Patient characteristics
Table 1 summarises the characteristics of the two cohorts of patients. In the cohort of 312 patients, the median age was 62.6 years (±9.4) and 79.8% were male. A total of 288 patients (92.3%) had ECOG-PS 0/1, and 24 patients (7.7%) had ≥2. A total of 212 patients (67.9%) had a history of former or current smoking; 155 patients diagnosed with adenocarcinoma (49.7%), 142 patients with squamous cell carcinoma (45.5%), and 15 patients with other subtypes of NSCLC (4.8%). A total of 188 patients (60.3%) had distant metastasis (diagnosed as M1b/c) before immunotherapy initiation. A total of 280 patients (89.7%) were administered PD-1 agents, and the remaining 32 patients were administered PD-L1 agents. ICI was administered in first-line treatment in 103 patients (33.1%), second-line treatment in 159 patients (51.1%), and third-line or further treatment in 49 patients (15.8%); 215 patients (68.9%) were administered PD-1/PD-L1 single agents, 75 patients (24.0%) were administered chemo-immunotherapy combinations, and the other 22 patients (7.1%) were administered PD-1/PD-L1 inhibitors combined with other treatments including anti-angiogenesis, target TKIs or CTLA-4 inhibitors. The median baseline BMI was 23.9 kg/m2 (±3.3) and was categorised into the following four groups according to the WHO Chinese criteria: 19 patients in the underweight group (6.1%), 139 patients in the normal-weight group (44.6%), 122 patients from overweight group (39.1%), and 32 patients from obese group (10.3%).
The general characteristics (confounding factors) of 194 patients with documented weight changes during the first 12 weeks were recorded and compared with those of the total cohort of 312 patients using the chi-squared test to show the distribution of confounding factors in the two patient cohorts. The median age was 62.3 years (±9.1, P=0.694 compared with the total 312 cohort); 82.5% were male and 17.5% were female (P=0.532). 186 patients with ECOG-PS scores of 0/1 and 8 others with greater than or equal to 2 (P=0.157), and 134 patients (69.1%) were former or current smokers (P=0.868). Other aspects were also balanced between the two cohorts, including the specific histological types (P=0.609), status of distant metastasis (P=0.486), agent type of ICI (P=1.000), treatment line of immunotherapy (P=0.478), combination anti-tumour strategy (P=0.052), and baseline BMI (P=0.428 for median value, P=0.863 for categorised group). During the follow-up 12 weeks, the median variation in BMI was -0.3% (±4.8%). Patients were categorised based on the BMI variation during the first 12 weeks into three groups according to the following: 1) decrease group: BMI >0.5 kg/m2 loss; 2) stable group: BMI maintained within ± 0.5 kg/m2; and 3) increase group: BMI >0.5 kg/m2 gain. A total of 132 patients (68.0%) were categorised as the stable group, 34 (17.6%) as the decrease group, and 28 (14.4%) as the increase group.
Clinical outcomes analysis
Out of 312 patients, 94 achieved response (30.13%), and the cohort of dynamic 194 patients was 40.20% (78/194). Median follow-up was 13.68 months in the entire cohort and 35.97 months in survival patients; median OS and PFS for the 312 patients were 18.5 months (95% confidence interval [CI]: 18.5-22.0; 216 dead, 14 censored due to loss to follow-up) and 5.2 months (95% CI:4.3-6.4; 275 progression events), respectively. The 194 patients had a median OS and PFS of 25.2 months (95% CI:22.0-31.5; 117 deaths, no lost patients) and 6.6 months (95% CI: 5.7-8.2; 162 progression events), respectively.
Objective response rate
Table 2 summarises the ORRs of the two patient cohorts from the univariate analysis and the statistical differences between the different subgroups as analysed using the chi-square test. In 312 patients, the ORR was 36.8% in the underweight group (95%CI:16.3-61.6; 7/19 response ratio), 25.9% in the normal group (95%CI:18.8-34.0; 36/139 response ratio), 35.2% in the overweight group (95%CI:26.8-44.4; 43/122 response ratio), and 25.0% in the obesity group (95%CI:11.5-43.4; 8/32 response ratio). There was no statistical difference among the baseline BMI groups (P=0.320, χ2 test, normal group regarded as the comparator). However, among the groups of patients grouped according to the maximum change in BMI, the ORR for the decrease group was 14.7% (95%CI:5.0-31.1; 5/34 response ratio), stable group, 39.4% (95%CI:31.0-48.3; 52/132 response ratio), and increase group, 75.0% (95%CI:55.1-89.3; 21/28 response ratio), showing a significantly statistical difference (P<0.001, χ2 test, stable group regarded as the comparator).
Table 3 shows the multivariate regression analyses for ORR after adjusting for confounding factors: age (<60 vs. ≥60 years old), sex (male vs. female), ECOG-PS score (0–1 vs. ≥2), distant metastasis (yes vs. no), smoking status (yes vs. no), histological subtype (adenocarcinoma vs. squamous vs. other NSCLC), and the ICI agent type (PD-1 vs. PD-L1). In the baseline BMI group, neither the overweight/obesity group nor the underweight group showed significant differences compared with the normal-weight group. However, the increase in BMI during the first 12 weeks led to a significantly better outcome (hazard ratio [HR]=1.21, 95%CI:1.11-1.32, P<0.001) meaning that each percent increase in BMI was associated with a 21% increase in reaching CR or PR. Furthermore, the predictive value of the variation in BMI was validated in the grouping condition of the multivariate analysis. The increase group showed a significantly higher ORR than the stable group (OR=5.42, 95%CI:2.02-14.54, P<0.001), and the decrease group had a significantly lower ORR than the stable group (OR=0.31, 95%CI:0.11-0.89, P=0.030).
Progression-free survival
The univariate analysis of PFS is shown in Figure 2. Based on baseline body mass index (BMI), no significant difference in PFS was observed between the subgroups (P=0.43) (Figure 2a). Patients in the obese and overweight subgroups did not show longer PFS. However, among those grouped according to BMI changes, the survival curves for PFS were significantly longer in the increase group than those of the stable and decrease groups, with a statistically significant difference between the three groups (P=0.028) (Figure 2b).
Cox proportional hazards regression analyses of baseline BMI and BMI change, adjusted for the aforementioned confounders, are shown in Table 4. In subgroup analyses of baseline BMI, compared with the normal-weight group, the overweight, obese, and underweight groups dd not show significant differences in PFS (underweight: P=0.228; overweight: P=0.682; obese, P=0.514). When BMI variation was considered as a continuous variable, Cox proportional hazards regression showed that each percentage point increase in BMI during the first 12 weeks of immunotherapy was associated with a 7% reduction in the probability of disease progression, and this result was statistically significantly different (HR=0.93, 95%CI:0.89-0.96, P<0.001). Similarly, the increase group showed a significant advantage in PFS compared with that of the stable group after adjusting for confounding factors (HR=0.57, 95%CI:0.35-0.92, P=0.022). Although no statistical difference was found between the decrease and stable groups, there was a trend toward shorter PFS for patients in the decrease group, considering its small sample of patients (HR=1.14, 95%CI:0.73-1.79, P=0.557).
Overall survival
Figure 3a/3b shows the KM survival curve for overall survival in the univariate analysis of subgroups according to baseline BMI and BMI variation, respectively. The intergroup chi-square analysis of baseline BMI revealed a statistically significant difference in OS between the groups, with the baseline overweight and obese group having a better OS than the normal and underweight groups as shown by survival curves (P=0.022, χ2 test). The intergroup chi-square analysis based on changes in BMI during the first 12 weeks showed statistical difference in OS which was more significant between groups, with OS in the increase group superior to that of the stable group, and the stable group OS superior to that of the decrease group in the KM survival curve (P=0.0033, χ2 test).
Table 5 presents the multivariate analysis utilizing Cox proportional hazards regression analyses of the association between baseline BMI and BMI variation with OS. In the subgroup analysis of baseline BMI, when normal-weight groups were considered as the reference, only the overweight groups showed a significant benefit in OS (HR=0.67, 95%CI:0.49-0.91, P=0.011). In the subgroup analyses of BMI variation, compared with the stable group, a more significant OS benefit was seen in the increasing weight group (HR=0.38, 95%CI:0.20-0.74, P=0.004). However, the decrease group did not reach a statistically significant difference. During the first 12 weeks of immunotherapy, there was a statistically significant difference in the likelihood of death by 9% for every percentage point increase in BMI variations when BMI variation was taken into account as a continuous variable, according to Cox proportional hazards regression (HR=0.91, 95%CI:0.87-0.96, P<0.001).
Finally, we analysed the curvilinear relationship between baseline BMI and the variation in BMI with OS (Supplement Figures 1-2); however, no significant difference was observed (baseline BMI: P/non-linear=0.128; BMI variation: P/non-linear=0.520). Therefore, a linear relationship was still observed in our cohort.