A multicenter study of body mass index in cancer patients treated with anti-PD-1/PD-L1 immune checkpoint inhibitors: when overweight becomes favorable

Background Recent evidence suggested a potential correlation between overweight and the efficacy of immune checkpoint inhibitors (ICIs) in cancer patients. Patients and methods We conducted a retrospective study of advanced cancer patients consecutively treated with anti-PD-1/PD-L1 inhibitors, in order to compare clinical outcomes according to baseline BMI levels as primary analysis. Based on their BMI, patients were categorized into overweight/obese (≥ 25) and non-overweight (< 25). A gender analysis was also performed, using the same binomial cut-off. Further subgroup analyses were performed categorizing patients into underweight, normal weight, overweight and obese. Results Between September 2013 and May 2018, 976 patients were evaluated. The median age was 68 years, male/female ratio was 663/313. Primary tumors were: NSCLC (65.1%), melanoma (18.7%), renal cell carcinoma (13.8%) and others (2.4%). ECOG-PS was ≥2 in 145 patients (14.9%). PD-1/PD-L1 inhibitors were administered as first-line treatment in 26.6% of cases. Median BMI was 24.9: 492 patients (50.6%) were non-overweight, 480 patients (50.4%) were overweight/obese. 25.2% of non-overweight patients experienced irAEs of any grade, while 55.6% of overweight/obese patients (p < 0.0001). ORR was significantly higher in overweight/obese patients compared to non-overweight (p < 0.0001). Median follow-up was 17.2 months. Median TTF, PFS and OS were significantly longer for overweight/obese patients in univariate (p < 0.0001, for all the survival intervals) and multivariate models (p = 0.0009, p < 0.0001 and p < 0.0001 respectively). The significance was confirmed in both sex, except for PFS in male patients (p = 0.0668). Conclusions Overweight could be considered a tumorigenic immune-dysfunction that could be effectively reversed by ICIs. BMI could be a useful predictive tool in clinical practice and a stratification factor in prospective clinical trials with ICIs. Electronic supplementary material The online version of this article (10.1186/s40425-019-0527-y) contains supplementary material, which is available to authorized users.


Introduction
Although the interaction between malnutrition and chronic inflammation has been widely investigated, whether this association is causative or correlative is still debated [1]. Historically, body mass index (BMI) has been considered the major surrogate of nutritional status and its correlation with clinical outcomes in advanced cancer patients has already been investigated without conclusive results [2][3][4][5].
It is now becoming clear that the nutritional assessment, which should include BMI, could be seen in a "new light" in the era of immune checkpoint inhibitors (ICIs). A large retrospective study has recently found an association between BMI and improved progression free survival (PFS) and overall survival (OS) in melanoma patients treated with either targeted therapy or immunotherapy [6]. Another study has reported that overweight sarcopenic melanoma patients treated with anti-PD1 (Programmed cell death protein 1) inhibitors experienced early acute limiting toxicity [7].
Additionally, another retrospective analysis by Richtig et al. revealed that overweight (BMI ≥ 25) melanoma patients (76 total) treated with ipilimumab had significantly higher response rate (p = 0.024) and a trend for longer OS (p = 0.056), when compared to non-overweight patients [8].
Lastly, Wang and colleagues have recently reported an improvement in terms of PFS (p = 0.003) and OS (p = 0.049) in a cohort of obese advanced cancer patients (BMI ≥ 30) treated with ICIs [9].
To further dissect this question, we conducted a large, multicentre, retrospective transverse study to evaluate clinical outcomes of patients with advanced solid tumors treated with ICIs according to baseline BMI.

Patient eligibility
This study enrolled patients with confirmed diagnosis of measurable stage IV cancer, who consecutively underwent treatment with single agent anti-PD-1/PD-L1 as 1 st or subsequent line, at the medical oncology departments of 17 Italian centers (Additional file 1), between September 2013 and May 2018.

Anthropometric measurements
Weight and height were obtained from the patient's medical records at the time of immunotherapy initiation. BMI was calculated using the formula of weight/height 2 (kilograms per square meter) and classified according to the World Health Organization (WHO) categories: underweight, BMI < 18.5; normal, 18.5 ≤ BMI ≤ 24.9; overweight, 25 ≤ BMI ≤ 29.9; obesity, BMI ≥ 30. For the study purpose, the binomial cut-off for BMI </≥ 25 was used, and patients were categorized into non-overweight (< 25) and overweight/obese (≥ 25) for the final analysis. Underweight patients were included in the non-overweight group.

Study design
We conducted a "real-life", multicenter, retrospective observational study aimed at comparing the clinical outcomes of cancer patients treated with ICIs according to baseline BMI levels.
Primary outcomes measures were: objective response rate (ORR), time to treatment failure (TTF), PFS and OS. ORR was defined as the proportion of patients experiencing an objective response (either complete response or partial response) as best response to immunotherapy. TTF was defined as the time from treatment's start to discontinuation for any reason. Progression-free survival (PFS) was defined as the time from the start of immunotherapy to the date of disease progression or death, whichever occurred first. Patients who were alive without disease progression were censored on the date of their last disease assessment. Overall survival (OS) was defined as the time from the start of immunotherapy to death. Patients who were still alive were censored at the date of last contact. Patients were treated according to the tumor type indication with pembrolizumab, nivolumab or atezolizumab with standard doses and schedules.
A subgroup analysis comparing clinical outcomes in males and females patients, using the binomial cut-off (BMI </≥ 25) was also conducted as secondary analysis.
To determine ORR and PFS, scans were reviewed by a dedicated thoracic oncologist at each Institution using Response Evaluation Criteria In Solid Tumors (RECIST) version 1.1. [14]. χ2 was used to compare ORR and incidence of irAEs among subgroups [15]. In the multivariate analysis, logistic regression was used to evaluate the role of parameters proven to be significant at the univariate analysis of ORR [16]. Median TTF, median PFS, and median OS were evaluated using the Kaplan-Meier method [17]. Median follow-up was calculated according to the reverse Kaplan-Meier method [18]. Cox proportional hazards model [19] was used to evaluate predictor variables in univariate and multivariate analysis for TTF, PFS and OS. The data cut-off was October 29 th , 2018. All statistical analyses were performed using MedCalc Statistical Software version 18.6 (MedCalc Software bvba, Ostend, Belgium; http://www.medcalc.org; 2018).

Patient characteristics
Nine hundred and seventy-six, consecutive advanced cancer patients were evaluated. Patient characteristics are summarized in

Safety analysis
In the entire cohort, 393 patients (40.3%) experienced irAEs of any grade. Sixty-three patients (6.5%) experienced G3/G4 irAEs. Overweight/obese patients were significantly more likely to experience any grade irAEs compared to non-overweight patients (55.6% vs. 25.2%, p < 0.0001). However, no difference in the rate of G3/G4 irAEs was observed between Overweight/obese patients and non-overweight patients (7.6 vs. 5.3%, p = 0.1338). The safety profile of ICIs according to BMI is summarized in Additional file 2.

Activity analysis
Univariate and multivariate analyses for ORR are detailed in Additional file 3. Among 910 patients evaluable for activity, 283 patients had a response to ICIs (ORR: 31.1%). Overweight/obese patients had a significantly higher ORR compared non-overweight patients (41.3% vs. 20.9%, p < 0.0001). Similarly, we found a significantly higher ORR among patients who experienced at least 1 irAE compared to those without irAEs (45.1% vs. 21.1%, p < 0.0001). Both BMI (overweight/obese vs. non-overweight) and the development of irAEs of any grade, were independently associate with higher ORR in the multivariate analysis (p = 0.0239 and p < 0.0001, respectively).
When these outcomes where analyzed according to BMI, we found that median TTF was significantly longer in overweight/obese patients compared to non-overweight patients ( (Fig. 1c).
After adjusting for PS, treatment line, n°of metastatic sites, gender, primary tumor subtype and development of irAEs, a BMI of ≥25 retained a significant association with a longer TTF (p = 0.0009), PFS (p < 0.0001) and OS (p < 0.0001) in multivariate models (Table 2, Table 3, Table 4) Table 5 reports the univariate and multivariate gender analyses for TTF, PFS and OS of male patients (Table 5A) and female patients (Table 5B). As shown overweight/obese male patients had significantly longer TTF (p = 0.0330) and OS (p = 0.0013), but not PFS (p = 0.0668), when compared with non-overweight patients, while overweight/obese female patients had significantly longer TTF (p = 0.0037),    Table 6 reports the univariate and multivariate analyses of TTF, PFS and OS, comparing overweight (non-obese) patients and obese patient with non-overweight patients. Figure 2 reports the Kaplan-Meier survival curves of obese, overweight and non-overweight patients.

A B C
When we analyzed the clinical outcomes of normal weight vs. underweight patients, we found a significantly longer median TTF ( Table 7 reports the univariate and multivariate analyses of TTF, PFS and OS, comparing overweight (non-obese) patients and obese patients with normal weight patients. Figure 3 reports the Kaplan-Meier survival curves of obese, overweight and normal weight patients.

Discussion
In this study we demonstrated that patients with a BMI ≥ 25 experienced a better clinical outcome compared to those with a BMI < 25. Recently, the association between BMI and OS of metastatic renal cell carcinoma patients, has been reported regardless of the use of anti-PD-1/PD-L1 therapy [4,20]. However, in our study we found a strong correlation between overweight and improved clinical outcomes with anti-PD-1/PD-L1.
Some authors have already speculated about the negative impact of body composition alteration on immune cells activity [21]. Interestingly, it has been increasingly recognized that white adipose tissue, which is the most related to the fattening process [22], is also involved in the induction and/or coordination of host defenses, being a source of cytokines and chemokines [23]. In fact, adipose tissue modulates the Th1/Th2 balance, decreases  [24][25][26]. Moreover, a recent preclinical study revealed that white adipose tissue might also play a role in immune homeostasis [27]. In this study, white adipose tissue of mice was reported to accumulate pathogen-specific memory T-cells after a microbial infection, including tissue-resident cells expressing a distinct metabolic profile. Intriguingly, these data support the hypothesis that adipose tissue can act as a reservoir of tissue-specific memory T-cells, which can undergo a rapid response to reactivation against exogenous stimuli. This evidence raises an interesting question, can these adipose tissuespecific T-cells be promptly reactivated against cancerspecific antigens as they do against microbial antigens?
In a recent meta-analysis of patients with immune-mediated inflammatory diseases treated with anti-TNF (tumor necrosis factor), the authors reported a trend towards a lower response rate to treatment among overweight patients [28]. This is likely to reflect the reduced responsivity of T-cells of obese individuals, which has also been confirmed in preclinical models showing a significant increase in dysfunctional exhausted T-cells in obese mice [9]. Nevertheless, such inflamed and immune-exhausted status may be more likely susceptible to the immune checkpoint blockade. In support of this, in preclinical models, T-cell dysfunction in obese mice was proven to be partly mediated by the PD-1 axis and driven by leptin, strengthening the already known correlation between JAK/STAT pathway and immune checkpoint inhibition [9,29].     Importantly, in our study we also found a significantly higher incidence of irAEs of any grade among overweight/ obese patients. In light of the emerging association between the development of irAEs and improved clinical outcomes with ICIs across different tumor types, our findings are not unexpected [30][31][32][33][34][35]. In our cohort, the development of irAEs of any grade was independently associated with improved clinical outcomes along with a BMI ≥25 in multivariate analyses.
The analysis performed by separating overweight and obese patients, demonstrated that a linear relationship between BMI and positive outcomes cannot be assumed. Even though we found no statistically significant differences in TTF, PFS and OS between overweight and obese patients, when separately comparing obese patients to non-overweight patients (Table 6), we observed the loss of significance regarding TTF and PFS, while not regarding OS. This result is of particular interest, considering the possibility of a negative impact on survival of obese patients due to cardiovascular and metabolic complications of obesity itself. Noteworthy, the HRs are concordantly lower for overweight (non-obese) patients in each survival analysis, compared to obese patients, thus supporting the hypothesis that the prognostic weight of obesity, could have partially influenced the final results.
On the other hand, despite the small sample size (4.1% of the entire population), underweight patients had significantly shorter TTF, PFS and OS, when compared to normal weight patients, confirming that malnutrition (and cachexia) is an independent negative prognostic factor. Nonetheless, when we compared obese and overweight patients (Table 7) with normal weight patients we observed significantly improved clinical outcomes favouring the overweight group, which suggests that overweightness has a direct impact on the efficacy of ICIs.
In our study we also carried a gender-based analysis. Previously, it has been reported that female patients tend to have lower benefit from ICI compared to males [36,37]. However, whether the gender plays a key role in determining the clinical outcome to immunotherapy is still in need of further investigation. In our study, we found that overweight female patients derived a greater clinical benefit form immunotherapy as compared to the male counterpart ( overweightness was associated with improved outcomes in both males and females in the multivariate analysis. This led us to speculate that in our population the predictive role of BMI was to be stronger than the predictive role of gender. Certainly, the relationship between sex, adipose tissue and immunity is complex and ambiguous. Sex-hormones, in particular estrogens, could affect adipose tissue functions [38], but in some respects their influence on the immune systems does not seem unidirectional [39]. Furthermore, the median age of female patients in our study was 67, indicating a prevalence of postmenopausal patients. In this specific population the adipose tissue becomes a major source of circulating estrogens [40].
Another way to explain how BMI might affect sex hormones levels and the immune response, is through diet regimens which underpin the weight gain. Indeed, gut microbiota may be influenced by the different "modifying pressures" of various diet types. Interestingly, males and females have recently been reported to have gender-specific differences in their immune system and gut microbiota composition. Whether these differences in gut microbiota composition might impact the efficacy or the safety profile of immunotherapy is subject of intense research and is expected to provide us further insight in the optimal management of our patients [41][42][43].
Our study is certainly flawed by several caveats, including the retrospective design with the risk of selection and data collection biases, the heterogeneity of the analyzed population, the lack of a centralized imaging review for response assessment and the lack of data about patient comorbidities. In addition, the lack of control group of patients who did not received ICIs further limit the power of our analysis. On the other hand, a unique strength of our study is that we evaluated the predictive role of baseline assessment of BMI in a "real life" population of individuals candidate to receive ICIs.