Association of Personal Characteristics and Effectiveness of Immunotherapy in Late-Stage Non-Small Cell Lung Cancer: A Systematic Review

Abstract Background Although immunotherapy can increase survival in non-small cell lung cancer (NSCLC), response rates are low. It is unclear which characteristics contribute to variability in immunotherapy efficacy and survival. Research is needed to identify reasons for heterogeneity in response rates to better tailor treatments. Methods Web of Science, Ovid EMBASE, and MEDLINE were queried from 2013 to January 2021, and all studies reporting overall or progression-free survival for patients treated with immunotherapy for NSCLC of at least stage IIIB were screened. Results Included were 18 randomized controlled trials (RCTs; 6534 immunotherapy RCTs; 11 192 nonimmunotherapy RCTs) and 16 observational studies (n = 9073 immunotherapy patients). Among RCTs, there was improved survival with the addition of immunotherapy in patients aged younger than 65 years in 10 of 17 studies; smokers in 8 of 15 studies; and males in 10 of 17 studies and 6 of 17 females. Only 5 studies reported outcomes by race. Among observational studies, younger patients (aged younger than 60, younger than 65, or younger than 70 years in most studies) had better survival than older patients (aged 60 years and older, 65 years and older, or 70 years and older) in 4 of 13 studies, ever-smokers in 7 of 13, and females in 2 of 14. Three studies reported race with mixed results. Conclusion Although evidence is mixed, younger patients, smokers, and males may derive more benefit from immunotherapy. Evidence on racial differences is limited. Physicians should be mindful of personal characteristics when formulating treatment plans. Further research is needed to understand underlying mechanisms and to identify the best immunotherapy candidates and alternative treatments for those unlikely to benefit.

Immunotherapy is now the standard of care for many of the approximately 200 000 non-small cell lung cancer (NSCLC) patients diagnosed annually in the United States (1). For some late-stage, unresectable cancers, immunotherapies can be less toxic and more targeted than chemotherapy. Multiple approaches combining immune checkpoint inhibitors with chemotherapy and radiotherapy continue to be tested (2). Compared with other cancers, NSCLC is relatively amenable to immunotherapy, but many patients still do not benefit. Reported response rates for pembrolizumab and nivolumab for patients with Programmed death-ligand 1 (PD-L1) of more than 50% are 41% and 19%, respectively, and as low as 15% for atezolizumab (3). Research is ongoing to identify personal characteristics, biomarkers, and clinical or histological features that may explain heterogeneity in immunotherapy response and predict which patients may benefit most.
However, much remains poorly understood about how various prognostic factors contribute to differences in efficacy and disparities in survival among those receiving immunotherapy, especially in the presence of clinical characteristics that can act as confounders. There is substantial heterogeneity of clinical characteristics, tumor mutation expression, and line of therapy between existing studies, which heightens the need for synthesis across experimental and observational studies. Patients enrolled in clinical trials tend to be White, younger, healthier, and more tolerant of treatment-related toxicities (4). Women have long been underrepresented in clinical trials (4), as have minority groups, and there has not been a systematic attempt to

Data Collection and Extraction
Search results were exported into EndNote X9 (Clarivate, Endnote X9) (7), and all duplicates were removed. Titles and abstracts were assessed for eligibility, and full-text articles were referenced when no exclusion criteria were met during this initial screen.
For experimental studies, the following data were extracted: prognostic factor of interest, hazard ratio for all patients, hazard ratio stratified by prognostic factor subgroups, histology, line of treatment, stage, and immunotherapy and control treatments. For immunotherapy groups, the number of patients, percent male, and median age were extracted. For observational studies, extracted data included the following: prognostic factor of interest, immunotherapy used, lines of treatment, survival outcome (OS or PFS), univariate hazard ratio, multivariate hazard ratio, covariate adjustments, stage, histology, and median age.
Data extraction forms for age, smoking status, sex, and race differed slightly based on the way each prognostic factor was categorized. Data about race were not extracted if more than 90% of study participants were of one ethnicity, and hazard ratios for minority subgroups were not reported. Data on race extracted as "Other" refers to studies where "Other" was used by the authors as a category without further breakdown. Quality review of all included experimental and observational studies was completed by 2 reviewers who scored the studies independently, using National Institutes of Health-developed tools (8,9). All discrepancies were resolved between the 2 reviewers, and a third, independent reviewer confirmed the final judgements (Supplementary Tables 1 and 2, available  online).

Search Results
The search and eligibility screening process is illustrated in Supplementary Figure 1 (available online). After removing duplicates, 1124 titles and abstracts were screened. The full texts of 65 studies were assessed along all inclusion and exclusion criteria; 18 experimental studies (all randomized controlled trials [RCTs]) and 16 observational studies met all inclusion criteria (Tables 1 and 2).

Quality Assessment
All 18 included experimental studies fulfilled at least 10 of 14 checklist items; all were randomized, had groups comparable across all covariates at baseline, and used reliable outcome measures (9). Although many of the included studies were open-label and unblinded, the focus on hard survival outcomes reduces the potential for bias in outcome measurements. Multiple studies did not report details about allocation concealment (Supplementary Table 1, available online) (15,16,18,23,24). All 18 experimental studies specified a priori any patient subgroup analyses for survival outcomes.
All 16 observational studies met 10 of 14 checklist items (9). Areas of potential bias included retrospective analyses, lack of sample size justifications, and sparse reporting around whether hypotheses were specified a priori (Supplementary Table 2, available online).

Age
Of the experimental studies, 17 reported data across age groups (  immunotherapy for patients aged younger than 65 years, compared with 3 studies for patients aged 65 years and older. Six studies reported age as a trinomial variable (16,18,19,21,22,25); 2 of which reported statistically significantly improved survival with immunotherapy for those aged younger than 65 years and those 65-75 years (Table 3) (19,21). One study reported improved survival only in patients aged younger than 65 years (22), and 1 did not report confidence intervals by age (18). No study reported statistically significant OS or PFS improvements with immunotherapy for patients aged older than 75 years.

Race
Of the 5 RCTs that reported OS hazard ratios by race (16,17,(24)(25)(26), only the PACIFIC study (durvalumab and chemoradiotherapy combination) (26) reported statistically significant improvement with immunotherapy. Asian and White patients derived similar benefit from immunotherapy, but the improvement was statistically significant only for White patients (26).
Two of the remaining studies reported hazard ratios for Asian, Black, and White patients (16,25), and 1 reported this data for only Asian and White patients (17). Distinctly, Barlesi et al. (24) reported race either as Hispanic or Latino, Japanese living in Japan, non-Hispanic or non-Latino, or Other. There were no statistically significant differences in OS in any of these studies (Supplementary Table 3, available online).

Narrative Synthesis of Observational Studies
Of the observational studies where all patients received immunotherapy, 16 assessed at least one of age, smoking status, sex, or race. Six studies reported hazard ratios based on PFS, and 10 studies reported hazard ratios for OS (Table 2). Pembrolizumab and nivolumab (PD-1 inhibitors) were the most frequently studied immunotherapy agents. Both were used in all lines of treatment as monotherapies (5,33,39) and as part of combination therapy (31). Atezolizumab, ipilimumab, durvalumab, sintilimab, and tislelizumab were the other evaluated immunotherapy regimens.

Age
Of the 13 studies reporting hazard ratios for age, 4 reported statistically significantly better OS or PFS for younger patients compared with older patients (  (29). However, in a US-based study (n ¼ 5807) of stage IV NSCLC patients who received first-line immunotherapy treatment, those aged 18-59 years had longer overall survival than older patients (41). A study of 61 hospitalized patients in China treated with pembrolizumab, nivolumab, atezolizumab, or ipilimumab (28% first-line; 72% second-line and later) showed statistically significantly worse survival for patients aged 65 years or older compared with those aged younger than 65 years, before and after adjustments for metastasis, adverse effects, line of therapy, and other biomarkers (32). In a multivariate model, Lichtenstein et al. (30) found immunotherapy patients aged older than 80 years had worse OS than those aged younger than 60 years. However, no statistically significant differences in OS were observed for patients in other age groups.

Smoking Status
Of 13 studies, 8 reported statistically significant OS or PFS differences (either univariate or multivariate) for smoking status.  Seven showed that among immunotherapy recipients, former and current smokers had statistically significantly better OS and PFS than never-smokers (29,33,35,37,39,40,42). Distinctly, only 1 study (n ¼ 101) of stage III-IV NSCLC patients receiving third-line, immunotherapy with anlotinib combination therapy showed statistically better PFS for never-smokers relative to ever-smokers (Table 7) (31). The 2 largest studies examining multiple immunotherapy agents showed statistically significantly better survival for smokers than nonsmokers (29,35). In Smit and colleagues' (29) analysis of Dutch patients (n ¼ 2302), nonsmokers had statistically significantly worse survival than smokers. In the Japanese Okayama Lung Cancer Study of NSCLC patients in all lines of treatment with metastatic or recurrent disease, univariate results were not statistically significant. However, after adjustments, never-smokers had statistically significantly worse survival than former and current smokers (35).

Sex
Of 14 studies reporting on sex, 12 found no statistically significant differences in OS or PFS in males compared with females treated with immunotherapy (Table 8). However, the largest study (n ¼ 5807) of patients receiving first-line immunotherapy found females had better survival than males (41). Another retrospective analysis of patients (n ¼ 257) treated with singleagent pembrolizumab, nivolumab, or atezolizumab showed statistically significantly better OS for females than males only in the adjusted model (5).

Race
Only 3 observational studies assessed the association between survival and race, among immunotherapy patients (Supplementary Table 4, available online) (5,40,41). Foster et al. (41) found non-White patients had statistically significantly better OS than White patients. Nazha et al. (5) observed no differences in OS by race. A univariate analysis of patients in Colorado and Shanghai with targetable driver mutations in all treatment lines showed better survival for non-Asian compared with Asian patients, but the difference was not statistically significant in the multivariate analysis (40).

Discussion
This systematic review was performed to evaluate whether age, smoking status, sex, and race modify the effectiveness of immunotherapy in late-stage NSCLC and to assess variability in survival among patients who received immunotherapy. We reviewed 18 experimental studies comparing immunotherapy and nonimmunotherapy groups (n ¼ 6534 and 11 192, respectively) and 16 observational studies of 9073 patients who received immunotherapy. Because of study design differences, results from experimental and observational studies are complementary, rather than comparable. Experimental hazard ratios reflect benefit from the addition of immunotherapy (compared with those without) and allow us to assess factors that may modify immunotherapy efficacy. Immunotherapy was universal across patients in observational studies, which focus on directly comparing survival among patient subgroups. As such, observational studies may provide insight into whether differential efficacy of immunotherapy contributes to survival disparities.
Experimental studies generally suggested that younger patients derive more benefit from immunotherapy than their older counterparts. In particular, survival outcomes of patients aged 75 years and older tended to be less favorable, and trends were less consistent than for patients aged younger than 75 years. For subgroups of patients aged younger than 65 years, all experimental studies except CheckMate 026 (20) showed a trend toward improved survival with receipt of immunotherapy, even if the improvement in survival was not statistically significant. Likewise, with the exception of the CA184-104 trial (25), all subgroups of patients aged 65-75 years trended toward receiving benefit from immunotherapy. This is consistent with research showing older adults are more susceptible to the onset of immunosenescence and reduced intrinsic immunity (43). Increased age is also linked to a longer period of carcinogenesis as well as increased vulnerability and sensitization of cells to environmental carcinogens (44,45).
However, when directly comparing survival by age group, the majority of observational studies found no statistically significant difference. This is somewhat paradoxical, as younger lung cancer patients generally survive longer (44,46). The 2 largest observational studies had inconsistent findings, although they assessed different treatment lines. This may suggest that the marginal benefit of immunotherapy by age may differ across treatment lines. More uniform reporting of outcomes would be needed to address this hypothesis.
For smoking status, experimental studies more frequently showed increased survival benefit in smokers compared with nonsmokers. Likewise, most observational studies reported improved survival for current or former smokers but not neversmokers. These results are consistent with previous reviews and may reflect the fact that smokers tend to have more tumor mutations, corresponding to greater immunogenicity and a higher likelihood of immune cell recognition of tumor cells (47). For instance, whole-genome sequencing revealed smokers had a tenfold higher mutation frequency than never-smokers (48). Considering nonsmokers tend to have fewer comorbidities and are generally healthier than smokers (49), this suggests nonsmokers may not be ideal candidates for immunotherapy, especially in the absence of a high tumor mutational burden.
Two clinical trials-KEYNOTE-189 (13) and PACIFIC (26)yielded statistically significant improvement in OS for neversmokers, and more so than in ever-smokers. These findings differ from the majority of included studies and may be partially attributable to the unique characteristics of the trials. KEYNOTE-189 was the only trial to have an immunotherapy intervention arm with both pembrolizumab and pemetrexed (13). The PACIFIC trial, which enrolled patients in second-line and later treatment lines, was the only included trial to enroll patients with unresectable stage III NSCLC or to assess either durvalumab or chemoradiotherapy (26). As such, combination therapies like pembrolizumab and pemetrexed as well as durvalumab and chemoradiotherapy ought to be further examined as potential options for immunotherapy treatment of patients who may otherwise be poor responders. Findings around sex were mixed, reflecting the ongoing debate in the current literature (50). Among the 17 experimental studies reporting data stratified by sex, 10 reported statistically significant improvements with the addition of immunotherapy for males, compared with only 6 for females. However, every experimental study trended toward survival improvement for males with the addition of immunotherapy, even when results did not rise to the level of statistical significance. Results for female patients were less consistent, although trends from most studies did suggest some potential benefit. Notably, the KEYNOTE-189 trial, which uniquely evaluated pembrolizumabpemetrexed combination therapy, reported especially high survival gains for females (HR ¼ 0.29) that exceeded those seen in male patients (HR ¼ 0.70) (13). Given the different trends for KEYNOTE-189 with regard to both smoking status and sex, this combination may warrant future examination. Most observational studies found no statistically significant difference in survival by sex, although the largest found statistically significantly better survival for females, compared with males, consistent with previous research (51). Although there is uncertainty, one interpretation of these findings is that even if males derive more benefit from immunotherapy than females, the increased benefit is insufficient to overcome the existing female survival advantage (41,51,52). Although a specific mechanism for NSCLC is still unclear, sex hormones such as estrogen and testosterone, which have immunogenic and immunosuppressive properties, respectively, are known to modulate gene expression, immune system agents, and treatment-related adverse effects (53,54). Additional research is needed to further examine moderating factors that may favor positive prognoses by sex. Moreover, only 1 clinical trial had approximately equal enrollment of males and females (21): the rest were at least 59% male, with 6 trials enrolling 70% males or more. Therefore, these results reinforce the need for proper representation of females in clinical trials to better evaluate how sex may affect survival and response to immunotherapy.
Only 5 experimental studies and 3 observational studies provided race information. More data is needed on outcomes according to race in clinical trials, considering recent evidence of differential tumor mutation burden (55,56). Trials reporting outcomes by race were severely underpowered because of their disproportionate underenrollment of non-White patients. Despite representing 13% of the US population, Black patients comprised no more than 4% of any clinical trial. Although in Foster and colleagues' study (41), non-White patients survived longer than White patients, the classification of non-White makes interpretation difficult and further highlights the need for more diverse, inclusive clinical trials with details on patients' ethnic and racial backgrounds.
This review highlights several gaps in current research on immunotherapy efficacy. There was uneven reporting of univariate and multivariate hazard ratios among the included observational studies. Even when multivariate hazard ratios were reported, not all studies specified adjustment variables, and no 2 studies used the same set of covariate adjustments. To minimize some of this variation, specific covariates that will be adjusted for in multivariate hazard ratios or adjusted risk calculations should be chosen and reported a priori. Establishing a consensus around adjustments will facilitate comparisons between different studies. Lastly, more frequent reporting of interaction terms (ie, age/sex/smoking status/race*immunotherapy) in observational studies may allow for more direct comparison with experimental studies and provide more information about effectiveness outside of a clinical trial setting, where male, younger, and healthier patients are frequently overenrolled (57). As such, greater consistency in methodology and reporting of risk calculations can further the understanding of clinical confounders that may contribute to differential survival outcomes, and ultimately, help tailor patient therapies. To our knowledge, this systematic review is the first to aggregate evidence across both experimental and observational studies and analyze outcomes with respect to multiple personal characteristics, including race. Although head-to-head comparisons with previous reviews are difficult because of differences in inclusion and exclusion criteria and patient populations, many of our findings from observational studies pertaining to age and smoking status align with those in previous reviews of experimental studies (47,(58)(59)(60).
This review is not without limitations. Because of the heterogeneity in patient populations, immunotherapies administered, and uneven reporting of outcomes, meta-analysis was not possible. Moreover, although there were a couple of national database studies that met all inclusion criteria, most observational studies were retrospective analyses of patient cohorts at individual hospitals. Therefore, some studies may have limited generalizability to the broader patient population receiving immunotherapy. Publication bias toward negative findings is also a concern as both observational studies and clinical trials that fail to identify statistically significant differences or associations are less likely to be published.
Overall, evidence is mixed around which personal characteristics may be associated with increased benefit from immunotherapy treatment in advanced NSCLC. In aggregate, the findings from this review confirm those of previous reviews and individual studies, which suggest immunotherapy may increase survival in younger patients, smokers, and males.
We also highlight gaps in the literature, as very few experimental and observational studies have reported outcomes by race and those which have are severely underpowered. Further research into the moderating effects of race on immunotherapy efficacy can fill this gap while potentially creating more opportunities to better target immunotherapies and improve clinical care. To further explore disparities in immunotherapy effectiveness, future research should also include patient-level analyses of national registries (41). This will enable direct adjustment for individual characteristics and assessment of treatment interactions in patients more representative of the general late-stage NSCLC population.

Funding
Not applicable.

Notes
Role of the funder: Not applicable.
Disclosures: All authors have no conflicts of interest, disclosures, to report for this work.