Baseline neutrophil to lymphocyte ratio combined with serum lactate dehydrogenase level associated with outcome of nivolumab immunotherapy in a Japanese advanced melanoma population

Dear Editor, Although immune checkpoint inhibitors significantly improve the survival of patients with advanced melanoma, more than half of patients receive no benefit.1,2 To predict outcomes, efforts to associate baseline peripheral blood biomarkers were started in patients treated with ipilimumab. Among the most critical markers is an increased neutrophil‐to‐lymphocyte ratio (NLR), which is negatively correlated with outcome.3,4 Although several baseline factors have been reported to correlate with outcome in patients treated with nivolumab/pembrolizumab [eosinophil count, lymphocyte count, lactate dehydrogenase (LDH) and C‐reactive protein (CRP)],5,6,7 a positive link between NLR and outcome has yet to be shown. To investigate the correlation between baseline laboratory test parameters with nivolumab treatment outcome, data from patients with melanoma treated with nivolumab were collected from eight clinical sites in Japan. Data taken 0–4 weeks before the first dose of nivolumab were considered as baseline data. The baseline factors determined in this analysis were tumour type (cutaneous, acral or mucosal melanoma), stage (unresectable, M1a, M1b or M1c), high LDH (LDHhi), CRP and blood cell counts (total white blood cell count and absolute counts of neutrophils, monocytes, lymphocytes and eosinophils). To examine the association between outcome and each factor, logistic regression analyses were used. To determine the cut‐off values for each factor, we used the receiver–operating characteristic curve. Statistical tests were done with Stat Flex version 6·0 (Artec, Osaka, Japan) and Prism version 6 (GraphPad Software, La Jolla, CA, U.S.A.). The study was approved by the relevant institutional ethics committees. Of the 90 enrolled patients, overall response, as determined with RECIST (Response Evaluation Criteria In Solid Tumors), was one complete response and 10 partial responses; 27 patients had stable disease and 52 had progressive disease. Of the various baseline factors, we identified LDHhi, CRP, absolute lymphocyte count and NLR as candidate factors, as they yielded P‐values <0·1 on univariate logistic regression analysis. These factors were then evaluated with multivariate analysis. As a result, only an NLR > 2·2 was shown to be a significantly poor prognostic factor [odds ratio (OR) 4·16; P =0·003]. Apart from NLR, LDHhi showed a tendency toward poor outcome (OR 2·53; P =0·081). By using NLR > 2·2 and LDHhi, patients were divided into four groups (Fig. 1a,b). Responders tended to fall within the NLR ≤ 2·2/low LDH (LDHlo) quadrant, whereas nonresponders tended to fall within the NLR > 2·2/LDHhi quadrant. As a result, whereas 18 of 27 patients (67%) in the NLR ≤ 2·2/LDHlo quadrant responded (Fig. 1c), only three of 21 patients (14%) in the NLR > 2·2/LDHhi quadrant responded. The number of responders in the NLR > 2·2/LDHhi quadrant was statistically lower than those in both the NLR ≤ 2·2/LDHhi and the NLR ≤ 2·2/LDHlo quadrants (χ2P =0·023 and <0·001, respectively), indicating that this population was less likely to respond to nivolumab. Interestingly, although not significant, patients in the NLR > 2·2/LDHlo quadrant whose NLR decreased 6 weeks after the first dose showed a better outcome than those with increased NLR [P =0·097; Fig. 1c (right panel)]. We did not see a similar tendency in the other quadrants. (a) Lactate dehydrogenase (LDH) and neutrophil‐to‐lymphocyte ratio (NLR) development among responder patients. (b) LDH and NLR development among nonresponder patients. (c) Percentage of responders in each quadrant classified by high LDH (LDHhi) and NLR > 2·2. Patients in NLR > 2·2/LDHhi quadrant further developed NLR change at week 6 of nivolumab administration. (d) Percentage of responders in each quadrant classified by LDHhi and NLR > 3. (e) Percentage of responders in each quadrant classified by LDHhi and NLR > 5. *P <0·05; **P <0·01; ***P <0·001; #P <0·1, χ2 test compared NLRhi/LDHhi quadrant. In this study, we have shown that the baseline NLR level was associated with outcome in patients treated with nivolumab, which is similar to the findings of previous studies of patients with melanoma treated with ipilimumab.3,4 Intriguingly, a similar trend has been reported in patients with non‐small‐cell lung cancer treated with nivolumab,8 indicating that a possible correlation between increased NLR and poor outcome could be a common phenomenon in some cancer types. The reported cut‐off values of NLR in those previous studies were 3 and 4 for melanoma treated with ipilimumab, respectively,3,4 and 5 for lung cancer treated with nivolumab.8 When these values were applied to our cohort, a higher cut‐off value seemed to be superior in detecting patients who are unlikely to respond (Fig. 1d,e). Collectively, baseline NLR combined with LDH is a useful marker to predict response, at least within the reported NLR cut‐off values of 2·2–5. Although we could not correlate novel factors with outcome, our findings might help identify patients who may be better helped by clinical trials rather than nivolumab monotherapy. However, there is a possibility that high NLR and LDH reflects an advanced stage of melanoma, and not specifically resistance to nivolumab. In addition, there are several limitations to this study that should be considered: its retrospective nature and the small study population. Thus, our results should be considered exploratory in nature and need to be validated in larger prospective studies. We thank F. Miyamasu, Medical English Communications Center of the University of Tsukuba, for the English revision of the manuscript. E‐mail: fujisan@md.tsukuba.ac.jp Funding sources: none. Conflicts of interest: none declared.