Real-world effects of liraglutide versus dulaglutide in Japanese people with type 2 diabetes

Glucagon-like peptide-1 (GLP-1), an incretin hormone secreted by L-cells in the distal ileum and colon¹, has various effects, including glucose-dependent enhancement of insulin secretion², inhibition of glucagon secretion³, appetite suppression⁴, and suppression of gastric emptying⁵. Various GLP-1 receptor agonists (GLP-1 RAs) are available in the market at present. GLP-1 RAs can effectively lower blood glucose levels and facilitate weight loss with a low risk of hypoglycemia; hence, they are commonly used in patients with type 2 diabetes mellitus (T2DM). Furthermore, some GLP-1 RAs are reported to reduce the risk of 3-point major adverse cardiovascular events (MACE)^6,7, and a meta-analysis of four large-scale clinical studies that assessed GLP-1 RAs concluded that these agents can reduce the risk of cardiovascular events and all-cause mortality⁸.

Controlled clinical trials have demonstrated that GLP-1 RAs have distinct effects in terms of lowering glycated hemoglobin (HbA1c) levels and facilitating weight loss, as well as adverse reactions, such as nausea, vomiting, and diarrhea^8,9. In Japan, dulaglutide and liraglutide are the most widely used GLP-1 RAs in routine clinical practice¹⁰. In a randomized controlled trial (RCT), dulaglutide (up to 1.5 mg/week) was reported to be non-inferior to liraglutide (up to 1.8 mg/day)¹¹. In a Japanese comparative study of these agents used at different doses, HbA1c-lowering effects after 52 weeks of treatment were greater with dulaglutide (up to 0.75 mg/week) than liraglutide (up to 0.9 mg/day)¹². However, the above study evaluated patients who were not taking antidiabetic drugs or had discontinued antidiabetic monotherapy, thereby differing greatly from real-world clinical situations. In a real-world setting, only a few studies have investigated the effectiveness of liraglutide and dulaglutide in patients with T2DM, including those treated concomitantly with more than one oral hypoglycemic agent and insulin. To establish real-world evidence, we retrospectively examined changes in HbA1c after 12 months of GLP-1 RA treatment and compared liraglutide and dulaglutide. In this study, we used stabilized propensity score-based inverse probability of treatment weighting (IPTW) in order to reduce selection bias and confounding by observed covariates.

The present study compared the efficacy and safety of GLP-1 RAs, liraglutide and dulaglutide, in Japanese patients with T2DM in a real-world setting. We used stabilized propensity score-based IPTW in order to reduce selection bias and confounding by observed covariates. In a previously published RCT in Japanese patients, dulaglutide had greater HbA1c-lowering effects than liraglutide¹². However, the use of other drugs and patient characteristics in that RCT differed from those in real-world clinical practice. The present study included several patients who used more than one oral glucose-lowering agents, unlike previously reported RCT in Japanese patients. In this study using stabilized propensity score-based IPTW, the change in HbA1c level was lower in the liraglutide treatment group at 6 months, but there was no difference in the HbA1c level at 12 months, indicating similar HbA1c-lowering effect by the two agents.

Our study identified three characteristics that were associated with changes in HbA1c over the 12-month treatment period (Table 3). First, baseline HbA1c; patients with high baseline HbA1c levels were more likely to show improved post-treatment HbA1c levels. Next, GLP-1 RA treatment modality; compared with ‘add-on’, the rate of improvement in HbA1c at 12 months was modest in ‘switch’ and ‘reduce’. Similar findings have been reported following liraglutide treatment¹³. In clinical practice, treatment is often switched from dipeptidyl peptidase-4 inhibitors to GLP-1 RAs. In such cases, it is necessary to consider that any improvement in HbA1c after 12-month treatment would be smaller than that after the addition of GLP-1 RAs. Third, initiation of GLP-1 RA treatment during hospitalization was more effective in lowering HbA1c levels than applying the same treatment at the outpatient setting. Although patients who started GLP-1 RA treatment during admission to the hospital had higher baseline HbA1c levels than those who underwent outpatient initiation (9.7% [SD: 1.5] vs. 8.3% [SD: 1.4], P < 0.001), their HbA1c levels at 12 months were lower (7.1% [SD: 1.4] vs. 7.6% [SD: 1.5], P = 0.032). This result is quite interesting, but it is likely that unmeasured confounding factors and/or differences in follow-up care patterns contributed to this effect, and therefore, further investigation is warranted in the future.

Several studies reported the weight-reducing effects of liraglutide and dulaglutide, with weight loss of 2.3 kg (95% confidence interval [CI] 2.0–2.5) in one study⁶ and 1.46 kg (95% CI 1.25–1.67) in another¹⁴. However, in the AWARD-6 study using the same two agents, liraglutide (up to 1.8 mg/day) was significantly more effective in reducing body weight than dulaglutide (up to 1.5 mg/week)¹¹. Furthermore, in a Japanese phase III clinical study, body weight reduction by dulaglutide was negligible after 26 weeks of treatment¹⁵. Our results also showed that liraglutide was more effective in reducing body weight than dulaglutide. These results may partly explain why liraglutide is commonly used in obese diabetics in real-world clinical practice.

GLP-1 RAs are reported to help reduce HbA1c, body weight, and insulin dose among insulin users^16–18. Although there are no studies that directly compared changes in insulin dose after GLP-1 RA initiation (liraglutide vs. dulaglutide), our results showed total and bolus insulin dose significantly decreased in liraglutide and dulaglutide, no significant differences were observed between the two agents.

Several reports compared the real-world use of GLP-1 RAs, but most comparisons were restricted to treatment maintenance^19,20 and cost-effectiveness²¹. In contrast, our study examined changes in HbA1c levels. A real-world study from Taiwan²² showed there was a statistically significant change in HbA1c at 12 months from baseline in each treatment group (dulaglutide: − 1.06% [SD: 1.70] and liraglutide: − 0.83% [SD: 1.61]), with a significant between-group difference in HbA1c reduction of − 0.23% (95% CI − 0.38 to − 0.08%). A study from United States also showed that treatment with dulaglutide significantly reduced HbA1c compared with liraglutide²³. These results are different from our study, probably due to differences in drug dosage; the dose of dulaglutide in about 40% of the patients of the above study was 1.5 mg/week.

Few studies have investigated the real-world use of GLP-1 RAs in Japanese patients. Although more than 900 patients were enrolled in the JDDM-57 study¹⁰, HbA1c was analyzed after only 6 months of treatment with GLP-1 RAs. Despite the small sample size, the present study examined the contributing factors to changes in HbA1c induced by GLP-1 RAs in Japanese patients. We anticipate our data to help establish real-world evidence for the role of these factors in GLP-1 RA treatment.

Our study has several limitations. First, the sample size was relatively small, as described earlier, and the study was conducted in patients at a single institution treated by diabetologists; therefore, the results may not be generalizable to the entire population. However, since diabetologists often initiate the administration of GLP-1 RAs in Japan, our findings are considered plausible as real-world data at least in Japan. Second, the plasma glucose data were somewhat unreliable because not all blood samples were collected in a fasting state. However, evaluation of HbA1c, which was the primary endpoint, should be sufficient for the purpose of this study. Third, since liraglutide was the first to be launched in the Japanese market, there was a selection bias that dulaglutide could not be used in patients enrolled early in this study. In addition, the approved maximum dose of liraglutide in Japan under the public health system is 1.8 mg/day, and thus a higher dose of liraglutide was prescribed only in a few patients in the present study. Therefore, it is important to update this study in future investigations to establish more-up-date real-world evidence. Finally, propensity score is only appropriate when the strongly ignorable treatment assignment assumption is satisfied²⁴. However, there is no reliable way to test this assumption. Conventionally, consideration is given to C-statistics and propensity score-adjusted estimates of variables, but these also have no standard criteria. In this study, the estimates of the propensity score-adjusted estimates of variables are very close in each group compared to the pre-adjustment, although some variables remained significant, but it is uncertain whether these sufficiently satisfied the strongly ignorable treatment assignment assumption.

In conclusion, we have demonstrated in the present retrospective study the presence of significant differences in the characteristics of patients treated with liraglutide or dulaglutide in a real-world setting, implying that GLP-1 RAs were selected according to individual patient characteristics. Even after reducing the selection bias and confounding using stabilized propensity score-based IPTW between liraglutide and dulaglutide treatments, the effects of these two agents on HbA1c levels after 12 months of treatment were comparable. Our results also suggest that baseline HbA1c level, GLP-1 RA treatment modality, and inpatient initiation of GLP-1 RA treatment may be associated with reduction in HbA1c levels in a real-world setting.