Real-World Use and Survival Outcomes of Immune Checkpoint Inhibitors in Older Adults With Non–Small Cell Lung Cancer

Abstract

BACKGROUND:

Limited data exist regarding the characteristics and survival outcomes of older adults with non–small cell lung cancer (NSCLC) who receive immune checkpoint inhibitors in routine oncology practice.

METHODS:

Using the Surveillance, Epidemiology, and End Results–Medicare linked database, we identified 1256 patients aged ≥65 years who were diagnosed with pathologically confirmed stage I to stage IV NSCLC between 2002 and 2015 and initiated nivolumab or pembrolizumab in 2016. We examined patient characteristics and overall survival from the time of immune checkpoint inhibitor initiation through December 31, 2017.

RESULTS:

The median patient age at the time of immune checkpoint inhibitor initiatiton was 75.3 years (interquartile range, 8.5). A substantial percentage of patients were initially diagnosed with stage IV disease (42.6%) and had ≥2 comorbid conditions (48.7%). Using a claims-based proxy, 11.5% of patients had poor performance status and 12.6% had a history of autoimmune conditions. The median overall survival after initiation of immune checkpoint inhibitor was 9.3 months (95% CI, 8.5–10.5 months). The 1-year survival rate was 43.0% (95% CI, 40.2–45.7%). In multivariable analyses, multiple comorbid conditions, squamous histology, a history of nonplatinum doublet systemic therapy, recent radiotherapy, and a shorter time from initial diagnosis to treatment initiation were found to be statistically significantly associated with an increased hazard of death. Demographics, poor performance status, and prior autoimmune conditions were not significantly associated with the hazard of death.

CONCLUSIONS:

Many older adults with NSCLC who initiated immune checkpoint inhibitors had multiple comorbidities, a history of autoimmune disease, or poor performance status. Factors associated with poor prognosis among patients with advanced NSCLC were also associated with worse survival in older adults treated with immune checkpoint inhibitors.

INTRODUCTION

The introduction of immune checkpoint inhibitors has changed the treatment approach in patients with advanced non–small cell lung cancer (NSCLC). In 2015, the US Food and Drug Administration approved the first 2 programmed cell death protein 1 (PD-1)–targeting monoclonal antibodies, nivolumab and pembrolizumab, for patients with advanced NSCLC. In the CheckMate-057 and CheckMate-017 trials, nivolumab significantly improved overall survival compared with docetaxel among patients with advanced NSCLC with disease progression on or after platinum-based chemotherapy.^1,2 The median overall survivals in patients who received nivolumab were 12.2 months and 9.2 months, respectively, for those with nonsquamous and squamous histology.^1,2 In the KEYNOTE-010 trial, pembrolizumab improved overall survival by 1.9 months compared with docetaxel among patients with at least 1% programmed death–ligand 1 (PD-L1) expression.³ The median overall survival in patients treated with pembrolizumab was 10.4 months.³

These seminal clinical trials demonstrated the clinical efficacy of immune checkpoint inhibitors. However, to the best of our knowledge, little is known regarding the characteristics and prognosis of older adults receiving these therapies in routine oncology practice because they were underrepresented in trials.^1–4 Furthermore, patients with a poor performance status and autoimmune conditions were excluded from the trials. In this study, we used the Surveillance, Epidemiology, and End Results (SEER)–Medicare linked database to determine the characteristics and prognosis of older adults with NSCLC who were treated with immune checkpoint inhibitors in routine oncology practice.

MATERIALS AND METHODS

Study Design and Data

SEER-Medicare–linked data were used to identify patients who were diagnosed with pathologically confirmed stage I to stage IV NSCLC between 2002 and 2015 and were treated with nivolumab or pembrolizumab between 2014 and 2016. We identified receipt of nivolumab and pembrolizumab using the Healthcare Common Procedure Coding System (HCPCS) and National Drug Codes in Medicare claims. To ensure the completeness of information regarding treatment initiation, we included patients who initiated these therapies ≥2 months after specific HCPCS J codes became available in January 2016 (ie, J9299 for nivolumab and J9271 for pembrolizumab). Patients with a first immune checkpoint inhibitor claim prior to March 2016 were excluded from survival analyses because they were considered to have potentially incomplete information concerning the initiation date (eg, may have had prior claims with nonspecific HCPCS J codes or temporary billing codes). We only included patients who were continuously eligible for Medicare Parts A and B and were not enrolled in a Medicare Advantage plan from the time of diagnosis to immune checkpoint inhibitor initiation (or at least 12 months prior to initiation, whichever was longer) to identify baseline characteristics and treatment history. We excluded patients aged <65 years and those with a history of malignancies other than lung cancer for which immune checkpoint inhibitors also are indicated. We followed patients from the time of immune checkpoint inhibitor initiation through December 31, 2017.

Outcomes and Covariates

The outcome was overall survival; covariates included patient and tumor characteristics and treatment history. We obtained information regarding patient age, sex, and race; the original reason for Medicare entitlement; Medicaid dual eligibility; rural residence; region; diagnosis year and month; stage of disease at the time of diagnosis; and histology. We also identified the number of Charlson comorbid conditions^5–7; a proxy of performance status^8,9; the presence of 31 autoimmune conditions during the year before initiation of immune checkpoint inhibitors¹⁰; and central nervous system (CNS) metastasis within 6 months prior to initiation from Medicare claims.¹¹ The proxy of performance status was based on claims for health care services commonly used by patients with poor performance status, such as use of home oxygen, wheelchair, or home health services.⁸

We extracted information regarding treatment history, including prior systemic therapy and oral targeted therapy, recent radiotherapy, care at National Cancer Institute (NCI)–designated cancer centers, and time from initial diagnosis to immune checkpoint inhibitor initiation. Prior systemic therapy included systemic chemotherapy or infusion targeted therapy since diagnosis, and was classified as none, single-agent chemotherapy, platinum doublets with or without bevacizumab, and other. We assessed prior oral targeted therapy with epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors; and anaplastic lymphoma kinase (ALK), ROS1, BRAF, and MEK inhibitors. Patients who did not have continuous Medicare Part D enrollment from the time of diagnosis and had no oral targeted therapy claims were classified as unknown. The codes and definitions of the study variables are provided in Supporting Table 1.

Statistical Analysis

We summarized patient and tumor characteristics and treatment history. We estimated survival curves using the Kaplan-Meier method and identified prognostic factors for death using Cox proportional hazards regression. In exploratory descriptive analyses, we examined subgroups by histology and stage of disease at the time of the initial diagnosis. We also examined the percentage and median survival of patients with baseline steroid or other immunosuppressant use by prior autoimmune disease status. To assess the robustness of our findings, we applied different criteria for cohort selection and variable definitions. First, we excluded patients with a history of any other cancer prior to the initiation of immune checkpoint inhibitors and examined whether the findings were similar to those of the main analyses. Second, we included patients with a first immune checkpoint inhibitor claim prior to March 2016. Third, we used a 1-year covariate assessment period to identify CNS metastasis. We performed analyses using SAS statistical software (version 9.4; SAS Institute Inc, Cary, North Carolina) and R statistical software (version 3.5.2; R Foundation for Statistical Computing, Vienna, Austria). The Brown University institutional review board waived approval requirement for the study.

RESULTS

Among 4324 patients who received immune checkpoint inhibitors, a total of 1256 met the study eligibility criteria (Fig. 1, Supporting Table 2 and Table 1 summarize the characteristics of all treated patients and the final cohort of those who initiated immune checkpoint inhibitors, respectively). Of the 1256 patients, 1156 patients (92.0%) initiated nivolumab and 100 patients (8.0%) initiated pembrolizumab. The median age at the time of immune checkpoint inhibitor initiation was 75.3 years (interquartile range, 8.5) and 8.4% of patients were aged ≥85 years. Approximately one-half of the patients (48.7%) had ≥2 comorbidities, 11.5% had a poor performance status based on the claims-based proxy variable, and 12.6% had a history of autoimmune conditions. Approximately one-half of the patients (49.7%) initiated immune checkpoint inhibitors after platinum doublets and 8.1% had received no systemic therapy prior to immune checkpoint inhibitor initiation. Approximately 20% of patients received care at an NCI-designated cancer center. The median times from diagnosis to immune checkpoint inhibitor initiation were 14.1 months and 19.3 months, respectively, for those initially diagnosed with stage IV and stage I to stage III disease.

Figure 1.

Cohort selection diagram. ^aFirst immune checkpoint inhibitor claim prior to March 1, 2016. Specific Healthcare Common Procedure Coding System J codes used to identify immune checkpoint inhibitors from Medicare claims (ie, J9299 for nivolumab and J9271 for pembrolizumab) became available in January 2016. A 2-month washout period was used to ascertain treatment initiation. ^bLung cancer was not the first primary malignant cancer, but information regarding prior disease was unavailable.

TABLE 1.

Baseline Characteristics and Their Association With Mortality

Variables	Cohort (N = 1256) No. (%)	Univariablea			Multivariableb
		HR	95% CI	P	HR	95% CI	P
Patient Characteristics
Age at initiation, y
65–74	605 (48.17)	Reference group		.52	Reference group		.65
75–84	545 (43.39)	0.98	(0.85–1.13)		1.00	(0.86–1.16)
≥85	106 (8.44)	0.86	(0.67–1.11)		0.88	(0.67–1.16)
Sex
Male	653 (51.99)	Reference group		.08	Reference group		.76
Female	603 (48.01)	0.89	(0.77–1.01)		0.98	(0.85–1.13)
Race
White	1072 (85.35)	Reference group		.32	Reference group		.38
Black	89 (7.09)	1.24	(0.96–1.60)		1.16	(0.89–1.51)
Asian	37 (2.95)	1.16	(0.79–1.70)		1.35	(0.89–2.05)
Other	58 (4.62)	0.92	(0.66–1.30)		0.99	(0.70–1.41)
Original reason for Medicare entitlement
Age	1094 (87.10)	Reference group		.54	Reference group		.90
Disability/ESRD	162 (12.90)	1.06	(0.87–1.30)		1.01	(0.82–1.26)
Medicaid dual eligibilityc
No	1071 (85.27)	Reference group		.71	Reference group		.15
Yes	185 (14.73)	0.96	(0.80–1.17)		0.85	(0.69–1.06)
Rural residence
No	1228 (97.77)	Reference group		.41	Reference group		.27
Yes	28 (2.23)	0.82	(0.51–1.32)		0.76	(0.46–1.24)
Region
Northeast	307 (24.44)	Reference group		.07	Reference group		.15
Midwest	143 (11.39)	0.97	(0.76–1.23)		0.95	(0.74–1.22)
South	308 (24.52)	1.19	(0.99–1.45)		1.19	(0.97–1.45)
West	498 (39.65)	0.96	(0.80–1.14)		0.98	(0.81–1.17)
No. of comorbid conditionsd
0	253 (20.14)	Reference group		<.001	Reference group		.002
1	392 (31.21)	1.21	(0.99–1.48)		1.17	(0.95–1.43)
≥2	611 (48.65)	1.43	(1.19–1.73)		1.40	(1.15–1.70)
Proxy of performance statusd
ECOG 0–1	1112 (88.54)	Reference group		.42	Reference group		.40
ECOG 2–4	144 (11.46)	1.09	(0.88–1.35)		1.10	(0.88–1.38)
Prior autoimmune conditionsd
No	1098 (87.42)	Reference group		.24	Reference group		.51
Yes	158 (12.58)	1.13	(0.92–1.37)		1.07	(0.88–1.31)
Tumor Characteristics
Stage of disease at diagnosis
I	229 (18.23)	Reference group		.53	Reference group		.42
II	75 (5.97)	1.21	(0.88–1.66)		1.12	(0.81–1.55)
III	417 (33.20)	1.11	(0.91–1.35)		1.03	(0.83–1.28)
IV	535 (42.60)	1.03	(0.85–1.25)		0.92	(0.74–1.14)
Histology
Nonsquamous	793 (63.14)	Reference group		.01	Reference group		.04
Squamous	372 (29.62)	1.24	(1.07–1.44)		1.24	(1.05–1.45)
NSCLC NOS	91 (7.25)	1.20	(0.93–1.55)		1.09	(0.84–1.42)
Central nervous system metastasise
No	1056 (84.08)	Reference group		.08	Reference group		.25
Yes	200 (15.92)	1.17	(0.98–1.40)		1.12	(0.92–1.36)
Treatment History
Prior systemic therapyf
Platinum doublets	624 (49.68)	Reference group		.47	Reference group		.01
None	102 (8.12)	0.92	(0.71–1.20)		1.05	(0.79–1.38)
Single-agent chemotherapy	46 (3.66)	1.32	(0.93–1.86)		1.53	(1.07–2.20)
Platinum doublets plus bevacizumab	155 (12.34)	1.05	(0.85–1.30)		1.26	(1.00–1.59)
Other	329 (26.19)	1.07	(0.91–1.26)		1.31	(1.10–1.56)
Prior oral targeted therapyg
No	706 (56.21)	Reference group		.47	Reference group		.68
Yes	114 (9.08)	0.87	(0.67–1.11)		1.01	(0.77–1.31)
Unknown	436 (34.71)	1.02	(0.88–1.18)		1.07	(0.92–1.25)
Recent radiotherapyh
No	1124 (89.49)	Reference group		.001	Reference group		.001
Yes	132 (10.51)	1.44	(1.17–1.77)		1.47	(1.18–1.83)
Care at NCI-designated cancer centeri
No	1017 (80.97)	Reference group		.83	Reference group		.97
Yes	239 (19.03)	1.02	(0.86–1.21)		1.00	(0.84–1.20)
Time from diagnosis to initiation
2–6 mo	46 (3.66)	Reference group		.001	Reference group		<.001
6 mo to 1 y	319 (25.40)	0.83	(0.58–1.17)		0.83	(0.58–1.18)
1–3 y	658 (52.39)	0.68	(0.48–0.95)		0.64	(0.45–0.91)
>3 y	233 (18.55)	0.57	(0.40–0.82)		0.51	(0.35–0.76)

Abbreviations: ECOG, Eastern Cooperative Oncology Group; ESRD, end-stage renal disease; HR, hazard ratio; NCI, National Cancer Institute; NOS, not otherwise specified; NSCLC, non-small cell lung cancer.

^aResults were derived from univariable Cox proportional hazards models. P values were calculated from Wald tests.

^bResults were derived from a multivariable Cox proportional hazards model with all the variables listed in the table. P values were calculated from Wald tests.

^cAny state buy-in within the year prior to initiation.

^dWithin the year prior to initiation.

^eWithin the 6 months prior to initiation.

^fSince diagnosis. Included chemotherapy and infusion targeted therapy and excluded oral targeted therapy.

^gSince diagnosis. Included EGFR tyrosine kinase inhibitors; ALK, ROS1, BRAF, and MEK inhibitors. Unknown status for those without continuous Medicare Part D coverage.

^hWithin the 30 days prior to initiation.

ⁱDefined as having any claim originating at an NCI-designated cancer center within the 90 days prior to initiation.

The median overall survival from the time of immune checkpoint inhibitor initiation was 9.3 months (95% CI, 8.5–10.5 months), the 1-year survival probability was 43.0% (95% CI, 40.2–45.7%), and the 18-month survival probability was 31.3% (95% CI, 28.5–34.1%). Survival curves stratified by the number of comorbid conditions, histology, and treatment history are shown in Figure 2.

Figure 2.

Kaplan-Meier survival curves by the number of comorbid conditions, histology, and treatment history. P values were derived from log-rank tests. Prior systemic therapy included chemotherapy and infusion targeted therapy (excluding oral targeted therapy). Recent radiotherapy included radiation within the 30 days prior to the initiation of immune checkpoint inhibitors. Cell sizes of <11 patients were suppressed as per National Cancer Institute rules to preserve patient confidentiality. NOS indicates not otherwise specified; NSCLC, non–small cell lung cancer.

Multiple comorbid conditions, squamous histology, a history of nonplatinum doublet systemic therapy, recent radiotherapy, and a shorter time from the initial diagnosis to treatment initiation were found to be statistically significantly associated with an increased hazard of death in multivariable survival analyses (Table 1). For example, patients who received multiple prior systemic therapies or nonstandard regimens had worse survival outcomes compared with those who received only platinum doublets prior to immune checkpoint inhibitors. Patient demographics, performance status on the claims-based proxy, prior autoimmune conditions, stage of disease at the time of the initial diagnosis, and prior oral targeted therapy were not found to be significantly associated with an increased hazard of death.

The results of exploratory analyses stratified by histology and stage of disease at the time of the initial diagnosis are described in Supporting Tables 3 to 6. The percentages of patients with baseline injection steroid use were 46.8% and 32.0%, respectively, for those with and without a history of autoimmune disease. The percentages of patients who had received any steroid (injection or oral) or other immunosuppressants are described in Supporting Table 7. The results of median survival stratified by prior autoimmune disease and baseline injection steroid use are described in Supporting Table 8.

The findings of our study were robust to changes in the criteria for cohort selection and variable definitions. First, the results were similar to those of the main analyses when we excluded patients with a history of any malignancy other than lung cancer prior to immune checkpoint inhibitor initiation (see Supporting Fig. 1, Supporting Table 9). Second, the results also were similar when we included patients with a first immune checkpoint inhibitor claim prior to March 2016 (see Supporting Fig. 2, Supporting Table 10). The median overall survival after the first claim date was 8.7 months (95% CI, 8.4–9.6 months). Third, using a 1-year covariate assessment period to identify CNS metastasis did not change the main findings (see Supporting Table 11).

DISCUSSION

In this study, we used real-world data to examine the characteristics and prognosis of a large cohort of older adults with NSCLC who initiated immune checkpoint inhibitor therapy. Median survival after initiation of immune checkpoint inhibitors was 9.3 months. Older adults in our study were likely to have underlying health problems other than NSCLC, including multiple comorbid conditions, prior autoimmune disease, or a proxy of poor performance status. The percentage of patients receiving care at NCI-designated cancer centers was higher than the previously reported rate of NCI-designated cancer center attendance in patients with advanced lung cancer,¹² and it is likely to reflect the earlier adoption of treatment with immune checkpoint inhibitors in these centers.

We found that factors that are generally related to poor prognosis among patients with advanced NSCLC in routine practice, such as comorbid conditions, squamous histology, failure of aggressive prior treatment or nonstandard regimens, or the need for (palliative) radiation, remain predictors of poor prognosis in older adults who are treated with immune checkpoint inhibitors.^13,14 A shorter time from the initial diagnosis to immune checkpoint inhibitor initiation was also associated with poor prognosis and possibly indicates patients with rapid disease progression, given that nivolumab and pembrolizumab were used mainly as later lines of treatment during the study period. Similar to prior findings from community-based cancer clinics,¹⁵ we found no differences in prognosis between different age groups or patients who received these therapies as first-line or second-line treatment after platinum doublets. We also found no difference based on prior oral targeted therapy use.

In our multivariable analyses, autoimmune conditions were not found to be associated with the hazard of death. Although it is likely that patients with severe autoimmune disease were not included in our cohort, the results suggest that autoimmune conditions may not be an absolute contraindication for immune checkpoint inhibitor therapy.¹⁶ Patients with a history of autoimmune conditions were more likely to have received injection steroids at baseline compared with patients without such a history, although our descriptive analyses could not examine the reasons for the use of these medications. Further studies are warranted to determine which patients with autoimmune conditions can be treated safely with immune checkpoint inhibitors.

Our study has several limitations. First, SEER-Medicare data do not provide information regarding genetic mutation status, including PD-L1 expression level and smoking status. The majority of patients in our cohort received nivolumab, which was approved regardless of the PD-L1 expression level, and our analysis reflects the practice patterns during the study period. Second, our results regarding oral targeted therapy may have been influenced by the lack of information for patients without Medicare Part D coverage. Third, due to data availability, the survival analyses in our study were restricted to patients who were diagnosed until December 31, 2015, and initiated immune checkpoint inhibitors between March 1, 2016 and December 31, 2016. Patients included in survival analyses were likely to have a longer time from their initial diagnosis to the initiation of immune checkpoint inhibitor treatment compared with those who were not. Fourth, our claim-based algorithms for identifying selected baseline characteristics, including CNS metastasis and prior autoimmune conditions, may not be completely accurate. We used algorithms based on prior literature for these conditions^10,11; however, the validity of these algorithms has not been fully assessed. We used previously validated algorithms for comorbid conditions and a proxy for performance status to reduce missclassification.^6,9 Fifth, SEER-Medicare data are not geographically representative of the US population; patients in the Midwest and Northeast are underrepresented and those in the West and South are overrepresented.¹⁷ The Midwest and South regions have higher rates of smoking, lung cancer incidence, and mortality compared with the West and Northeast regions.^18–20 Although the impact of these sample characteristics on our findings is difficult to fully evaluate, we found no difference in survival outcomes with respect to geographic region. Sixth, we could not directly determine the line of therapy from SEER-Medicare data. However, the prior systemic therapy variable in our analysis can be considered a surrogate for the line of therapy because it identifies regimens received prior to immune checkpoint inhibitor initiation. Seventh, our results may not apply to patients aged <65 years. Nevertheless, our findings were generally similar to those of a prior study that included younger patients who were treated with immune checkpoint inhibitors in routine practice.¹⁵ Similar to our results, that study did not find an association between age and survival among patients who were treated with immune checkpoint inhibitors.

In conclusion, we identified predictors of mortality in older adults with NSCLC who were receiving immune checkpoint inhibitors in routine practice. Many older adults with NSCLC who initiated immune checkpoint inhibitors in routine oncology practice had multiple comorbidities or a history of autoimmune disease. Factors associated with poor prognosis among patients with advanced NSCLC were also found to be predictive of decreased survival in older adults who were treated with immune checkpoint inhibitors.