The PubMed, EMBASE and Web of Science databases were searched in March 2025 for peer-reviewed English-language randomised controlled trials (RCTs). Searches were not limited by date restrictions. Search terms were: “Oxaliplatin induced peripheral neuropathy treatment” OR “Oxaliplatin induced neurotoxicity treatment” OR “Chemotherapy induced peripheral neuropathy and therapy” OR “Chemotherapy induced neurotoxicity and therapy”. Additional references were also obtained from the references cited by a recent relevant meta-analysis (13). Two investigators (M.S. and Z.A.) independently read and selected the retrieved abstracts. Discrepancies between the reviewers’ selections were resolved by discussion. Full text versions of potentially eligible studies were then read by M.S. to confirm each conformed to the inclusion and exclusion criteria.

RCTs were included if the worst OIPN grade on treatment in adult CRC patients was evaluated using the CTCAE scale and patients were randomised to a pharmacological agent or traditional herbal medicine being tested as a preventative treatment for OPIN. To be included in the meta-analysis, studies needed to report how many participants in each arm had experienced either <grade 2 OIPN or ≥grade 2 OIPN and specify the time point at which this was measured. The CTCAE version used in each study was not considered since the scoring criteria are consistent across all versions.

Studies were excluded if: there was no full text published in English; OIPN was only evaluated using an alternative to the CTCAE scale; it was not possible to calculate counts <grade 2 and ≥grade 2 OIPN (n=4) (14–16); analysis included patients with baseline peripheral neuropathy (PN) (n=2) (17, 18) the study aimed to treat rather than prevent PN (n=1) (19) patients with ECOG performance status 3 were included in the analysis (n=1) (20); or ≤10 patients were available for analysis (n=1) (21).

Different trials assessed OIPN at different time points. Where available, we extracted data from trials that assessed OIPN grade midway through treatment (cycles 4-6) and at the end of treatment (cycles 8-12) separately. In the Ca²⁺/Mg²⁺ infusion trials (22–25), highest PN CTCAE grades on treatment were provided. We combined results for CTCAE grade 2 and above since grade 2 is the usual threshold for clinical interventions, such as dose reduction of chemotherapy or treatment deferral.

Data was extracted from the 20 included studies by two of 5 investigators independently (M.S., M.M, D.B., A.G. and C.P.). Data was imported into RevMan (26).

Inverse weighted random effects meta-analysis was performed in RevMan (26) for pharmacological agents or classes of drugs if there were data from a minimum of 100 patients across 3 or more studies. Heterogeneity was estimated using the restricted maximum likelihood method within RevMan. Risk of bias was also performed using RevMan, assisted by the RoB-2 tool (27).

Figure 1 shows the PRISMA flow chart of how RCTs were identified and reviewed. We reviewed the full-text report of 29 studies ( Supplementary Table 1 ), of which 20 met the inclusion criteria (summarised in Table 1 with full details supplied in Supplementary Table 2 ). 12 studies were double-blinded and placebo controlled (22–25, 28–35), three were placebo controlled but blinding was unclear (37–39) and five studies compared to a control arm with no or unclear blinding (36, 40–43). Median incidence of grade ≥2 OIPN, as measured at the end of treatment (8-12 cycles), was 57.95% (range 31.2-100%) in the placebo or control arms.

The studies were conducted in multiple countries, six in Japan (24, 30, 31, 33, 41, 42), four from China (34, 35, 37, 43), three from Egypt (22, 36, 40), two from each of USA (23, 25), Iran (28, 32) and Italy (29, 38), and one from each of USA and Canada (39) ( Table 1 ). A total of 1989 patients were given interventional treatments and 1972 control patients were included. 1917 males versus 2008 females were enrolled into the trials and where ethnicity was reported, the majority were White. In most studies the enrolled patients had stage III or IV CRC. The doses of oxaliplatin used to treat patients were 85 (22–25, 30, 31, 33–41), 100 (29) or 130 (28, 32, 42, 43) mg/m² every 2-3 weeks. 14 different pharmacological interventions were investigated: N-acetylcysteine (28, 37) glutathione (29, 38) metformin (36), L-carnosine (40), Ca²⁺/Mg²⁺ infusions (22–25) goshajinkigan (30, 31) monosialotetrahexosylganglioside (35), cystine and theanine (41), recombinant thrombomodulin (ART-123) (33), guilongtongluofang (34), ninjin’yoeito (42), neurotropin (43), duloxetine (32) and celecoxib (39). Included in two separate meta-analysis were first, N-acetylcysteine, glutathione, metformin and L-carnosine, grouped because each is able to support anti-oxidative stress responses in cells either directly via scavenger activity, by inducing antioxidant transcription or by supporting mitochondrial health and function; and secondly, Ca²⁺/Mg²⁺ infusions. OIPN was assessed every 2 weeks in most studies (12/20 studies, Table 1 ; range: weekly to every 3 months).

The risk of bias analysis showed that 35% of studies had potential high risk of bias with 45% of studies showing low risks of bias ( Figures 2A, B ). Only domain 2, which assessed risk of bias due to deviations from the intended interventions, show that all studies possessed a low risk of bias. However, all other domains showed potential for high risk of bias in 10-35% of the studies.

Eight interventions were investigated by only a single study: monosialotetrahexosylganglioside (35), ART-123 (33), duloxetine (32) and celecoxib (39). None showed a significant reduction in grade ≥2 OIPN in clinical trials of 192, 51, 32 and 2,450 patients with CRC, respectively. In the celecoxib trial (39), patients were randomised to 6 or 12 weeks of oxaliplatin ± 3 years of celecoxib treatment. 1,225 patients received celecoxib and there was no difference in incidence of grade ≥2 OIPN in this group compared to the placebo-controlled group either during or after completion of treatment with oxaliplatin. There was also no difference in severity of OIPN measured 12 months after completion of treatment with oxaliplatin using the Functional Assessment of Cancer Therapy/Gynecologic Oncology Group-Neurotoxicity-13 (FACT/GOG-NTX-13). This well-powered trial provides strong evidence against the use of celecoxib to treat OIPN in CRC patients. The other single trials evaluating agents were too small to draw definitive conclusions.

Oral administration of cystine and theanine (41), guilongtongluofang (34), ninjin’yoeito (42) and neurotropin (43) showed statistically significant association with reduced grade ≥2 OIPN (results summarised in Supplementary Table 3 ). The sample sizes employed by these clinical trials ranged from 28-120. Each of these interventions needs further evaluation in future trials.

Goshajinkigan (TJ-107) is a traditional Japanese herbal remedy used for pain relief. Previous meta-analyses of studies investigating whether goshajinkigan is effective in reducing taxane-induced PN concluded there was no significant effect (44). We identified two fully blinded, placebo-controlled trials that evaluated whether goshajinkigan could ameliorate OIPN (30, 31). The first, Kono et al., 2013 (30) enrolled 93 patients, of which 89 could be included in the analysis. A non-significant reduction in grade ≥2 and grade ≥3 OIPN at cycle 8 was observed in the arm randomised to receive goshajinkigan (39% vs. 51% in placebo arm and 7% vs. 13%, respectively). The authors also investigated tumour response in 27 patients receiving goshajinkigan and 23 receiving placebo and observed a non-significant improvement in response in the patients receiving goshajinkigan. The second study, Oki et al., 2015 (31) performed an interim analysis of 142 patients and identified a significantly higher incidence of grade ≥2 OIPN in the arm receiving goshajinkigan compared to placebo (50.6% vs. 39%), and a significantly shorter time to grade ≥2 OIPN in the goshajinkigan arm (RR=1.908, 95% CI: [1.81-3.083], p=0.007). The interim analysis was performed using assessments of OIPN up until the time when 50% of the planned recruitment had been achieved. All patients had received at least one cycle of oxaliplatin but a breakdown of the number of cycles patients included in the interim analysis had received was not provided. Since it had been determined that goshajinkigan could not significantly outperform placebo even if the study was allowed to continue, it was terminated early. Collectively, these two studies do not support goshajinkigan being used to treat grade ≥2 OIPN.

Two interventions trialled in three or more studies were analysed by random-effects meta-analysis. First, multiple therapies able to increase anti-oxidative stress responses or defences in cells were evaluated together in a meta-analysis (28, 29, 36–38, 40). Second, trials of Ca²⁺/Mg²⁺ infusions were evaluated together (22, 23, 25).

We analysed 6 trials (28, 29, 36–38, 40) investigating 4 drugs that have anti-oxidative stress effects: N-acetylcysteine (28, 37), glutathione (29, 38), L-carnosine (40) and metformin (36). N-acetylcysteine is a precursor of a key antioxidant, glutathione, and administration of N-acetylcysteine raises cellular glutathione levels. Metformin and L-carnosine induce antioxidant defences indirectly through effects on mitochondria and transcription and L-carnosine also has direct free radical scavenger effects and chelates Zn²⁺ (45, 46). Interestingly each of the studies investigating a drug with anti-oxidative stress effects found significant reduction in CTCAE grade ≥2 OIPN in their intervention arms. Five of the studies assessed OIPN at the end of 8-12 cycles (6 months) therapy (28, 29, 36–38) and three assessed OIPN after 4-6 cycles (2-3 months of treatment) (29, 36, 40) so two separate meta-analyses were performed. In the meta-analysis of the effect on OIPN at 8-12 cycles (6 months of treatment), a statistically significant reduction of large effect was observed ( Figure 3 ; OR: 0.04, 95% CI: [0.01-0.12], p<0.00001) with no heterogeneity (I² = 0%). High heterogeneity (I² = 66%) in the meta-analysis of the effect of anti-oxidative stress agents after 4-6 cycles (2-3 months) of treatment with oxaliplatin hindered interpretation, but no significant reduction in grade ≥2 OIPN was observed ( Supplementary Figure 1 ; OR: 0.19, 95% CI: [0.02 -1.60], p=0.13).

Given that metformin only has indirect effect on ROS scavenging, we repeated both meta-analyses without the El-Fatatry (36) study. Very similar results were seen after 4-6 cycles (2-3 months) of treatment with oxaliplatin ( Supplementary Figure 2 ; OR: 0.10, 95%CI: [0.00-1.98], p=0.13) as well as after 8-12 cycles of treatment ( Supplementary Figure 3 ; OR: 0.03, 95%CI: [0.01-0.11], p<0.00001).

Two out of the three studies that tested assessed the effect on OIPN at 4-6 cycles had a control arm rather than a placebo arm and blinding was unclear. A sensitivity analysis assessing the impact of study design was not possible for the 4-6 cycles timepoint. Two of the studies included in the analysis of OIPN at 8-12 cycles were double-blind, placebo-controlled trials: Bondad et al., 2020 (28) and Cascinu et al., 2002 (29), which tested N-acetylcysteine and glutathione respectively. In a meta-analysis including only these two trials and excluding the three trials with unclear blinding, very similar overall measures of effect are seen compared to the 5-study meta-analysis presented in Figure 3 , (OR: 0.02, 95% CI: [0.00-0.15], p<0.0001; Supplementary Figure 4 ).

Three double-blinded, placebo-controlled trials (22, 23, 25) investigated patients randomised to receive either an infusion of Ca²⁺/Mg²⁺ pre- and post- oxaliplatin treatment (198 patients in total) or placebo (200 patients). One rationale for Ca²⁺/Mg²⁺ infusions was that acute PN may be due to the disruption of ion levels that affects neuronal function and homeostasis (47). Meta-analysis of the three trials revealed no association between administration of Ca²⁺/Mg²⁺ infusions pre- and post- oxaliplatin treatment and risk of CTCAE grade ≥2 PN, as assessed after 6 months of oxaliplatin-based chemotherapy ( Figure 4 ; OR: 0.57, 95% CI: [0.29- 1.11], p=0.10). This is in keeping with the results of a meta-analysis by Peng et al., 2022 (13). The Loprinzi et al., 2014 (25) trial also included an arm where patients received Ca²⁺/Mg²⁺ infusions before oxaliplatin and placebo after. It also failed to show any significant reduction of OIPN (p=0.3383). The risk of bias analysis ( Figures 2A, B ) identified high risk of bias for the Grothey et al., 2011 study (23) and noted a high attrition rate, as well as premature termination following interim results from the Combined Oxaliplatin Neurotoxicity Prevention Trial (CONcepT) which suggested Ca²⁺/Mg²⁺ infusions affected the response to chemotherapy (48).

While the beneficial effect of administrating anti-oxidative stress drugs alongside oxaliplatin was clear from the meta-analysis that we performed, the drugs do not all act by the same mechanism.

Given that oxaliplatin treatment results in mitochondrial dysfunction and excess production of ROS, it seems probable, therefore, that anti-oxidative stress compounds protect against PN by scavenging excess ROS and/or by inducing Nrf2-based antioxidant transcription. If true, it would be interesting to test targeted forms of antioxidants that concentrate in mitochondria e.g. mitoQ, as these may be more effective at lower doses.

Previous studies have uncovered potentially protective effects of anti-oxidative stress agents against CIPN (57, 58). They included patients with different cancer types, treated with various chemotherapy agents and scored PN using different methods. This study is, to our knowledge, the first to systematically review potentially protective treatments for severe PN in patients with the same cancer type, treated with the same chemotherapy drug and where the same method of scoring PN was used in each trial.

None of N-acetylcysteine, glutathione, L-carnosine or metformin are currently recommended for treating OPIN in the current ASCO guidelines (12). Despite glutathione treatment significantly reducing OIPN in an RCT conducted over 20 years ago (29) none of the anti-oxidative stress agents assessed here have been tested in a large clinical trial. The large effect size of anti-oxidative stress compounds revealed in our meta-analysis, coupled with the lack of heterogeneity in the meta-analysis when OIPN was assessed at 6 months of treatment, supports a larger study to confirm this protective effect. This study must also address potential changes to the efficacy of oxaliplatin. These anti-oxidative stress compounds are cheap and well-tolerated and a successful trial would open options for a simple way to treat OPIN simply with large potential benefit for CRC patients.

The studies included in the anti-oxidative stress therapies meta-analysis were all small and each had a high rate of grade ≥2 OIPN in the placebo arms (88-100%). It is possible that these interventions will not prove effective at reducing incidence of grade ≥2 OIPN if the incidence in the placebo arm is lower. We only considered trials using clinician-scored CTCAE criteria to score OIPN to facilitate comparability across studies. However, we accept that it is a subjective measure and will vary between the scorers and this may not be the optimal measure of OIPN symptom severity. Patient reported outcomes may be preferable, particularly if combined with CTCAE scores (59) and incorporating novel biomarkers such as neurofilament light (NFL) chain has potential for a further objective method for assessing the severity of OIPN (60).

Despite strict inclusion criteria there were potential sources of bias identified in 4 out of the 6 studies that investigated anti-oxidative stress agents (36–38, 40). Blinding was not clearly described for four of the studies (36–38, 40). However, when unblinded trials were excluded, the two remaining studies (28, 29) still showed clear evidence of a significant reduction in grade ≥2 OIPN (P<0.0001) in the intervention arm in 50 patients ( Supplementary Figure 4 ).

The variation in delivery of the drugs (intravenous and oral) and in trial designs, as well as the high risk of bias in the studies, means that there is an urgent need for further validation through a large-scale, high quality RCT to establish a definitive beneficial effect and to identify which class of antioxidants may be most effective and best tolerated. Despite the promise that antioxidants may help to reduce the severity of OIPN, the evidence for this comes from small trials with differences in trial designs. There remains an urgent need for novel interventions to be developed and tested to reduce peripheral neuropathy and enhance the quality of life for patients receiving oxaliplatin-based therapies.