The systemic review and meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines (27). Two investigators (SP and AFY) independently searched, without language restriction, PubMed-MEDLINE, Embase and Scopus for randomized controlled trials (RCTs) published since 1 Jan 2005 to 8 Aug 2020. A detailed list of the search terms can be found in the Supplementary Material .

Human RCTs that assessed the efficacy of any form of protective measure to try to prevent and/or reduce the incidence of chronic OIPN as compared with placebo, no intervention or other interventions were included. These trials must have enrolled adult cancer patients with the planned chemotherapy regimen containing oxaliplatin as the only neurotoxic chemotherapeutic agent and reported the incidence and/or severity of chronic OIPN measured using clinical scales or electrophysiological studies. Trials were excluded if they were single-armed, retrospective, non-randomized, enrolled patients with pre-existing peripheral neuropathy, enrolled patients not naïve to chemotherapy, did not report outcomes specific to OIPN, or only reported acute OIPN outcomes. The primary outcome was the incidence of grade 2 and above peripheral neuropathy as assessed by the Common Terminology Criteria for Adverse Events (CTCAE). Information on other outcomes such as neuropathy grades assessed by other neuropathy scales or electrophysiology measurements were also extracted. For trials that did not use CTCAE in the outcome assessment, the primary outcome measures reported by trials were used for analysis. For the trials that did not report the full details of outcomes for analysis, the authors were contacted for more information. If such data were not available, then such trials were excluded from analysis.

The same investigators (SP and AFY) independently extracted data from the selected trials. Discrepancies were resolved by consensus amongst both investigators and advice from a third independent investigator (RS). Information extracted included: intervention dose and route, nature of the control (placebo, no additional intervention, or another intervention), trial primary and secondary outcomes and their assessment methods (clinical neuropathy scales or neurological electrophysiological studies) and time of assessment, follow up duration, and cumulative dose of oxaliplatin. Information was also collected for trial quality assessment, as outlined below.

Quality of the trials was assessed using the Risk of Bias Tool in Review Manager version 5.3 (RevMan 5.3) software by Nordic Cochrane Centre, and scored according to the domains of selection bias, performance bias, detection bias, attrition bias and reporting bias. The risk of bias results were scored on three levels, namely “high bias”, “low bias” and “unclear bias”, and then included into a rating approach proposed by the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) Working Group to evaluate the certainty of the result across four domains, namely risk of bias, inconsistency, indirectness and imprecision (28). The final GRADE scores represent four levels of certainty, namely “very low” (1), “low” (2), “moderate” (3) and “high” (4) ( Supplementary Material )

For trials using dichotomous outcome measures, the treatment effect was calculated across trials in the form of Risk Ratio (RR) based on the intention-to-treat sample size. If there were two or more RCTs that reported the same outcome measure by comparing the same combination of interventions, the respective RR were summarized in Forest plots using the random effects model of meta-analysis, in RevMan 5.3. All reported p-values were two-sided and p<0.05 indicated statistical significance. Heterogeneity was assessed via Cochran’s Q test and Higgins’ I² test. If significant methodological or statistical heterogeneity was detected, sensitivity analysis was performed based on factors that contributed to heterogeneity. For continuous outcomes, such as pain or neuropathy scores, the difference in means was calculated and the results were expressed with 95% confidence intervals (CI). For outcomes that did not fit the above criteria, significant outcomes were summarized as they were presented in the results section of the original trial.

A network meta-analysis was also conducted using R programming (netmeta package, frequentist method and random effects model), to obtain indirect evidence of treatment effect from the trials that compared two types of interventions. The main purpose of performing a network meta-analysis is to be able to derive indirect comparisons between a type of intervention and control (either placebo or no additional intervention), as such information was not available from the original trials (29).

The search strategy ( Figure 1 ) derived a total of 44 RCTs that examined 30 interventions for OIPN prevention, of which 43 trials examined 29 pharmacological interventions and 1 trial examined 1 non-pharmacological interventions. Of those that compared pharmacological interventions against placebo or no additional intervention, seven trials examined calcium and magnesium (Ca/Mg) infusions, ten examined neuro-active pharmaceuticals, six examined other pharmaceuticals, eleven examined antioxidants, and seven examined herbal medicines. One of the trials, Dong et al. (30), assessed two different types of interventions; thus, it was classified under two sections. In addition, three trials compared two types of interventions (or combination interventions) and their results were analyzed separately using network meta-analysis.

Among the excluded studies, seven RCTs that enrolled patients receiving oxaliplatin, and whose primary outcome was the prevention of chronic CIPN, were also identified via full text screening. Of these, two trials (31, 32) enrolled patients with prior platinum exposure and the remaining five (21, 33–36) did not provide sufficient data for analysis. Hence, all seven of these trials were excluded from the final analysis.

Based on the RR calculated from primary outcome measures and the GRADE evaluation of certainty of evidence, the interventions were assigned to one of three levels of recommendation, namely “Has Potential”, “Insufficient Evidence” and “Not Recommended” ( Figure 2 ). An intervention was rated to have “Potential” if the RR < 1.00, the CI did not cross 1.00, and the GRADE evaluation score was ≥ 3 (“moderate” or “high”). An intervention was rated to have “Insufficient Evidence” if the RR < 1.00, but either the CI crossed 1.00 or the GRADE evaluation score was = 2 (“low”). An intervention was rated to be “Not Recommended” if the RR > 1.00 or the GRADE evaluation score was = 1 (“very low”). When there was more than one outcome measure reported, the primary outcome measures using CTCAE was prioritized in the overall evaluation, while evaluations based on other outcome measures were recorded in the Table S1 and the scales explained in Table S2 . In summary, six interventions were rated as having “Potential”, nine interventions as “Insufficient Evidence” and fifteen interventions as “Not Recommended”.

In addition, among the fourteen interventions with statistically significant lower RR of OIPN incidence, including L-carnosine ( Figure S1 ), six were classified as “insufficient evidence” while two were classified as “not recommended” as their overall ratings were limited by “low” and “very low” GRADE certainties respectively ( Table 1 )

Refer to the Supplementary Material for summaries of trial characteristics and risk of bias of all trials ( Table S3 ), dichotomous or continuous outcome results analyzed from single trials ( Tables S4 , S5 ) and table presentation of network meta-analysis results ( Table S6 ).

For the ease of presentation, the pharmacological interventions were classified into five groups based on mechanisms of action, namely Ca/Mg, neuro-active pharmaceuticals, other pharmaceuticals, antioxidants, and herbal medicine.

Given the lack of definitive evidence on preventive interventions for OIPN, it is paramount to examine the relevant theoretical background and pre-clinical studies, so as to guide future trials.

Mechanisms for OIPN can be classified into platinum-related or oxaliplatin-specific. Platinum-related mechanisms include direct platinum-induced DNA damage in dorsal root ganglion, mitochondrial dysfunction, oxidative stress, and increased neuronal apoptosis (10, 83). On the other hand, oxaliplatin-specific mechanisms and targeted neuroprotective interventions have been examined in various pre-clinical studies, as summarized in a recent systematic review by Kawashiri et al. (84). Unfortunately, success in animal-based models has not consistently translated into clinical effectiveness. Various agents including anti-oxidants have been shown to reduce oxaliplatin-induced behavior alterations in mice (85), but when assessed in clinical trials, they range from being not recommended [e.g. Vitamin E (α-tocopherol)] to requiring further evidence to validate neuroprotective potential (e.g. glutathione and n-3 PUFA). Nonetheless, there are many emerging therapeutic targets worthy of further clinical assessment. Notable mechanisms in pre-clinical studies include remediation of oxaliplatin-induced peroxisome alterations by activating Peroxisome Proliferator Activated Receptor-γ (PPAR-γ) (86), antagonism of noxious hypersensitivity mediated by interactions between astrocytic vascular endothelial growth factor-A (VEGF-A) and VEGF receptor-1 (87), and alternatively via the α9α10 nicotinic acetylcholine receptors (nAChR) (88), and reduction of neuronal inflammation by promoting Sphingosine 1-Phosphate Receptor 2 (S1P₂) signaling (89).

Lastly, a paucity and need for more studies on non-pharmacological interventions was noted. Only one trial, exploring air wave pressure in preventing OIPN, fulfilled the selection criteria (59), however it was unable to demonstrate any significant benefit. In other CIPN prevention studies, four main non-pharmacological interventions have previously been assessed in RCTs, namely acupuncture (17), exercise (18, 90), compression therapy (91), and cryotherapy (15, 16, 19–21, 92, 93). Of these, the most promising intervention has been cryotherapy (7, 8). In a recent meta-analysis on the effectiveness of cryotherapy by Bailey et al. (94), it was noted that all seven trials assessed taxane-induced peripheral neuropathy, although the trial by Beijers et al. also recruited patients on oxaliplatin (21). Given that cold exposure is generally avoided to prevent acute OIPN, it is interesting that this study explored cryotherapy for CIPN prevention including an oxaliplatin cohort. One-third of the patients in this trial discontinued the intervention due to intolerance, however further details are unavailable. This preliminary evidence suggests future studies can be conducted explore cryotherapy for OIPN prevention.