Heterogeneity, MCID, and missed trials: comments on “efficacy and safety of glucagon-like peptide-1 receptor agonists in Parkinson’s disease: A systematic review and meta-analysis”

Dear Sir,

We read with great interest Singh et al.^[1] on Efficacy and safety of glucagon-like peptide-1 (GLP-1) receptor agonists (RAs) in Parkinson’s disease (PD): A systematic review and meta-analysis. Efficacy and safety of GLP-1 RAs in PD: A systematic review and meta-analysis. This study was pre-registered with PROSPERO, strictly followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 guidelines, assessed risk of bias (RoB) using RoB-2, and assessed the quality of evidence using GRADE. A total of 5 randomized controlled trials (RCTs) with a total of 570 participants were included from 1126 studies, providing a broader perspective on GLP-1 RA effects in PD.^[2] We acknowledge that the NLY01 trial was included by Singh et al., but the search strategy itself may not have been optimally designed to capture all relevant agents such as dulaglutide, tirzepatide, and others.^[1]

The core problem that reduces the credibility of the conclusion. First, regarding the search strategy, while the authors retrieved a substantial number of records, the final inclusion of only five RCTs suggests potential limitations in search sensitivity. The original article retrieved 1126 articles but only included 5 RCTs, which is very limited if only RCTs are available at this stage. The sparse number of included articles may be due to the irregularity of search queries and the reduction of synonym searches, and the search queries in Supplementary Material 2 only cover 4 databases, omitting a variety of commonly used GLP-1 RAs (e.g., dulaglutide, albiglutide, tirzepatide, benaglutide, NLY01) and PD synonyms (e.g., Parkinsonism). Emtree controlled words are not used in Embase, and the Boolean logic nesting is not clear enough. According to the PRESS 2015 guidelines,^[3] such defects can impair retrieval sensitivity and affect reproducibility. While the authors correctly reported the non-significant pooled mean difference (MD) under the random-effects model (MD = −2.00, P = 0.08) in their abstract, they subsequently emphasized the significant result from the fixed-effect model (P = 0.002) within the body text to support claims of efficacy. Given the considerable heterogeneity, the fixed-effect model is unlikely to be appropriate, and reliance on its result is misleading.

Second, a major concern is the substantial statistical heterogeneity observed in the primary outcome (Unified PD Rating Scale [UPDRS]-III, I² = 67%) and secondary outcomes (e.g., UPDRS-II, I² = 85%). Similar or even higher heterogeneity is also seen in UPDRS-II (I² = 85%). The authors did not do further subgroup or meta-regression to explore sources of heterogeneity (e.g., drug class, dose, and course of treatment).

Third, the clinical significance of the findings requires careful interpretation. The minimum clinically important improvement in (movement disorder society-UPDRS-III was −4.83 points, and the minimum clinically important deterioration was 4.38 points,^[4] while the pooled effect of the original article was only −2.0 points, which was not enough to meet the clinical benefit threshold.

Fourth, safety data reporting could be more refined. We found that the authors mentioned an interesting data that “GLP-1 RAs increase the risk of weight loss by 2.49 times.” Unexpected weight loss may exacerbate frailty and sarcopenia in PD patients,^[5] and strengthening discussions in this area would enhance the feasibility of the current study.