Summary of main results
Acetylcysteine (N-acetylcysteine, NAC)
The single eligible study evaluating the use of N-acetylcysteine (NAC) against the neurotoxicity of oxaliplatin was described as a pilot study and included only a small number of participants (five receiving NAC and nine controls) (Lin 2006). In addition, uncertainty about the randomisation methods and presence of blinding (participants or investigators), report only of subjective toxicity scales, and inclusion of nerve conduction results from the NAC group but not the control group are issues that limit interpretation of the preliminary study results.
The eligible studies evaluating the use of amifostine as a neuroprotective agent against the neurotoxicity of cisplatin and other chemotherapy agents are inconclusive in demonstrating efficacy, primarily because few studies utilised quantitative measures of neurotoxicity. Although the studies were generally well done, participant masking was unclear, perhaps because amifostine was given intravenously in conjunction with interval chemotherapy and felt to be of little interest to the recipient. The authors of one study acknowledged that the trial was not conceived as a double-blind study, but that the physicians assessing non-haematological toxicity usually were not the same physicians involved in administering treatment and they therefore could not be influenced by the evaluation of symptoms and signs associated with chemotherapy (Lorusso 2003). It is unknown whether this belief about the low likelihood of inadvertent influence on judgment is correct. Paclitaxel was used together with carboplatin in two of the amifostine trials that were included (Kanat 2003; Lorusso 2003).
Planting 1999 was the single study that included quantitative sensory testing (QST), our primary outcome measure, among their assessment instruments. That trial showed a favourable outcome in terms of amifostine neuroprotection, but the subclinical result was based on only 14 participants in the amifostine group and 20 participants in the control group and the results were not particularly robust, showing statistical but unclear clinical significance. Among the secondary outcome measures we selected, the only other quantitative measure used in any of the four studies was related to evaluation of peripheral nerve electrophysiology. A single study utilised electrophysiological measures, recording sensory nerve action potential (SNAP) amplitudes from two upper extremity sensory nerves and one lower extremity sensory nerve at baseline and after completing chemotherapy (Kanat 2003). This study, which also included a relatively small number of participants (19 participants in each study arm), found that the quantitative measures of large fibre sensory axons failed to identify evidence of significant amifostine neuroprotection. Unfortunately, the study also showed an unexpected and almost implausible low level of neurotoxicity among the control participants who received carboplatin and paclitaxel but not amifostine, rendering the amifostine efficacy result uninterpretable.
Despite the lack of quantitative measures showing amifostine efficacy, the authors of all seven trials concluded that pretreatment with amifostine reduced, prevented, or at least exerted some protection from the cumulative neurotoxicity associated with cisplatin or carboplatin and paclitaxel. This conclusion reflected the results obtained from the various neurological examination scales or neurotoxicity scales, most often the National Cancer Institute-Common Toxicity Criteria (NCI-CTC) neuropathy rating. The neurotoxicity rating scale was not specified by Gallardo 1999, but presumably represented the NCI-CTC scale. For the most part, the other scales utilised were of unknown sensitivity or specificity, and the clinical relevance of the results uncertain. For example, it is unclear whether any of the sensory symptoms or signs resulted in substantial functional impairment or persisted. Some of the neurotoxicity results were reported in terms of each evaluation, rather than for each participant, suggesting fluctuation during the trial. Such fluctuation suggests inclusion of non-specific symptoms, as opposed to the persistent distal predominant and symmetrical sensory symptoms and signs characteristic of most toxic neuropathies. One exception is the functional activities of daily living (ADL) scale reported by Kanat 2003. Although based on a small number of participants (19 participants in each study arm), the results of this sensorimotor neurotoxicity score suggested a small but statistically significant decrease in the ADL in 2 of 19 participants in the amifostine group compared to 9 of 19 participants in the control group (risk ratio (RR) 0.22, 95% confidence interval (CI) 0.6 to 0.90). Another exception reflected the use of the NCI neurological toxicity rating among 242 participants with ovarian cancer treated with cisplatin and cyclophosphamide (Kemp 1996). The significant decrease in the NCI rating among participants pre-treated with amifostine compare to control participants was an impressive finding, despite the lack of QST or electrophysiological evaluations. The limited availability of clinical trials utilising conventional QST and nerve conduction study measures of peripheral nerve function is disappointing, at least in part because of the relative simplicity of the peripheral nervous system evaluation relative to other neurological functions, such as behavior. Similarly, De Vos 2005 concluded that amifostine showed minor but significant activity in diminishing neurotoxicity (but not in preventing paclitaxel plus carboplatin-induced bone marrow toxicity) without interfering with chemotherapy efficacy.
The largely positive results based on subjective neuropathy ratings generally led to conclusions that amifostine could reduce the incidence and severity of peripheral neurotoxicity caused by cisplatin or oxaliplatin chemotherapy (e.g., Kemp 1996; and Lu 2008). These conclusions were supported by the pooled data involving secondary neurotoxicity rating scales, most often based on the NCI-CTC neuropathy scale. Although the additional studies identified in the 2013 update permitted pooling of data, there still were only a small number of studies that could be included, and those studies involved a relatively small number of participants. For example, the pooled data from three studies involving the equivalent of NCI-CTC grades ≥2 neurotoxicity included only 148 participants, of which 10 of 74 receiving amifostine showed ≥ grade 2 neurotoxicity relative to 39 of 74 control participants showed a similar grade of neurotoxicity (DeVos 2005; Gallardo 1999; Kanat 2003). In this analysis, the Kanat 2003 results represented outlying (positive) values relative to the other studies. That the pooled data showed substantial heterogeneity is perhaps not surprising considering the different cancers and primary treatment protocols described in these studies. Similar results were obtained from pooled analysis of the available data from three other studies using the equivalent of NCI-CTC grades ≥3 although the results were no longer statistically significant (De Vos 2005; Kemp 1996; Lu 2008), demonstrating how sensitive the results are to inclusion of data from different trials. In this analysis, the Lu 2008 results represent very outlying (positive) values relative to the other studies.
In terms of direct relevance to chemotherapy, the expectation that reduced neurotoxicity would result in increased dosing was not clearly realised in any of the trials reviewed, although Lu 2008 reported that the proportion of chemotherapy schedule adjustment because of chemotherapy-induced neurotoxicity was significantly lower in the amifostine group relative to the placebo group (4.3% versus 23.9%, P = 0.007).The expectation of the beneficial result of increased dosing is based on recognition that neurotoxicity is the primary dose-limiting adverse effect attributed to cisplatin.
De Vos 2005 conducted a randomised phase II study of paclitaxel and carboplatin with and without amifostine and, as part of their discussion, pooled the results of three studies, all involving chemotherapy with paclitaxel and carboplatin (De Vos 2005; Kanat 2003; Leong 2003). We had excluded from our review the data from the Leong et al. study because participants received only two doses of carboplatin. We included the data from the De Vos et al. study, but noted that we were unable to separate the effects of carboplatin versus paclitaxel. The pooled results reported by De Vos et al. identified an odds ratio (OR) for developing ≥ grade 2 neurotoxicity of 0.30 (95% CI 0.15, 0.63; P < 0.05, random-effects model. This compared to our pooled results (after change the RR to an OR), which showed an OR of 0.08 (95% CI 0.01to 0.79, random-effects model, P = 0.03) for developing ≥ grade 2 neurotoxicity. Notwithstanding the different studies included, the two analyses based on pooled data (our analysis and the one reported by De Vos et al.) suggested potential amifostine protection against chemotherapy-induced neurotoxicity.
In conclusion, the results of the available trials suggest the possibility of potential amifostine neuroprotection against cisplatin and other chemotherapy drugs. However, data involved a relatively small number of trials and participants, did not include improvement of primary measures or objective quantitative secondary measures, and the neuroprotection appears to be of small magnitude. Given this, in our opinion the overall efficacy results, while promising, remain inconclusive.
Calcium and magnesium (Ca/Mg)
The studies of Ca/Mg infusions as a chemoprotective agent against oxaliplatin neurotoxicity included subjective NCI-CTC grading for neuropathy and Debiopharm Neurotoxicity Scale (DEB-NTS) or similar neurotoxicity grading criteria (Chay 2010; Grothey 2011; Ishibashi 2010). Enrolment in all three studies was terminated prematurely, owing to reports that treatment results were poorer in the Ca/Mg group than in the control group. in This was according to the authors' interim analysis in the Ishibashi 2010 study, results not confirmed in the final analyses and, in the Chay 2010 study, according to reports from other studies indicating negative results. Grothey 2011 was terminated early because of reports of treatment interference by the study medication (Grothey 2011).The early discontinuations resulted in a small sample size and limited the data available to determine if Ca/Mg infusions had neuroprotective potential. The NCI-CTC rating of Grothey 2011, the largest of the available studies, reported a borderline significant result favouring Ca/Mg infusions for preventing neuropathy, defined as an NCI-CTC grading ≥2 (OR 0.42; 95% CI 0.17 to 0.99). However, the combined available data using the same grading scale did not support a significant beneficial effect. In a non-randomised retrospective analyses, Knijn 2011 reported that early ≥ NCI-CTC grade 2 neurotoxicity (occurring during six cycles of oxaliplatin chemotherapy) occurred in 218 of 551 (40%) Ca/Mg-treated participants versus 81 of 181 (45%) non-Ca/Mg-treated participants. Similarly, late ≥ grade 2 neurotoxicity present at the last cycle before going off of study was present in 148 of 551 (27%) Ca/Mg-treated participants versus 62 of 181 (34%) non-Ca/Mg-treated participants. The retrospective Knijn 2011 described a decrease in the frequency of late ≥ grade 2 neurotoxicity present at the last cycle before going off of study in Ca/Mg-treated participants relative to those who did not receive Ca/Mg. When we pooled these positive results with the other studies reporting ≥ grade 2 neurotoxicity at the end of chemotherapy, we found a modest but statistically significant result favouring the use of Ca/Mg, OR 0.68 (95% CI 0.49 to 0.94) (Chay 2010; Grothey 2011; Ishibashi 2010; Knijn 2011).
Our limited pooled results showed a non-significant reduced risk of developing a NCI-CTC ≥ grade 2 favouring Ca/Mg (RR 0.84, 95% CI 0.62 to 1.05), results different from those two published meta-analysis involving the efficacy of Ca/Mg infusions (Ao 2012; Wen 2013). Ao 2012 analyses included one study requiring translation that we are waiting to review (Dong 2010), a double-blind study involving oxaliplatin-induced neuropathy in which 4 of 20 Ca/Mg-treated participants developed chronic neuropathy versus 11 of 27 control participants (a borderline significant difference). The pooled analyses performed in Ao 2012 reported an OR of 0.44 (95% CI 0.23 to 0.85, fixed-effect model), indicating a significant result favouring treatment with Ca/Mg. After converting our analyses to an OR and using a fixed effects analyses, we showed a non-significant OR of 0.58 (95% CI 0.27 to 1.21), results still different from those reported by Ao et al. Similarly, Wen 2013 reported the results of a second meta-analysis involving oxaliplatin-related neurotoxicity. Their analysis is potentially flawed because of the variability of the studies included, in which the study by Chay et al. was the only RCT included (it had a non-significant outcome for total cumulative subjective sensory neurotoxicity). Their analyses reached significance only after including the non-randomised trials of Knijn 2011 (discussed immediately above) and Gamelin 2004. We did not include data from either the Knijn et al. or the Gamelin et al. studies in our analyses because both represented retrospective non-randomised studies (and because it is a Cochrane Neuromuscular Disease Group policy to not include data from non-randomised studies in the results section). When we investigated the effect of pooling the retrospective data from Knijn et al. with the data we reported in the results section, we found a borderline significant RR of 0.80 (95% CI 0.62 to 1.05) favouring the use of Ca/Mg, a result qualitatively similar to the results of the two published meta-analyses. Regardless, the conclusion of Wen 2013a that Ca/Mg infusions tend to decrease the incidence of oxaliplatin-induced cumulative neurotoxicity and thus enhance patients' tolerance to oxaliplatin is based primarily on subjective data derived from retrospective studies, not RCTs.
In summary, the results of the best available trials selected for analyses are difficult to evaluate because of the small number of subjects, due primarily to early termination of several important studies (for reasons unrelated to the study intervention or treatment protocols). Based on the remaining data however, our results of the available RCT data do not show suggest statistically significant effect of Ca/Mg in preventing oxaliplatin-induced neurotoxicity. Inclusion of results from several retrospective non-randomised studies suggest more positive, borderline significant results favouring the use of Ca/Mg. At present, however, the limited data are not convincingly positive in favour of Ca/Mg neuroprotection and in our opinion, the overall efficacy results are promising but inconclusive.
The single study of DDTC as a neuroprotective agent suffers from having no measures of neurotoxicity other than subjective reporting (NCI). To gain full appreciation of the magnitude of the difference between the two arms (DDTC and placebo), one should probably add those withdrawn for toxicity, to patient request, and adverse experience (Gandara 1995).
Overall, six out of seven studies reported a significant protective effect of GSH. All measures of peripheral neuropathy favoured the GSH group, including the measure of VPT (one study), sural SNAP amplitudes (four studies) and improvement in functional measures or various neurotoxicity rating scales (six studies). Even though the overall effect of GSH appears to be beneficial and protective, the variable dosages used with different malignancies and different combinations of chemotherapy, high drop out rate, predominant reliance on subjective measures, limited statistical analyses, and lack of long-term follow-up, make the overall effect of GSH difficult to judge.
Overall, the few eligible studies evaluating the use of Org 2766 as a neuroprotective agent against the neurotoxicity of cisplatin are inconclusive in demonstrating efficacy. A major concern is that the total number of participants enrolled in the studies is rather low (188 Org 2766-treated and 123 control participants) and, moreover, participants are not homogenously distributed among the four trials, since one of them (Roberts 1997) included 68% of the Org 2766-treated and 54% of the control participants. All the trials included QST, our primary outcome measure, among their assessment instruments, while none of the studies reported an evaluation of peripheral nerve electrophysiology or effects on ADL. Neurological examination was based on non-validated scales in all the four trials.
The first study suggesting a protective effect of Org 2766 (van der Hoop 1990) is based on an inadequate statistical analysis. In fact, analysis was performed after six cycles of cisplatin on only 28 participants out of the 55 admitted to the study, while the others were not eligible or had not yet received the planned chemotherapy cycles. Intermediate analysis (i.e. after four cycles of cisplatin) was performed on 39 participants. The authors of the second study (Hovestadt 1992) admitted that the number of participants was too low to allow a reliable formal statistical analysis. The third study was the only one performed mostly on males (22 men versus 1 woman). The authors used two different methods of statistical analysis and the results were conflicting (van Gerven 1994). The largest study had adequate subject, outcome assessor and observer blinding and also the statistical analysis was adequate (Roberts 1997). Instead of providing evidence of protection induced by Org 2766, the authors suggested that high doses of the compound might even increase the rate of change and degree of neuropathy induced by cisplatin (P value > 0.05).
In conclusion, although the results of the first trial (van der Hoop 1990) suggested the possibility of potential Org 2766 neuroprotection, effects appear to be of small magnitude or not convincingly positive in favour of Org 2766 neuroprotection, particularly in view of the results reported in the most recent trial (Roberts 1997). Furthermore, the combined data from the three trials using the same measure showed no significant group difference at the follow-up QST examination (mean difference -1.77 95% CI -4.78 to 1.23). The overall efficacy results are negative.
The only study evaluating the efficacy of OXY for prophylaxis against oxaliplatin-induced neuropathy reported a favourable effect (Argyriou 2006a). The results were based on an open label evaluation (randomised but not placebo-controlled), a small sample size, and without quantitative sensory testing as a primary outcome measure. However, validated clinical instruments (NSS and NDS) and appropriate neurophysiological measures were incorporated and showed several significant group differences, all favouring the OXC group. The significant nerve conduction results involved a change in the baseline to six month recordings for the lower extremity SNAP amplitude measures (sural and superficial peroneal), but not the ulnar sensory or peroneal motor measures, results consistent with those expected to represent the most sensitive indicators of an oxaliplatin-induced neuropathy. Comparisons of the mean SNAP amplitudes for the treatment versus control groups post treatment (six-month recordings after 24 cycles) showed no significant differences in any of the SNAP amplitudes, however. Although the significant neurophysiological group differences based on the change from baseline to six month records were modest and of uncertain clinical importance, the overall results support further investigation of OXC in a larger placebo-controlled randomised trial.
The results from the single study involving retinoic acid as a neuroprotective agent are limited by methodological issues resulting in uninterpretable nerve conduction study results, unbalanced and unexplained baseline group differences, use of the NCI-CTC grading as the only measure of neuropathy, and an inadequate follow-up interval (Arrieta 2011). The NCI-CTC assessment of neuropathy grades ≥2 were present in 23 of 45 ATRA participants and in 37 of 47 placebo participants, showing a borderline significant difference favouring the ATRA treatment group over the placebo group (RR 0.75 95% CI 0.55 to 1.02) (Analysis 18.1).
Although the results involving vitamin E as a neuroprotective agent are encouraging, methodology issues, the small size of the study, the use of multiple chemotherapeutic regimens (including taxane in the largest available study), lack of blinding, and lack of primary outcome measures make the data less than convincing. This conclusion is despite the statistically significant results from the pooled data involving subjective measures of neuropathy (NCI-CTC neurotoxicity rating). The changes noted in median SNAP but not the in sural SNAP amplitudes and the use of a non-validated toxicity measure suggest that additional more definitive studies are needed. In our opinion the overall efficacy results, while promising, remain inconclusive.
The quality and characteristics of the trials reviewed were quite variable, and included different measures of neuropathy (qualitative and subjective), different durations of follow-up, and different analyses. The duration of follow-up must be sufficient to identity cisplatin-induced sensory nerve deterioration, and therefore should extend beyond the last cisplatin treatment. How long after the last treatment is open to debate, as all toxic neuropathies demonstrate some progression after exposure ceases, and patients with cisplatin-induced neuropathy can show improvement (depending on the initial severity) after cisplatin is discontinued. We included all evaluations performed zero to six months after the last treatment, selecting the evaluation closest to three months after treatment to the extent possible. We believe that a two to three month interval after treatment is completed is biologically reasonable. In all, 15 trials were included in our initial review, a further five trials in our 2010 review, and an additional 9 trials in our 2013 update. The combined trials involved nine separate, unrelated potential neuroprotective agents and included many disparate measures of neuropathy, resulting in sufficient data to combine the results for only a few measures, most of which were secondary measures such as the NCI-CTC neurotoxicity rating. Based on our review, we feel that the evaluation of agents intended to prevent cisplatin-induced sensory neuropathy should include validated measures, and not necessarily be limited to the primary measures we initially identified (tests of QST at the index finger and great toe and nerve conduction study evaluation of SNAP amplitudes in the median sensory and sural nerves). Before performing our review, considerable consideration was given to two measures (QST or nerve conduction studies) competing for the primary endpoint. We selected QST as the primary endpoint, in part, because it had been used in several prominent trials. While QST is an excellent quantitative measure of the endpoint of interest (sensation), the SNAP amplitude has the advantage of providing information about the actual cisplatin target, the peripheral sensory nerve, independent of patient co-operation or motivation. Sufficient information exists about both measures to perform power calculations to determine the number of subjects required to detect a meaningful group difference. A difficulty we experienced in performing our review related to the limited data available, even when QST or SNAP amplitude recordings were performed. Most studies provided only descriptive statistics (e.g., mean, SD) reflecting the baseline examination and the follow-up examination for treatment and control groups, without information about change (mean, SD) between baseline and follow-up examinations. Inclusion of the latter facilitates comparisons with subsequent studies.