Obesity-associated inflammation has been shown to dysregulate the immune response, potentially having profound effects on the toxicity and efficacy of immunotherapy across different type of malignancies.1–5 In non-small cell lung cancer (NSCLC), recent evidences have already confirmed the positive association between obesity and improved outcomes in patients receiving immune checkpoint inhibitors as further line of therapy, regardless of programmed death ligand-1 (PD-L1) tumor expression.6 7 Moreover, a pooled post-hoc analysis of data from four prospective trials (two of which randomized, with docetaxel as control arm) of pre-treated NSCLC patients receiving atezolizumab has provided robust evidences supporting this hypothesis.8 Importantly, Kichenadasse and colleagues confirmed that obesity was associated to survival benefit (compared with normal-weight patients) in patients treated with atezolizumab, but not in those who received chemotherapy, suggesting a predictive as well as a prognostic role of a high body mass index (BMI).8 Of note, the survival benefit was more pronounced in the PD-L1-positive subgroup, while a loss of statistical significance was reported within the PD-L1-negative patients.8

Accordingly, some evidences suggest that adipose tissue might play a pivotal role in regulating the immune homeostasis.9 10 In a recent meta-analysis including 203 articles, more than 6,000,000 patients across 15 different malignancies, the association between obesity and improved clinical outcomes was confirmed in those tumor types in which programmed death-1 (PD-1)/PD-L1 checkpoint inhibitors have proven to be more effective (such as lung cancer and renal cell carcinoma), despite studies involving patients who received immune checkpoint inhibitors were poorly represented in the meta-analysis itself.11 Therefore, assuming that the adipose tissue has somehow a role in the antitumor immune response, evaluating whether the variation of the BMI in patients receiving immune checkpoint inhibitors affects the clinical outcomes represents an interesting area of research.

Recently, we published a large real-world multicenter study of metastatic NSCLC patients with a PD-L1 expression ≥50%, receiving first line single agent pembrolizumab at 34 European institution, aimed at investigating the clinic-pathologic correlates of efficacy.12 Here, we present the clinical outcomes analysis according to baseline BMI and BMI variation (defined as ΔBMI) of the same study population. In order to confirm the results, we also evaluated a control cohort of metastatic NSCLC patients treated with first line platinum-based chemotherapy in clinical practice.

In this study we confirmed that also in a population of NSCLC patients with a PD-L1 expression ≥50% receiving first line pembrolizumab, obesity is associated to improved clinical outcomes. Obese patients had a significantly higher ORR, and prolonged PFS and OS compared with normal-weight patients. The fact that this trend was confirmed at each outcome analysis (ORR, PFS and OS) and with multivariate regression models further supports our results. More importantly, we did not observe such correlation among patients treated with chemotherapy, suggesting that the obesity-related clinical benefit is unique of patients treated with immunotherapy. Of note, obese patients who received chemotherapy had a non-significant trend towards a lower ORR and shorter OS, and a significantly shorter PFS, compared with normal-weight patients. These evidences are aligned to what already reported for NSCLC patients in clinical practice,6 7 20 and to the improved survival observed in obese PD-L1-positive NSCLC patients by Kichenadasse et al.8 Consistently with our results, also in the control cohort of the study by Kichenadasse and colleagues, no survival benefit was reported for obese patients who received chemotherapy.8 Remarkably, obese patients in the chemotherapy cohort did not have an improved OS despite 58.5% of them received immunotherapy on chemotherapy failure.

As previously mentioned, a growing body of evidence is indicating that the adipose tissue might play a critical role in shaping the antitumor immune responses. Wang and colleagues reported that obese mice showed a significant increase of dysfunctional exhausted T-cell.9 Such exhaustion could be partially mediated by the immune checkpoints (as the PD-1/PD-L1 pathway) and driven by leptin, and then might be more likely elicited by immune checkpoints inhibitors.9 10

Nonetheless, a high baseline BMI has been historically somehow associated to improved lung cancer survival across different disease stages, even in the ‘pre-immune checkpoint inhibitors era’, reflecting the potential prognostic nature of baseline BMI in lung cancer patients.21–24 However, we need to highlight that even if cancer cachexia mechanisms are not completely known yet, several evidences showed that the systemic inflammation plays a central role.25 Lung cancer is an aggressive disease, and metastatic NSCLC patients usually to present with poor clinical condition and weight loss at the time of first line treatment commencement, which underlies a systemic inflammatory over-activation.26 From this point of view, a higher baseline BMI might be considered a sign of functional reserve and a protective feature for NSCLC advanced patients.

To that end, in a study evaluating 2585 stage IV NSCLC patients from three trials with first line chemotherapy, obesity was associated with improved OS.27 Intriguingly, when time on-study exceeded 16 months, obese patients experienced a significant increase in their hazard rate for death, compared with normal/overweight patients, suggesting a possible negative role of the sarcopenic obesity in the long-term during chemotherapy.27

Also, the efficacy analysis according to ΔBMI revealed interesting results. A BMI variation of 1.4% during treatment was significantly (and concordantly) related to each measured outcome (ORR, PFS and OS) for both the pembrolizumab and chemotherapy cohorts. Looking to the forest plot graphs, we can notice that the aOR and the aHRs were concordantly more pronounced for the pembrolizumab cohort. Similarly, the Kaplan-Meier survival curves according to the ΔBMI (online supplemental figure 4) show a much more marked survival benefit for patients who experienced a BMI gain of ≥1.4% within the pembrolizumab cohort rather than in the chemotherapy cohort. These findings could make us assuming that the prognostic role ΔBMI is stronger during the immunotherapy rather than chemotherapy. However, we must not fail in taking into account that the optimal grouping computed with the recursive partitioning algorithm within the patients receiving pembrolizumab was applied to both the cohorts. Interestingly with respect to the optimal grouping, while for the pembrolizumab cohort a harmonic positive cut-off for ΔBMI was found for ORR, PFS and OS (+1.4%), three different cut-offs were found among the chemotherapy cohort using a recursive partitioning: a positive value for ORR (+7.1%), and two different negative values for PFS (−6.8%) and OS (−2.2%). Although using different cut-offs for the same variable may seem redundant, the results raise some questions. A harmonic and relatively small weight gain during pembrolizumab could be considered a concordant independent predictor for ORR, PFS and OS. Conversely, a more pronounced weight gain during chemotherapy is associated with ORR, while a weight loss is significantly associated with inferior PFS and OS. From this perspective, the BMI acts differently in the two cohorts; considering the adipose tissue an immune organ, patients who experienced a slight weight gain during immunotherapy had a significantly improved ORR, PFS and OS. Considering instead the adipose tissue as a functional reserve for patients receiving chemotherapy, a pronounced weight gain is related to an improved ORR, while a weight loss had a significant prognostic negative role for PFS and OS. In a post-hoc analysis from three phase III studies of patients receiving a platinum-based systemic therapy, Patel and colleagues evaluated the clinical outcomes according to a pre-planned weight gain cut-off (> vs ≤5%), assessed by monitoring post-baseline weight variation till the maximum weight during treatment or at the 30-day post-study discontinuation follow-up visit.28 The authors reported that a weight gain of >5% were significantly related to improved ORR, PFS and OS.28 However, the different methodology does not allow a direct comparison between our chemotherapy cohort and the cohort of Patel et al.28

Our study has several limitations beyond the retrospective design and the consequent selection biases. First, we cannot consider the chemotherapy cohort a proper validation cohort, because we did not perform any case-control matching; the percentage of elderly patients and patients with ECOG-PS ≥2 are in fact significantly different. Considering the good clinical outcomes achieved in absolute terms, the chemotherapy cohort might had been positively biased. These findings are likely to be related to the clinicians attitude to treat with doublet chemotherapy more fit patients, compared with single agent pembrolizumab. Moreover, this positive selection might partially subtend to the lack of association between survival and BMI in the chemotherapy cohort. From this perspective, the analysis of the chemotherapy cohort collides with the evidences suggesting that also in NSCLC patients receiving chemotherapy a higher BMI is associated with an improved survival.11 Additional limitations include the data lack availability regarding comorbidities, the different sample size of the two cohorts, as well as the median follow-up which was different. The chemotherapy cohort was not powered to detect statistically significant differences according to weight categories, moreover, being a historic cohort, we did not have data regarding PD-L1 expression for the subset of patients treated with chemotherapy. Nevertheless, considering the real-world prevalence of PD-L1 expression in NSCLC, we can assume that we one-third of the patients in the chemotherapy cohort had a PD-L1 expression of ≥50%.29 Without having the weight data at pre-specified time points, we computed the ΔBMI from treatment commencement to disease progression or to the last contact for censored patients, who, however, were progression free at the data cut-off and might had been positively selected with regards of the weight gain (due to better clinical condition). Conversely, progressed patients might have been negatively selected regarding the weight loss, even if the 4 weeks landmark selection was performed to mitigate this bias. In this respect, in our population, the probability of experience a BMI variation was conditional on the PFS, which was longer for the pembrolizumab cohort. Despite this, the median ΔBMI was the same in the two cohort.

We have also to consider that with the weight-based dose overweight/obese patients have been exposed to higher dose intensity of pembrolizumab. Nevertheless, pharmacokinetic studies did not report definitive findings in exposure differences nor in safety profiles between the regimens (weight-based and fixed dose).30 Furthermore, weight variation is inevitably related to the food intake, and chemotherapy is certainly associated to a higher incidence of anorexia, nausea and emesis compared with single agent pembrolizumab.

We confirmed that baseline obesity is associated to significantly improved ORR, PFS and OS also in metastatic NSCLC patients with a PD-L1 expression of ≥50%, receiving first line single agent pembrolizumab, but not among patients treated with chemotherapy. BMI variation during treatment is also significantly related to clinical outcomes, but it seems to have a different role for patients receiving first line pembrolizumab, compared with patients receiving first line platinum-based chemotherapy.