Study selection
The systematic literature search yielded 404 records (Supplementary Table S1). After the titles, abstracts, and duplicates were screened, 15 articles were considered potentially relevant. Of these, ten articles were excluded because of non-extractable data (n = 3), lacked assessment on the prevention of CIPN (n = 3), absence of appropriate treatment group (n = 2), a review article (n = 1), and a piece of scientific news (n = 1). Thus, four RCTs [16, 17, 27, 28] and one retrospective cohort study [15] were included after full-text review for further analysis. The flowchart of the search strategy and study selection is illustrated in Fig. 1.
Characteristics and quality assessment of included studies
The general characteristics of all included studies are summarized in Table 1. These five studies involved 868 patients (413 patients in the GM1 group and 455 patients in the control group). All of them were conducted in China, and published between 2012 and 2020. Sample size, types of cancer, follow-up duration were well-balanced and comparable in each study between the intervention and control group (Table 1). Among the four included RCTs, two [16, 17] were of good quality, one [27] was of poor quality, and one [28] was of fair quality (Fig. 2A); the quality of the retrospective study [15] included in our analysis was assessed as good (Fig. 2B).
Primary endpoints
Incidence of CIPN (CTCAE)
Four studies [15-17, 28] provided data regarding the incidence of grade ≥2 CIPN using the CTCAE. Of the four studies, three studies [15, 16, 28] used the oxaliplatin regimen, only one study [17] used the taxanes regimen. Pooled data showed a tendency to reduce the risk of grade ≥2 CIPN (CTCAE), but not statistically significantly (OR 0.34, 95% CI 0.11 - 1.11, P = 0.07; Fig. 3), and there was substantial heterogeneity (I2 = 88.1%, P < 0.0001; Fig. 3).
Incidence of CIPN (DEB‑NTC)
Two studies [16, 27] reported the data on the incidence of grade ≥2 CIPN measured with the Neurotoxicity criteria of Debiopharm (DEB-NTC), which was an oxaliplatin-specific neuropathy grading scale [29]. Pooled data showed that GM1 was not associated with a lower incidence of grade ≥2 CIPN (OR 0.25, 95% CI 0.01 - 7.10, P = 0.42) when compared with controls. Heterogeneity was substantial (I2 = 89.3%, P = 0.002; Fig. 4).
Secondary endpoints
Objective response rates to chemotherapy (RECIST)
Pooled data from three studies [15, 27, 28] that assessed the objective response rates to chemotherapy, including complete response (CR), partial response (PR), stable disease (SD), progressive disease (PD), overall response rate (ORR), disease control rate (DCR), showed that GM1 did not influence any response rate parameter (including CR, PR, SD, PD, ORR, and DCR), indicating that GM1 could not affect the antineoplastic property of chemotherapeutic agents. The heterogeneity was relatively low (Fig. 5).
Incidence of adverse events (CTCAE)
The incidence of adverse events was investigated using the CTCAE. Two studies [16, 17] reported the data on the risk of grade ≥2 fatigue, nausea, and diarrhea, and these data were pooled for meta-analysis. GM1 did not influence the incidence of grade ≥2 fatigue (OR 0.19, 95% CI 0.00 - 11.15, P = 0.42), the risk of grade ≥2 nausea (OR 0.67, 95% CI 0.08 - 5.37, P = 0.71), or the risk of grade ≥2 diarrhea (OR 0.70, 95% CI 0.35 - 1.38, P = 0.30). There was low heterogeneity in grade ≥2 diarrhea (I2 = 0.0%, P = 0.64); however, the heterogeneity in grade ≥2 fatigue was high (I2 = 95.2%, P < 0.001), and the heterogeneity in grade ≥2 nausea was moderate (I2 = 54.3%, P = 0.14, Fig. 5). Only one study [17] reported on the incidence of rash associated taxane and no statistically significant differences were observed in grade ≥2 rash (OR 4.04, 95% CI 0.44 - 36.82, P = 0.22, Fig. 5).
Chemotherapy dropout
Pooled data from two studies [15, 16] that assessed chemotherapy dropout suggested that GM1 did not influence the risk of chemotherapy dropout (OR 0.85, 95% CI 0.33 - 2.18, P = 0.74) with moderate heterogeneity (I2 = 72.8%, P = 0.06; Fig. 5).
Subgroup analyses
Because no significant effect was observed on the overall incidence of CTCAE grade ≥2 CIPN in the primary analysis, we further conducted subgroup analysis to explore the effect of the type of chemotherapy drugs on the results. It showed that for the endpoint of CTCAE grade ≥2 CIPN, GM1 was associated with a lower risk of taxane-induced peripheral neuropathy (OR 0.003, 95% CI 0.00 - 0.05) but not reduced risk of oxaliplatin-induced peripheral neuropathy (OR 0.63, 95% CI 0.35 - 1.13) (Fig. 2); indicating that GM1 might be more useful in preventing taxane-based peripheral neuropathy, and GM1 could not prevent oxaliplatin-induced neurotoxicity. Moreover, subgroup analyses stratified by other factors such as age, gender were not conducted because no sufficient data were available.
Trial sequential analysis (TSA) for the incidence of CIPN
To assess whether the results of taxanes and oxaliplatin in our meta-analysis were conclusive, TSA was conducted. The calculations were based on a 5% α with a two-tailed test. We used a random-effects model to construct the cumulative Z curve. The incidence in the control arm was set empirically at 40%, and heterogeneity correction was model variance-based. We used an anticipated relative risk reduction (RRR) of 25.0% with a power of 80% to calculate the required information size (RIS) to detect or reject an intervention effect. The TSA findings showed that the results of the taxane subgroup were robust, while those of the oxaliplatin subgroup were inconclusive (Fig. 6).