Correlation between patients’ age and cancer immunotherapy efficacy

ABSTRACT

Background: Although immunosenescence-induced difference on overall immune function and immune cell subsets between younger and older populations has been well characterized, the potential effect of patients’ age on the efficacy of immune checkpoint inhibitors (ICIs) remains little known. We performed a meta-analysis to investigate whether age differences play a role in cancer immunotherapy efficacy based on a large amount of clinical data.

Methods: We conducted a systematic search of PubMed, Embase and MEDLINE for relevant randomized controlled trials. The primary outcome was overall survival (OS) and progression-free survival (PFS) was secondary outcome. The interaction test was used to assess the heterogeneity of HR between younger and older groups.

Results: In total, 19 clinical randomized trials involving 11157 patients were included. The pooled HR for OS was 0.73 (95% CI 0.69–0.78) and 0.63 (95% CI 0.52–0.73) for PFS in younger patients receiving ICIs treatments, when compared with younger patients treated with controls. For older patients treated with ICIs, the pooled HR for OS compared with controls was 0.64 (95% CI 0.59–0.69) and 0.66 (95% CI 0.58–0.74) for PFS. The difference on OS efficacy between younger and older patients treated with ICIs was significant (P_{heterogeneity} = 0.025).

Conclusions: Immune checkpoint inhibitors significantly improved OS and PFS in both younger and older patients compared with controls, but the magnitude of benefit was clinically age-dependent. Patients ≥65 y can benefit more from immunotherapy than younger patients. Future research should take age difference into consideration in trials and focus on tolerance and toxicity of ICIs in older patients.

Introduction

Cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4) and programmed cell death protein-1 (PD-1) receptor pathways play an important part in tumor-induced immune tolerance, which are referred to as “immune checkpoints”. Cancer cells exploit these immunosuppressive pathways to evade immune attack.^1–6 Multiple monoclonal antibodies have been designed for targeting these checkpoints to enhance the function of the immune system. The advent of PD-1/L1 inhibitors and CTLA-4 inhibitors have been undoubtedly an inspiring breakthrough in cancer immunotherapy in recent years. Many clinical trials have evaluated the efficacy of immune checkpoint inhibitors (ICIs) and demonstrated a significant overall survival (OS) benefit and improved anti-tumor immune responses in several types of solid and hematologic malignancies.^7-12 CTLA-4 inhibitors, such as ipilimumab and tremelimumab, can break immune tolerance and activate the initial stages of T-cells, thus enhancing anti-tumor immune response. PD-1 inhibitors and PD-L1 inhibitors reactivate previously primed T lymphocytes that have lost effector and proliferative function, and disrupt the negative regulation of immune responses. The most typical PD-1 inhibitors are nivolumab and pembrolizumab while the PD-L1 inhibitor is atezolizumab.^13-15 Malignancies are a disease belonging to the older adults and the incidence and mortality are associated with the age of patients.^16,17 Notably, a previous meta-analysis showed that patients aged ≥75 y might not benefit from anti-PD1 monoclonal antibody (overall survival benefit). However, an original research found that the level of forkhead box protein P3 positive regulatory T cells (Tregs) in the melanomas of young mice receiving ICIs treatments was significantly higher compared to the aged mice while the CD8þ effector T-cell numbers was lower, which could result in a significant decrease in anti-tumor immune response of young mice. And a similar result was observed in younger patients treated with anti-PD1. Considering CD8þ effector T cells suppressed by Tregs and the changes of CD8þ cells types with age, it seems to partly explain that younger patients are more likely to resist anti-PD1 inhibition and older patients may derive a better benefit from ICIs than young patients.^18,19 The root cause of the conflict between the above findings is that age-dependent changes in intratumoral immune populations and response to immunotherapy in the tumor microenvironment remains little known.²⁰ Given that the correlation between patients’ age and cancer immunotherapy efficacy still remains hugely controversial, in this study, we conducted a systemic review and meta-analysis based on a large amount of clinical data to investigate whether age differences play a role in cancer immunotherapy efficacy.

Results

Search results and patient characteristics

The above database search and the examination of reference lists yielded a total of 13546 publications, from which we identified 153 potentially relevant studies. We excluded a study that did not report the efficacy of patients aged over 75.²¹ According to our selection criteria, we included a total of 19 studies for final analysis after abstract and full article review. This included 14 phase 3 trials, 4 phase 2 trials and 1 phase 2/3 trial. Among these studies, six investigated nivolumab, five investigated pembrolizumab, three investigated atezolizumab, three investigated ipilimumab, one investigated tremelimumab and one investigated ipilimumab plus nivolumab. The tumor types were melanoma in nine trials, non-small cell lung cancer in seven trials, and renal cell carcinoma, urothelial carcinoma and gastric tumors in one trial. A total of 11,157 patients were eligible for this meta-analysis, of whom 4930 (44.19%) had melanoma, 4371 (39.18%) had non-small-cell lung cancer, 821 (7.36%) had renal cell carcinoma, 542 (4.86%) had urothelial carcinoma and 493 (4.42%) had gastric tumors. The sample size in each study ranged between 138 and 2075 and the age of enrolled patients ranged from 15 to 90 y. A total of 2991 patients were enrolled in nivolumab trials, 3240 in pembrolizumab trials, 3212 in atezolizumab trials, 921 in ipilimumab trials, 655 in tremelimumab trials and 138 in nivolumab plus ipilimumab trials. The details of the study characteristics are summarized in Table 1.

Table 1.

Main characteristics of the studies included in the meta-analysis.

Study	Year	Phase	Tumour type	Treatment arms		Number of patients			Overall survival (HR, 95% CI)
						<65		≥65	Patients <65 y	Patients ≥65 y
Borghaei22	2015	3	Non-small-cell lung cancer	Nivolumab vs. docetaxel		339		243	0.81(0.62–1.04)	0.65(0.44–0.86)
Brahmer23	2015	3	Non–small-cell lung cancer	Nivolumab vs. docetaxel		152		120	0.52(0.35–0.75)	0.59(0.31–0.87)
Fehrenbacher24	2016	2	Non-small-cell lung cancer	Atezolizumab vs. docetaxel		175		112	0.70(0.48–1.01)	0.65(0.42–0.99)
Fehrenbacher25	2018	3	Non-small cell lung cancer	Atezolizumab vs. docetaxel	A: ITT850	453		397	A: 0.81(0.65–1.01)	0.69(0.54–0.87)
					B:ITT1225	661		564	B: 0.84(0.70–1.01)	0.75(0.61–0.91)
Herbst9	2016	2/3	Non-small-cell lung cancer	Pembrolizumab vs. docetaxel		604		429	0.63(0.50–0.79)	0.76(0.57–1.02)
Rittmeyer26	2016	3	Non-small-cell lung cancer	Atezolizumab vs. docetaxel		453		397	0.8(0.64–1.00)	0.6(0.52–0.83)
Ribas27	2013	3	Melanoma	Tremelimumab vs. chemotherapy		455		200	0.88(0.72–1.08)	0.87(0.64–1.18)
Robert28	2015	3	Melanoma	Nivolumab vs. dacarbazine		200		218	0.52(0.32–0.85)	0.33(0.14–0.52)
Larkin29	2018	3	Melanoma	Nivolumab vs. investigator’s choice chemotherapy		257		148	1.17(0.84–1.63)	0.62(0.41–0.94)
Hodi30	2014	2	Melanoma	Ipilimumab plus sargramostim vs. ipilimumab		138		107	0.66(0.41–1.07)	0.59(0.30–1.16)
Weber31	2016	2	Melanoma	Nivolumab followed by ipilimumab vs. ipilimumab followed by nivolumab		82		56	0.54(0.29–1.01)	0.40(0.16–0.97)
Hodi32	2010	3	Melanoma	A: Ipilimumab plus gp100 vs. gp100;		385		154	0.70(0.54–0.90)	0.69(0.47–1.01)
				B: Ipilimumab vs. gp100		189		84	0.65(0.47–0.90)	0.61(0.38–0.99)
Robert33	2015	3	Melanoma	A: Pembrolizumab every 2 weeks vs. ipilimumab		NA		NA	0.65(0.44–0.95)	0.56(0.36–0.87)
				B: Pembrolizumab every 3 weeks vs. ipilimumab		NA		NA	0.77(0.53–1.12)	0.66(0.44–1.01)
Kang34	2017	3	Gastric or gastroesophageal junction cancer	Nivolumab vs. Placebo		284		209	0.76(0.58–1.00)	0.53(0.38–0.74)
Motzer35	2015	3	Renal cell carcinoma	Nivolumab vs. everolimus		497		324	0.78(0.60–1.01)	0.68(0.46–0.90)
Bellmunt36	2017	3	Urothelial carcinoma	Pembrolizumab vs. chemotherapy		230		312	0.75(0.53–1.05)	0.76(0.56–1.02)
						Number of patients			Progression-free Survival (HR, 95% CI)
						<65	≥65		Patients < 65 y	Patients ≥ 65 y
Reck37	2016	3	Non–small-cell lung cancer	Pembrolizumab vs. Chemotherapy		141	164		0.61(0.40–0.92)	0.45(0.29–0.70)
Ribas38	2015	2	Melanoma	A: Pembrolizumab 2 mg/kg vs. chemotherapy		NA	NA		0.47(0.34–0.66)	0.70(0.48–1.01)
				B: Pembrolizumab 10 mg/kg vs. chemotherapy		NA	NA		0.42(0.30–0.59)	0.60(0.41–0.88)
Eggermont39	2018	3	Melanoma	Pembrolizumab vs. pacebo		768	251		0.57(0.41–0.80)	0.55(0.32–0.93)

Primary outcome: overall survival

A total of 16 trials involving 9795 patients were included in the meta-analysis of OS. Patients were divided into younger and older populations with an age cut-off of 65 y. For younger patients, the pooled HR of ICIs compared to control therapy was 0.73 (95% CI 0.69–0.78, Figure 1). The fixed-effects model was used because there was no significant heterogeneity between the individual studies in this analysis (p = 0.214, I² = 19.7%). For older patients treated with ICIs, the pooled HR for OS was 0.64 (95% CI 0.59 −0.69, Figure 1) compared to controls. The fixed-effects model was also used because there was no evidence of significant heterogeneity among the individual studies (P = 0.264; I² = 15.5%). ICIs significantly prolonged the OS in both younger and older populations in comparison with control therapies. Notably, the OS benefit obtained from immune checkpoint inhibitors in younger patients compared to older patients was smaller (HR: 0.73 vs. 0.64). There was a statistically significant difference in the OS efficacy of ICIs between younger and older patients regarding the pooled HR (P_{heterogeneity} = 0.025, Table 2). This result manifested that in older patients, ICIs demonstrated a more significant overall survival benefit compared to those in younger patients. We also performed a subgroup analysis according to cancer types and ICIs types. The results of subgroup analyses are shown in Table 2. For ICIs subgroups, both PD-1/L1 inhibitors and CTLA-4 inhibitors significantly prolonged the OS in younger and older patients compared to controls. Older patients treated with PD-1/L1 inhibitors experienced a longer OS than younger patients (HR: 0.64 vs. 0.74). The heterogeneity test for this age-related interaction, assessed between younger and older patients was statistically significant (P_{heterogeneity} = 0.025). Notably, age difference had no effect on OS in patients treated with CTLA-4 inhibitors. CTLA-4 inhibitors had comparable efficacy in younger vs. older patients (HR: 0.74 vs. 0.71). The OS difference between younger and older patients in PD-1/L1 inhibitor group was more obvious compared to CTLA-4 inhibitor group. For each of cancer histotype (melanoma, non-small-cell lung cancer and other tumor types), the magnitude of efficacy of ICIs was greater for older patients than for younger patients. The difference of OS between younger and older patients, evaluated within each subgroup, was not statistically significant. Notably, age-related difference had more effect on OS in melanoma patients compared to non-small-cell lung cancer patients (Table 2). We also conducted a separate analysis of PD-1 inhibitors in melanoma patients. The pooled OS HR for younger patients was 0.74 (95% CI 0.51–0.98) and 0.50 (95% CI 0.38–0.62) for older patients, when compared with patients treated with controls. The difference was not significant by interaction test (P_{heterogeneity} = 0.092, Table 2).

 Figure 1.

Forest plot of hazard ratios for overall survival by patients’ age

Table 2.

Pooled hazard ratios for OS and PFS according to ICI types and tumor types.

Subgroup	Number of trials	Age	HR (95% CI)	I2	Pheterogeneitya
Type of ICIs
PD-1/L1 inhibitors	12	<65	0.74(0.68, 0.79)	27.2%	0.025
		≥65	0.64(0.58, 0.69)	23.4%
CTLA-4 inhibitors	3	<65	0.74(0.64, 0.85)	11.9%	0.634
		≥65	0.71(0.56, 0.87)	0.0%
Type of tumors
Melanoma	7	<65	0.72(0.64, 0.80)	34.1%	0.089
		≥65	0.57(0.48, 0.66)	38.0%
Non-small-cell lung cancer	6	<65	0.73(0.66, 0.80)	39.6%	0.271
		≥65	0.69(0.62, 0.76)	0.0%
Others	3	<65	0.77(0.64, 0.89)	0.0%	0.059
		≥65	0.64(0.52, 0.75)	23.1%
PD-1 inhibitors in melanoma patients	3	<65	0.74(0.51, 0.98)	60.3%	0.092
		≥65	0.50(0.38, 0.62)	43.2%
Overall OS	16	<65	0.73(0.69, 0.78)	19.7%	0.025
		≥65	0.64(0.59, 0.69)	15.5%
Overall PFS	8	<65	0.63(0.52, 0.73)	66.3%	0.55
		≥65	0.66(0.58, 0.74)	44.3%

^a P value for difference on HR between younger and older patients

Secondary outcomes: progression-free survival

The analysis of PFS was based on eight trials comprising 4788 patients. The pooled HR for PFS in younger patients showed significant difference between ICIs treatments and control therapies (HR 0.63, 95% CI 0.52–0.73, Figure 2). The random-effects model was used due to the high heterogeneity between the individual studies (p = 0.002, I² = 66.3%). For older patients, ICIs significantly prolonged the PFS in comparison with controls (HR 0.66, 95% CI 0.58–0.74, Figure 2). The fixed-effects model was used because there was no evidence of significant heterogeneity among the included studies (p = 0.064, I² = 44.3%). The above results showed that both younger and older patients treated with immune checkpoint inhibitors had a longer PFS compared to controls. Although the pooled HR in younger patients was lower than that in older patients, the heterogeneity test for this difference was not significant (P_{heterogeneity} = 0.44, Table 2), which manifested that ICIs had comparable PFS efficacy in younger and older patients. We did not further perform a subgroup analysis according to the type of cancer and the type of ICIs because of the small sample size.

Figure 2.

Forest plot of hazard ratios for progression-free survival by patients’ age.

Publication bias and sensitivity analysis

The Begg’s test and Egger’s test showed there was no significant publication bias for primary outcomes and secondary outcomes in our meta-analysis. Considering the low heterogeneity among the included studies, we only conducted a sensitivity analysis on the pooled HR for PFS in young patients with high heterogeneity. Sensitivity analysis showed that the pooled HR of PFS in younger patients had high stability and removing any single study did not significantly change the combined results (data not shown).

Discussion

The previous studies showed no significant difference on overall survival benefits of ICIs between younger and older patients.⁴⁰ Recently, an original research by Kugel et al. indicated that in melanoma patients treated with PD-1 inhibitors, older patients (≥60 y) responded more efficiently to PD-1 inhibitors compared to younger patients. Their findings were quite surprising and transformed our understanding of cancer immunotherapy efficacy in older patients as it was well-known that the elderly experienced a process of age-related immunosenescence. Given that the clinical difference on immunotherapy efficacy between younger and older patients still remains little known, we performed this meta-analysis to systematically explore the correlation between patients’ age and cancer immunotherapy efficacy based on 11157 patients.

Our meta-analysis showed that by using 65 y as the cut-off age, ICIs can significantly improve overall survival in both younger patients and older patients in comparison with controls. The heterogeneity test suggested that there was significant difference on survival benefits between younger and older patients. Older patients treated with ICIs had a larger treatment effect than younger patients.

To the best of our knowledge, this is the first study to clearly show significant difference on the efficacy of ICIs between younger and older patients. Meanwhile, our study clinically demonstrated for the first time that the efficacy of ICIs in older patients was better than that in younger patients. A previous meta-analysis performed by Nishijima et al. showed the difference upon OS benefit for ICIs in both younger and older patients was not statistically significant. However, the OS analysis was based on only eight trials involving 4725 patients and the authors used ambiguous ages as the age cutoff to dichotomize patients into two groups. It was worth noting that in this meta-analysis, a significant improvement on PFS in older patients receiving ICIs treatments was not observed compared with controls. A recent meta-analysis by Elias evaluated the efficacy of PD1/L1 inhibitors in older adults based on nine studies and showed a comparable efficacy benefit in terms of OS [HR 0.68 (CI 0.61–0.75) vs. 0.64 (CI 0.54–0.76)] and PFS [HR 0.73 (CI 0.61–0.88) vs. 0.74 (CI 0.60–0.92)] between younger and older patients. The above two meta-analyses have potential limitation in the inclusion and exclusion criteria and there was significant heterogeneity between the individual studies. In this study, we have developed strict inclusion criteria to reduce the impact of potential confounding factors on the final outcomes. The patients must receive ICIs treatments or ICIs combined with other immunological compounds. Our study demonstrated with the best level of evidence that ICIs had a better therapeutic effect on elderly patients compared with younger patients. The overall survival in patients >65 y was significantly longer than younger patients. We attempted to perform a subgroup analysis to explore whether this particular relationship was related to tumor types and ICI types. We found this age-related difference on overall survival was more prominent in PD1/L1 inhibitors than in CTLA-4 inhibitors. Also, the effect of age difference on the overall survival of melanoma patients is greater than that of non-small cell lung cancer patients. Notably, we did not observe a significant improvement in the overall survival of patients aged more than 75 y receiving ICIs treatment. A total of four trials included in this meta-analysis reported the overall survival HR of patients older than 75 years, of which only one showed ICIs significantly improved overall survival in patients ≥75 years. However, any conclusions drawn from patients aged more than 75 years were not reliable enough because only 213 patients ≥75 years were included in the analysis, accounting for a quite small part of all patients ≥65 y.

Aging can disorder the proportion between subpopulations of immune cells and affect the overall function of immune system in the body, especially T cell-mediated immune responses, which in part results in the occurrence of immune-related diseases in older individuals. Decreased number of Tregs significantly reduced the suppressive of CD8 þ effector T cells and boosted antitumor immunity.^41,42 Although an increased number of Tregs has been found in the skin of older adults, age-dependent changes in the intratumoral number and function of Tregs are poorly understood.¹⁹ A recent study by Kugel et al. indicated that intratumoral CD8þ: Treg ratios of aged melanoma patients treated with PD-1 inhibitors significantly increased compared to younger patients and a similar result was observed in aged mice. This increase may result in the difference on response to immunotherapy between younger and older patients.¹⁹ However, when Kugel et al. considered complete response rate rather than any other clinical responses, they found no difference between younger and older patients (13% vs. 15%). Therefore, there could be other potential molecular mechanisms in the tumor microenvironment such as tumor-associated stroma involved in this age-induced difference on cancer immunotherapy efficacy. Additionally, the “quality“ of aging rather than age itself may also tune the cancer-immune set point. But only these clinical benefits translating into improved overall survival could be the biggest help for patients.^43-45

Our meta-analysis has several significant clinical and research values. We have demonstrated that older patients benefited more from ICIs compared with younger patients. Such results encourage older cancer patients to actively receive ICIs treatments and participate in the corresponding clinical trials. Future research should take age factor into consideration when exploring new immunotherapeutic approaches. Although the mechanism of age-mediated difference on the efficacy of immunotherapy in melanoma patients has been identified, whether such this mechanism exists in other malignancies remains unknown. Meanwhile, the process of intratumoral CD8þ: Treg ratio increasing in older patients still requires to be explored.

Strengths and limitations

Our meta-analysis has the following strengths. All analyses were based on the most comprehensive and the latest clinical trial data, which allowed us to perform a subgroup analysis to explore whether tumor types and ICI types were related to the difference on immunotherapy efficacy between younger and older patients. In addition, the combined results were fairly stable in this meta-analysis because there was no significant heterogeneity between the individual studies. Several potential limitations in our study should be acknowledged. We did not further perform a subgroup analysis on PFS because of the small sample size, and whether age difference had effect on PFS of immunotherapy still required a lot of research to confirm. Although we have demonstrated that older patients can benefit better from ICIs, we could not establish a comparison of tolerance and toxicity of ICIs between younger and older patients. Despite the above limitations, as far as we know, this is by far the largest and most comprehensive study that incorporates results from 19 clinical trials involving 11,157 patients.

Conclusions

By using a cut-off age of 65 y to dichotomize patients into younger and older groups, our meta-analysis showed that immune checkpoint inhibitors can significantly improve overall survival and progression-free survival in both younger and older patients compared with controls, but the magnitude of benefit was age-dependent. Based on a large amount of clinical data, our study clinically demonstrated for the first time that older patients benefited more from immunotherapy than younger patients. Future research in cancer immunotherapy should take age difference into account, guarantee inclusion of more older patients and focus on tolerance and toxicity of ICIs in older patients.

Methods

Search strategy and selection criteria

This systematic review and meta-analysis followed by PRISMA guidelines (Preferred Reporting Items for Systematic Review and Meta-Analysis).⁴⁶ We systematically searched PubMed, Embase, MEDLINE, Web of Science and Google Scholar for randomized controlled trials published from database inception to July 2018. The search terms included “PD-1”, “PD-L1”, “CTLA-4”, “nivolumab”, “pembrolizumab”, “atezolizumab”, “durvalumab”, “avelumab”, “ipilimumab”, “tremelimumab”, “Yervoy”, “Opdivo”, “Keytruda”, “Tecentric”, “immune checkpoint inhibitor” and “immunotherapy”. Two investigators (BZ and QW) performed an independent search of the above databases. We only included clinical trials published in English. In order to obtain more comprehensive clinical data, we also manually examined the references of final included studies, review articles and other relevant studies. Meanwhile, we conducted an additional search of European Society for Medical Oncology and American Society of Clinical Oncology meeting database using the same search terms.

Trials were included in our meta-analysis according to the following selection criteria: (1) Phase 2 or 3 randomized controlled trials including updated trials for treatment of solid cancer; (2) Trials must evaluate the efficacy of PD-1 inhibitors, PD-L1 inhibitors, CTLA-4 inhibitors, or their combination compared to placebo or other anti-tumor drugs; (3) Trials comparing newer and more effective ICIs (PD-1 inhibitors) with older and less effective ICIs (CTLA-4 inhibitors); (4) Trials clearly reporting the hazard ratio (HR) of overall survival (OS) or progression-free survival (PFS) according to patients’ age; and (5) Trials reporting subgroup efficacy by using 65 y as a cut-off age. We excluded trials that: (1) were single-arm trials (noncomparative clinical trials); (2) did not report the subgroup efficacy by age and not use 65 y as a cut-off age for subgroup analyses; (3) assessed the efficacy of ICIs in hxematologic malignancies; (4) arranged patients to receive ICIs combined with chemotherapy or other non-immunological drugs. Two investigators (BZ and QW) independently reviewed each title and abstract from the list of literature search and independently selected potentially relevant articles according to the inclusion and exclusion criteria. Any discrepancies were resolved through the consensus of all investigators.

Data extraction and quality assessment

The following information was extracted from the eligible articles: (1) Study characteristics (first author, year of publication, treatment regimens in the experimental groups and control groups, clinical trial phase); (2) Patient characteristics (total number of younger patients (＜65 y) available for analysis in each arm, total number of older patients (≥65 y) available for analysis in each arm; (3) Primary outcome (HR for OS in younger and older patients); (4) Secondary outcome (HR for PFS in younger and older patients).

We used the Cochrane risk of bias tool to assess the quality of each included study from the following criteria⁴⁷: (1) randomized Sequence Generation; (2) allocation concealment; (3) blinding of participants, personnel; (4) blinding of outcome assessment; (5) incomplete outcome data; (6) selective outcome reporting; and (7) other sources of bias. Each risk of bias was described as low risk, high risk or unclear risk. Three investigators independently performed data extraction and quality assessment and any disagreement was resolved by consensus. Details on the selection process are shown in Figure 3 and the assessment of risk of bias among included studies is summarized in Figure S1.

Figure 3.

Flow diagram of study selection process.

Statistical analysis

We calculated the pooled HR of OS and PFS in younger and older patients. We selected the corresponding model for calculating the summary HR and 95% confidence intervals (CI) according to heterogeneity between the individual studies. The Q test and I² statistics were used to assess between-study heterogeneity. I² value of <30%, 30–59%, 60–75%, and >75% was considered as low, moderate, substantial, and considerable heterogeneity, respectively. If I² value was less than 50%, we used the fixed-effects model to perform our meta-analysis, otherwise, we used the random-effects model. For trials that reported separate HR estimates for 65–75 and >75 y, a combined estimate (≥65-year-age) was established using the above statistical methods, and we used the combined estimate for final meta-analysis. We performed subgroup analyses to explore the specific correlation between patients’ age and immunotherapy efficacy. The subgroup was cancer types and ICIs types. If there was evidence of significant heterogeneity between the individual studies, we conducted a sensitivity analysis by removing one study at a time to examine the stability of the combined results. We evaluated the potential publication bias by the Begg’s test and Egger’s test.^48,49 We evaluated the heterogeneity in HR between the two groups by using the interaction test (P_{heterogeneity}). All statistical analyses were done with StataSE12.0 software (StataCorp, College Station, Texas) and the graphs of risk of bias were generated by Review Manager 5.3 software (Nordic Cochrane Center, Copenhagen, Denmark). P ≤ 0.05 was considered statistically significant.

Funding Statement

This study was funded by the National Natural Science Foundation of China (No. 81401796 and 81572267), Key Research and Development Projects of Jiangsu Province (BE2017681), Six Talent Peaks Program of Jiangsu Province (WSW-038), Jiangsu Post-doctoral Program (1701008A), Jiangsu youth medical talent project (QNRC2016702) and China Post-doctoral program (2017M621800).

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website.