In this study, both males and female stage IV NSCLC patients treated with chemoimmunotherapy were found to have better overall survival compared to those treated with chemotherapy alone. In the overall patient cohort, benefit from the addition of immunotherapy was similar between sexes. However, when stratifying by the main histological subtypes of NSCLC, we did not observe any difference in immunotherapy efficacy between male and female adenocarcinoma patients, although our results preliminarily suggest a greater degree of benefit in male versus female squamous cell patients. Lung adenocarcinoma and squamous cell carcinoma, while subtypes of the same disease, are in fact known to vary dramatically both in terms of their genomic and clinical attributes . More recently, key differences between these histological subtypes has been observed in terms of immune host response, with implications for immunotherapy response . Sample size of this sub-group was small, however, especially in the propensity-matched cohort; replication of this finding in a much larger cohort of squamous cell patients is necessary.
In meta-analyses conducted by Conforti et al., they found that males derive significantly more clinical benefit than females when treated with immune checkpoint inhibitors for NSCLC, yet found the opposite to be true when studying the effects of the addition of immunotherapy to chemotherapy, reporting pooled survival estimates of 0.76 (95% CI = 0.66 to 0.87) and 0.48 (95% CI = 0.35 to 0.67) for men and women, respectively [12, 14]. By comparison, our analysis showed the addition of immunotherapy treatment as having a smaller impact on survival, and we did not find any evidence of female NSCLC patients deriving more benefit from chemoimmunotherapy. Estimates from Conforti et al. are based on results of randomized clinical trials, the patients of which are traditionally known to differ both clinically and demographically from patients in real-world settings. The authors also acknowledge as a major limitation the use of data extracted from published studies, which precluded them from being able to adjust for patient-level characteristics that differ by sex and may affect treatment outcomes. Our study used population-based, patient-level data that included key clinical and demographic information that allowed for adjustment for potential confounders, such as age, comorbidity, and income, using several different statistical modeling techniques. Consideration of these confounders is critical when investigating the impact of sex on immunotherapy response; we show here the tremendous importance of taking into account histological subtype, which isn’t possible in most meta-analyses. Moreover, even when clinical trials are stratified by histological subtype, such as the KEYNOTE-407 trial investigating pembrolizumab plus chemotherapy in patients with metastatic squamous NSCLC, female patients are often dramatically underrepresented . In this recent trial, just 19% of study participants were female . While a limitation of our analysis is that only 2015 data was available, with FDA-approval of immunotherapy treatment for NSCLC beginning later in that year, the high percentage (47%) of chemoimmunotherapy patients being female attests to the enrichment of real-world patients over RCT participants in our data. Using real-world, patient-level data is necessary is necessary to more fully understand disparities in chemotherapy plus immunotherapy response, and we hope to see this work replicated in larger, more recent datasets in future.
NCDB is in fact one of the first large clinical datasets to include data for immunotherapy treatment along with long-term survival data. A significant limitation of immunotherapy trials is the lack of long-term follow up, resulting in the use of surrogate endpoints such as response rates and progression free survival to estimate benefit,  which may not be entirely representative of actual benefit . We consider it a notable strength of our study that NCDB has highly accurate mortality data that spans multiple years on a large sample size, creating an opportunity to assess the true benefits of immunotherapy in specific patient subsets.
While our results did benefit from the comprehensiveness of the NCDB data, there were important variables missing from the data, the exclusion of which may be influencing the results. Specifically, we did not have key elements such as ECOG status; if patients with lower ECOG status were less likely to receive immunotherapy, the results could be biased towards showing the combination of immunotherapy + chemotherapy to be more efficacious than it actually is. We were able, however, to adjust for comorbidity status which should be a good proxy for EGOG performance. In addition, patients’ sex is not known to be associated with ECOG status,  thus ECOG should not have affected the observed relative difference in treatment efficacy between the sexes. Because the NCDB is a clinical database sourced from registry data, other potential confounders, such as EGFR, ALK, ROS mutational status, which are known predictive biomarkers for NSCLC, and differ by sex, were missing in the NCDB. Since women are more likely to harbor these mutations  the inability to adjust for these markers is a considerable limitation. However, the incidence of driver mutations in squamous cell lung cancers is extremely low,  therefore EGFR/ALK/ROS1 mutations should have no impact on the findings reported for the stratified analysis on squamous cell lung cancer. Smoking status, too, was missing from the NCDB data. While all histologic types of lung cancer are significantly associated with cigarette smoking, this association is strongest for squamous cell carcinoma . Additionally, smoking history is more commonly reported for male patients than female. Stratifying according to histology and sex should reduce some bias, but we acknowledge that smoking status is an important potential confounder we could not fully adjust for. We were also unable to study the role of tumor mutation burden and the tumor microenvironment, which may be impacted by smoking . While the results reported here of our analysis using individual patients’ data are meant to overcome some of the limitations of the previous meta-analyses conducted on this topic, future research is needed to explore how some of the covariates lacking in the NCDB might influence immunotherapy response according to sex. As larger, more detailed datasets containing immunotherapy information become available, we recommend that this analysis be repeated. Use of electronic health registry data, for example, might be one avenue to overcome these limitations.
Another major limitation to this work is that the specific type of immunotherapy agent used was not reported in the NCDB. Moreover, several of the drugs included in the NCDB immunotherapy variable are not FDA-approved or conventional immunotherapies, although all agents included have been shown in preclinical and/or clinical studies to modulate the anti-tumor immune response. Never-the-less, we have reason to believe that this patient population is enriched for those who did in fact receive immune checkpoint inhibitors. The patient selection helped us to reduce the heterogeneity of drugs included under the “immunotherapy” variable in the NCDB. By limiting the analysis to only those patients diagnosed in 2015, the year the FDA approved the first immunotherapy drugs for NSCLC treatment, we have enriched for patients receiving immune checkpoint inhibition drugs compared to patients diagnosed in previous years. We also tried to get an estimate of which types of drugs would have been available for clinical use in 2015 and onward, and how likely it is that patients included in the present analysis were treated with immunotherapeutic agents that are now excluded from clinical use. We observed that many of the immunotherapeutic treatments included as part of the NCBD’s immunotherapy variable have never been approved for use among NSCLC patients, making it unlikely that our patient population were treated with these. For instance, the NCDB immunotherapy variable is inclusive of EGFR antibodies, but these drugs were never FDA approved for standard of care therapy in NSCLC, so it is doubtful that the patients in our dataset received this form of treatment. Moreover, we completed an extensive search of the literature, including clinicaltrails.gov and PubMed, looking for clinical trials using EGFR antibody treatment for NSCLC during our study period . All trials inclusive of EGFR antibodies were conducted only in combination with chemotherapy treatment. We therefore conduced a sensitivity analysis on immunotherapy only vs chemotherapy only treatment, and found that there was no statistically significant interaction between sex and immunotherapy efficacy (p > 0.05, data not shown). Stratification according to histologic type should have also helped to provide a cleaner picture of immunotherapy efficacy by sex. In squamous cell carcinoma cases angiogenesis inhibitors are contraindicated due to increased risk of bleeding,  thus we expect that the squamous cell lung cancer patients recorded as treated with immunotherapy would not have received anti-VEGF treatment. In this stratified analysis we observed a difference in treatment efficacy between sexes when immunotherapy was added to chemotherapy. Ideally this analysis will be replicated in a larger patient-level dataset reporting specific immune checkpoint inhibitor information; it would be interesting to see if the difference in efficacy observed among squamous cell patients is seen among those with adenocarcinoma when immunotherapy drugs are limited concisely to immune checkpoint inhibitors. The lack of details on type of immunotherapy remains a notable limitation of our work, yet we believe that our findings contribute to the general understanding of the association of sex and immunotherapy, and raise important questions for future research.
The role of sex as it pertains to immunotherapy response remains an important clinical question. The patient-level nature of our data allowed us to also do a head-to-head comparison of survival outcomes of females vs males receiving immunotherapy in addition to chemotherapy. We found that females in fact experienced improved survival with immunotherapy compared to males, after adjusting for potentially important clinical variables. This finding may be expected given that females experience better cancer outcomes overall than males . Lung cancers of male and females are known to have distinctive clinical and biological differences, notably in terms of histology, driver mutations and smoking exposure history, [33, 34] all of which may impact immunotherapy efficacy. We would like to note, however, that the major aim this work is not demonstrate the value of chemoimmunotherapy treatment, generally or for specific subgroups, but instead to investigate how chemoimmunotherapy treatment efficacy may vary by sex and/or histology.