; 95 % CI 0.37–1.39; p = 0.98) [18, 20, 23, 24, 26, 28, 40, 41], exenatide and liraglutide (RR 1.73; 95 % CI 0.72–4.15; p = 0.71) [20, 21, 27, 30, 33, 39, 42], insulin (RR 1.38; 95 % CI 0.76–2.52; p = 0.76) [22, 25, 31, 34, 36, 37, 43] and oral antidiabetic drugs (RR 0.73; 95 % CI 0.44–1.22; p = 0.70) (Fig. 4) [19, 23, 24, 28, 29, 32, 35, 38]. No heterogeneity (I 2 = 0 %) or publication biases were noted (p = 0.837 for placebo; p = 0.392 for exenatide and liraglutide; p = 0.523 for insulin; p = 0.561 for oral antidiabetic drugs) in these analyses. All these results were robust in the sensitivity analysis and were not affected by any single study. The results of total and subgroup analyses using the random-effects model are shown in Supplementary Table 1.

Fig. 4
figure 4

Subgroup analysis according to control groups. CI confidence interval, GLP-1RAs glucagon-like peptide-1 receptor agonists, M–H Mantel–Haenszel

Full size image

Additionally, once-weekly GLP-1RAs did not increase the tumour risk in any tissue (Table 2). A total of five cases of pancreatic cancer were reported: two cases of metastatic pancreatic carcinoma, one treated with albiglutide [18] and the other with liraglutide [39]; one case of pancreatic carcinoma treated with dulaglutide [42]; one case of adenocarcinoma pancreas treated with oral drugs [23]; and one case of benign pancreatic neoplasm treated with albiglutide [19]. No heterogeneity (I 2 = 0) was noted in these analyses. All of these results were robust and were not affected by any single study. Significant publication biases existed only in the analyses of breast and lung tumours (p = 0.07 and p = 0.06, respectively). However, no trimming was performed in the ‘trim and fill’ analysis and the data remained unchanged, which suggests that publication bias might not affect these results significantly.

Table 2 Subgroup analysis according to tumour tissue
Full size table

4 Discussion

Although GLP-1RAs are a novel class of antihyperglycaemic agents, their safety with respect to tumour risk has attracted a high level of concern in the past 5 years. In 2011, the FDA examined the adverse events database of studies that investigated these treatments [44, 45]. This report indicated that GLP-1RAs increase the risk for pancreatitis and raised caution about the potential long-term actions of these drugs in promoting pancreatic cancer. However, the FDA report had many issues. The FDA Adverse Event Reporting System database does not provide information regarding obesity, smoking habits, alcohol consumption or chronic pancreatitis, which are well-established additional risk factors for pancreatic cancer. At the same time, many fundamental studies had been performed to evaluate the safety of GLP-1RAs with respect to the risk of pancreatic, thyroid, prostate, colon and breast cancer [46–51]. However, there was neither firm evidence in favour of this hypothesis nor evidence strong enough to rule out the possibility of increased risk based on the results available at present.

Our systematic review also has many limitations. First, because the description of neoplasms in some trials was not sufficiently detailed and we included all types of tumours, namely benign, malignant and unspecified neoplasms. Second, it was unclear whether the neoplasms were present before treatment and we could not determine if there was a link between the neoplasms and treatment. Third, the duration of these randomized controlled trials was short for evaluating cancer risk, especially that of trials that evaluated exenatide-ER, with most being conducted for about 6 months. Besides, even though a total of 16,090 patients were included in our analysis, this sample size is still insufficient to evaluate the risk of site-specific cancer (e.g. pancreas or thyroid). Finally, only 149 tumour cases were reported in both the GLP-1RAs and control group and the small number of tumour cases decreased the reliability of our final conclusion.

To our knowledge, this is the first article that systematically evaluated the clinical safety of once-weekly GLP-1RAs with respect to tumour risk in type 2 diabetes patients. Our study showed that once-weekly GLP-1RAs did not increase tumour risk compared with other treatments and this result was independent of the type of once-weekly GLP-1RA administered and the treatment duration. Subgroup analyses performed according to the tumour tissue and types of control groups also revealed similar results.

5 Conclusion

We can conclude that compared with other treatments, once-weekly GLP-1RAs do not increase the risk of tumour occurrence and this is independent of the type of GLP-1RAs and treatment duration. However, there were limitations to our study, and further larger sample, long-term clinical trials should be conducted in future to confirm our results.