The associations of CD4 count, CD4/CD8 ratio, and HIV viral load with survival from non-small cell lung cancer in persons living with HIV

Abstract

The relationship between HIV status and survival from non-small cell lung cancer (NSCLC) is not fully elucidated. Among NSCLC patients in the Bronx, NY, we assessed 1) the associations of CD4 count, CD4/CD8 ratio and HIV viral load (VL) with survival and 2) prognostic factors among persons living with HIV (PLWH). We compared survival from NSCLC diagnosis (2004-2017) between HIV-negative persons (HIV-, n=2,881) and PLWH (n=88) accounting for clinical and sociodemographic factors. In separate comparisons to HIV-, PLWH were dichotomized by CD4 (<200 vs. ≥200cells/μL), CD4/CD8 (median, <0.43 vs. ≥0.43) and VL (<75 vs. ≥75copies/mL) at NSCLC diagnosis. Among PLWH, we assessed the relationships of CD4, CD4/CD8, and VL with survival, adjusting for age, sex, and cancer stage. PLWH with CD4<200cells/μL had lower survival than HIV- [hazard ratio, 95% confidence interval [HR(95%CI)]=1.86(0.98-3.55)]. Survival was similar between PLWH with CD4≥200cells/μL and HIV-[HR(95%CI)=0.90(0.61-1.33)]. Results were similar when categorizing PLWH by CD4/CD8 [vs. HIV-: low CD4/CD8: HR(95%CI)=1.74(1.07-3.89); high CD4/CD8: HR(95%CI)=0.63(0.37-1.07)] and VL [vs. HIV-: <75copies/mL: HR(95%CI)=0.74(0.46-1.21), ≥75copies/mL: HR(95%CI)=1.41(0.88-2.27)]. Among PLWH, CD4<200cells/μL was associated with worse survival [vs. CD4≥200cells/μL: HR(95%CI)=2.37(1.14-4.92)]. CD4, CD4/CD8, and VL may be prognostic markers for PLWH with NSCLC, suggesting immune status may be important in NSCLC survival among PLWH. Those with severe immune suppression from HIV may additionally face unique barriers to NSCLC care driving survival differences.

Introduction

In the United States (US), lung cancer incidence in persons living with human immunodeficiency virus (PLWH) is higher than that of the general population(Kirk, Merlo, & Lung, 2011; Keith Sigel et al., 2012). This has been attributed to higher smoking rates, chronic pulmonary inflammation, and immunodeficiency(Keith Sigel et al., 2017; Winstone, Man, Hull, Montaner, & Sin, 2013). Some reports have shown that among those with lung cancer, the leading cause of US cancer mortality (“Lung Cancer Statistics,”), PLWH have worse survival than those who are not infected(K. Sigel et al., 2013; Suneja et al., 2013); others have shown no difference(Powles et al., 2003; Rengan, Mitra, Liao, Armstrong, & Vachani, 2012). These conflicting findings may be attributed to between-study differences in the population of PLWH and the availability of highly active antiretroviral therapy (HAART) when these studies were conducted(Winstone et al., 2013). Prognostic factors among PLWH with lung cancer are relatively understudied. Cancer stage and performance status (i.e., patients’ level of functioning) have been established as prognostic factors, but there are mixed reports regarding HAART use and CD4 counts(Lavole et al., 2009; Makinson et al., 2011; Winstone et al., 2013; Zheng et al., 2018).

The Bronx, New York City has been at the epicenter of the HIV epidemic since it began(Lambert, 1989). In the Bronx, there were 31,280 living cases of HIV reported in 2019, representing 2.2 percent of Bronx residents (“New York City Annual HIV/AIDS Surveillance Statistics 2019,” 2020). Using an established cohort of Bronx lung cancer patients(Klugman et al., 2019), we evaluated the relationship between HIV status and survival from non-small cell lung cancer (NSCLC), which represents 85-90% of lung cancers in the US(Molina, Yang, Cassivi, Schild, & Adjei, 2008). In addition, we assessed prognostic factors among PLWH with NSCLC. We hypothesized that those with markers of worse immunosuppression and those with detectable HIV viral load (VL) have worse survival compared to HIV-negative (HIV-) patients and PLWH with less immunosuppression and undetectable VL.

Materials and Methods

Cohort and selection criteria:

The Lung Cancer Clinical Cohort at Montefiore Medical Center (MMC) has been described previously(Klugman et al., 2019). Briefly, the cohort included persons aged 18+ years diagnosed at MMC, a tertiary care referral center, between 2004-2017 with incident primary lung cancer (n=5,411). Those with NSCLC were included in this analysis [excluding those with unknown lung cancer type (n=976) or small cell lung cancer (n=516)]. Additionally, the study cohort included individuals who either had any of their first course of treatment administered at MMC, or had a decision to not treat the cancer determined at MMC (designated as “analytic cases” for treatment and survival analysis by the Commission on Cancer (“FORDS: Facility Oncology Registry Data Standards Revised for 2016,” 2016)). We excluded 950 “non-analytic” cases, representing 13% of PLWH and 23% of HIV-negative patients. The final analytic cohort contained 2,969 patients: 88 PLWH and 2,881 HIV-.

Variables:

Sociodemographic and tumor-related variables were obtained from the Montefiore/Einstein Cancer Registry and MMC’s electronic medical records. These variables included age at cancer diagnosis; diagnosis year; race; ethnicity; marital status; language preference; cancer stage, type and histology; and treatment history. Smoking status (never, former, current) was determined at cancer diagnosis. Socioeconomic status (SES) was based on census block group and was reported as a summary score of 6 variables relative to the New York State mean(Diez Roux et al., 2001; Klugman et al., 2019). HIV-related variables, including age at HIV diagnosis, HIV laboratory markers (CD4 count, CD4 nadir, CD4/CD8 ratio, and HIV VL), and history of intravenous drug use and positive hepatitis C virus serology, were obtained through the Einstein-Rockefeller-CUNY Center for AIDS Research (ERC-CFAR) Clinical Cohort Database, which contains data on all confirmed PLWH and all confirmed HIV- patients receiving care at MMC (Felsen, Bellin, Cunningham, & Zingman, 2014; Hanna et al., 2016). All PLWH who were included were part of the ERC-CFAR cohort, as all patients came from the same source population. Those with unknown HIV status (n=2,290) were presumed to be HIV-. We defined laboratory markers at cancer diagnosis as the temporally nearest measurements preceding the lung cancer diagnosis date [CD4 count (n=75/88): median measurement: 42 days prior to cancer diagnosis, interquartile range (IQR): 17 to 91 days prior; CD4/CD8 (n=75/88): 72 (22-250) days prior; VL (n=75/88): 49 (22-106) days prior]. CD4 count was dichotomized by the threshold for AIDS diagnosis: high (≥200 cell/μL) and low (<200 cells/μL), and CD4/CD8 ratio was dichotomized by the median: high (≥0.43) and low (<0.43). Detectable HIV VL was defined as ≥75 copies/mL, the lowest common detection limit of HIV RNA assays used during the study period.

Vital status was assessed using the Montefiore/Einstein Cancer Registry and MMC’s electronic medical record (EMR). Vital status was ascertained using reported patient hospitalizations, known patient readmissions, contact with the patient’s physician, and/or direct patient contact (“FORDS: Facility Oncology Registry Data Standards Revised for 2016,” 2016). In addition, a National Death Index (NDI) search is conducted periodically among PLWH in the ERC-CFAR Clinical Cohort Database who are lost to follow-up, with the most recent match occurring through 2017. Because this source of death information was available only for PLWH (identifying 9 additional deaths), NDI information was used only in analyses restricted to PLWH.

Statistical analysis:

Associations between baseline demographic and tumor-related characteristics and HIV status were tested using Pearson’s chi-square test (categorical variables) and Student’s t-test or Mann-Whitney U test (continuous variables). Associations between continuous variables among PLWH were assessed using Pearson correlation coefficient and Spearman’s rho. Survival was computed from lung cancer diagnosis date to death date. Patients who were still alive were censored at the date of last contact. Univariate survival analyses utilized the Kaplan-Meier method, comparing groups using the log-rank test. Using multivariable Cox regression, survival was compared between HIV- and PLWH, with PLWH categorized by various HIV laboratory measures at cancer diagnosis (high vs. low CD4 count, high vs. low CD4/CD8 ratio, and detectable vs. undetectable HIV VL). This association was adjusted for age, sex, race/ethnicity, cancer stage, treatment (chemotherapy, radiation, surgery, palliative care), smoking status, histology, marital status, language preference, and socioeconomic status, consistent with previously published analyses of NSCLC among MMC patients(Klugman et al., 2019). Additional analyses included an analysis restricted to only those with known HIV status and an analysis no longer adjusting for race/ethnicity. The purpose of the sensitivity analysis was to compare those who were confirmed HIV-negative with confirmed PLWH; those who have not been tested for HIV may differ from those who have been. Furthermore, removing race/ethnicity from the model allowed for increased sample size.

For analyses of PLWH only, variables associated with survival in univariate analyses (p<0.20) were initially considered for inclusion in the final multivariable model in addition to age and sex. HIV-related laboratory measures associated with survival in univariate analysis included CD4 count, CD4/CD8 ratio, and VL, and these variables were kept in the final multivariable model regardless of significance. Non-HIV-related variables (other than age and sex) were removed if no longer statistically significant when part of the multivariable model. The final model consisted of age, sex, cancer stage, CD4 count, CD4/CD8 ratio, and VL. In a sensitivity analysis, multiple imputation including potentially important clinical characteristics and patient demographics in the imputation process was performed to include PLWH with missing HIV laboratory data. The proportional hazards assumption was tested graphically for each of the covariates using log-log plots and statistically using Schoenfeld residuals. No violations were found. All statistical analyses were performed using Stata v15.1. Two-tailed p-values<0.05 were considered statistically significant.

Results

Descriptive characteristics:

From 2004-2017, PLWH (n=88) were diagnosed with NSCLC at a younger age than HIV- (n=2,881) [median (IQR): 57 (52-62) and 69 (61-77) years, respectively] (Table 1). Female sex was less common among PLWH than HIV- (40 vs. 51%, respectively). There were differences in race/ethnicity, with 44 (64%) Non-Hispanic Black patients among PLWH and 794 (39%) among HIV-. Fifty-eight percent of PLWH were current smokers vs. 37% among HIV-. There were no statistically significant differences in cancer stage and adenocarcinoma proportion (local stage: PLWH: 15%; HIV- patients: 22%; adenocarcinoma: PLWH: 58%, HIV-: 56%).

Table 1.

Descriptive characteristics of patients with non-small cell lung cancer in the Lung Cancer Clinical Cohort at Montefiore Medical Center, stratified by HIV status

Variable	PLWH (n=88)	HIV- (n=2881)	p
Age – median years (IQR)	57 (52-62)	69 (61-77)	<0.001
	N (%)
Female sex	35 (40)	1477 (51)	0.03
Race/ethnicity 1			<0.001
Non-Hispanic Black	44 (64)	794 (39)
Non-Hispanic White	8 (12)	692 (34)
Hispanic	17 (25)	528 (26)
Non-Hispanic Asian	0 (0)	43 (2)
Smoking status 2			<0.001
Never Smoker	5 (6)	406 (15)
Former Smoker	32 (36)	1337 (48)
Current Smoker	51 (58)	1041 (37)
Cancer stage			0.26
Local	13 (15)	632 (22)
Regional	22 (25)	645 (22)
Distant	33 (38)	1118 (39)
Other/unknown	20 (23)	486 (17)
Adenocarcinoma	51 (58)	1625 (56)	0.77
Cancer treatment receipt 3
Surgery	29 (33)	942 (33)	0.96
Chemotherapy	33 (38)	1153 (40)	0.63
Radiation	28 (32)	1041 (36)	0.41
Targeted therapy4	11 (13)	242 (8)	0.18
Diagnosis year			0.16
2004-2008	18 (20)	849 (29)
2009-2013	36 (41)	1104 (38)
2014-2017	34 (39)	928 (32)
Language preference 5			0.15
English	75 (90)	2260 (85)
Not English	8 (10)	410 (15)
Insurance 6			0.30
Public or None	65 (79)	1862 (74)
Private	17 (21)	657 (26)
Marital status 7			<0.001
Not married	70 (82)	1638 (62)
Married	15 (18)	995 (38)

¹Data available for 69 (78%) of PLWH and 2057 (71%) of HIV-

²Data available for 88 (100%) of PLWH and 2784 (97%) of HIV-

³Treatments are not mutually exclusive and can be from any course

⁴Prescription of at least one of 19 molecular therapies for lung cancer listed by the National Cancer Institute as of July 2019(“Targeted Cancer Therapies,”)

⁵Data available for 83 (94%) of PLWH and 2670 (93%) of HIV-

⁶Data available for 81 (92%) of PLWH and 2519 (87%) of HIV-

⁷Data available for 85 (97%) of PLWH and 2633 (91%) of HIV-

In PLWH, age was positively correlated with NSCLC diagnosis year (Spearman’s rho=0.22). The median CD4 count preceding lung cancer diagnosis for PLWH was 388, IQR: 212-640 cells/μL. Thirty-two PLWH (43%) had detectable VL nearest to lung cancer diagnosis. Seventy-three patients (83%) had evidence of HAART medication prescription by the time of death or censoring. CD4 count and CD4/CD8 ratio were highly correlated (Pearson’s r=0.73). Undetectable versus detectable VL was associated with CD4–related measures (chi-square test: CD4 count: p=0.005; CD4/CD8: p<0.0001).

The association of HIV status and survival:

While PLWH and HIV- experienced similar death rates [PLWH: n=61, 69%; HIV-: n=1,857, 64%), median survival time was lower among PLWH (1.1 years, IQR 0.6-1.3 years, person-time of follow-up: 141 years) than among HIV- [1.6 (1.5-1.7) years, person time of follow-up: 6,589 years]. Among PLWH, those with low CD4 counts, low CD4/CD8 ratios, and detectable VL at diagnosis had worse survival than HIV- and PLWH with high CD4 counts, high CD4/CD8 ratios and undetectable VL (Figure 1a–1c). When accounting for clinical and sociodemographic factors, the overall cohort of PLWH had similar survival compared with HIV- (HR=1.06 (0.77-1.45)). However, PLWH with CD4 count<200 cells/μL had a nearly 2-fold increase in risk of death compared with HIV- [HR=1.86, 95% confidence interval (CI): 0.98-3.55], while those with CD4≥200 cells/μL had similar survival to HIV- (Table 2). PLWH with low CD4/CD8 ratio had worse survival than HIV- (HR=1.74, 95% CI: 1.13- 2.68) while PLWH with high CD4/CD8 ratio had similar survival. There was a suggestion that VL distinguished PLWH with similar vs. worse survival than HIV- (undetectable VL vs. HIV-: HR=0.74, 95% CI: 0.46-1.21; detectable VL vs. HIV-: HR=1.41, 95% CI: 0.88-2.27). After restricting the analysis to those with confirmed HIV status (Table S1) and when no longer adjusting for race/ethnicity (Table S2), results were similar. When excluding cancer treatment from the analysis, the magnitude of the survival disadvantage in PLWH with low CD4 counts, low CD4/CD8 ratios, and high viral loads increased, whereas PLWH with high CD4 counts, high CD4/CD8, and low viral loads continued to have similar survival compared to HIV- individuals (Table S3).

Figure 1.

HIV status and survival from NSCLC diagnosis, with persons living with HIV (PLWH) stratified by a) CD4 count, b) CD4/CD8 ratio [median value], and c) HIV viral load [level of detection: 75 copies/mL] at diagnosis.

Table 2.

HIV Status and non-small cell lung cancer survival in the Lung Cancer Clinical Cohort at Montefiore Medical Center

	N	Ndeaths	HR1	95% CI	p	N	Ndeaths	HR2	95% CI	p
CD4 count	2917	1883			0.0007	1843	1137			0.18
HIV-	2842	1831	1.00	Ref		1786	1100	1.00	Ref
PLWH, ≥200 cells/μL	57	36	1.07	0.77-1.49		46	27	0.90	0.61-1.33
PLWH, <200 cells/μL	18	16	3.08	1.88-5.05		11	10	1.86	0.98-3.55
CD4/CD8 ratio	2917	1883			0.0001	1843	1137			0.01
HIV-	2842	1831	1.00	Ref		1786	1100	1.00	Ref
PLWH, ≥0.433	39	19	0.78	0.50-1.23		32	14	0.63	0.37-1.07
PLWH, <0.43	36	33	2.25	1.60-3.18		25	23	1.74	1.13-2.68
HIV viral load	2917	1882			0.0001	1843	1136			0.17
HIV-	2842	1831	1.00	Ref		1786	1100	1.00	Ref
PLWH, undetectable4	43	22	0.82	0.54-1.25		35	17	0.74	0.46-1.21
PLWH, detectable	32	29	2.50	1.73-3.61		22	19	1.41	0.88-2.27

¹Unadjusted analyses for HIV laboratory counts are based on the nearest measurement prior to lung cancer diagnosis.

²Three separate multivariable models, with each HIV status variable adjusted for age, sex, race/ethnicity, cancer stage, treatment, smoking status, histology, marital status, language preference, and socioeconomic status.

³Median measurement

⁴Limit of detection 75 copies/mL

Prognostic factors in PLWH:

In unadjusted analyses among PLWH, later cancer diagnosis year, surgery, earlier cancer stage, undetectable HIV VL, higher CD4 count, and higher CD4/CD8 ratio were associated with improved survival (Table 3). After adjusting for age, sex, and cancer stage, those with CD4 count<200 cells/μL experienced over twice the risk of death than those with CD4 count≥ 200 cells/μL (HR=2.37, 95% CI: 1.14-4.92; Table 4), while neither CD4/CD8 ratio nor VL were associated with survival (Table 4). When utilizing multiple imputation to include those with missing HIV markers, results were unchanged (Table S4). Most causes of death among PLWH were attributed to lung cancer (Table S5).

Table 3.

Factors associated with survival in persons living with HIV and NSCLC in the Lung Cancer Clinical Cohort at Montefiore Medical Center

Variable	Unadjusted HR	95% CI	p
Age (years)	0.99	0.96-1.02	0.33
Sex			0.50
Male	1.00	ref
Female	0.85	0.52-1.38
Diagnosis year	0.92	0.86-0.98	0.015
Race/ethnicity 1			0.15
Non-Hispanic Black	1.00	ref
Non-Hispanic White	0.55	0.19-1.55
Hispanic	1.52	0.81-2.89
Smoking status			0.64
Former/never	1.00	ref
Current	1.26	0.46-3.48
Language preference 2			0.18
English	1.00	ref
Not English	1.60	0.83-3.06
Insurance 3			0.64
Public or None	1.00	ref
Private	1.15	0.64-2.07
Marital status 4			0.79
Not married	1.00	ref
Married	1.09	0.59-2.00
Treatments (receipt vs. non-receipt)
Surgery	0.45	0.26-0.77	0.003
Radiation	1.07	0.65-1.75	0.80
Chemotherapy	0.89	0.55-1.43	0.62
Palliative care	1.61	0.83-3.09	0.16
Targeted Therapy5	0.88	0.45-1.72	0.70
Cancer stage			<0.001
Local (1 or 2)	1.00	ref
Regional (3)	2.83	1.20-6.69
Distant (4)	7.02	2.98-16.54
Other/unknown	1.45	0.53-3.95
Adenocarcinoma			0.54
No	1.00	ref
Yes	0.86	0.54-1.39
HIV viral load 3			0.005
Undetectable (<75 copies/mL)	1.00	ref
Detectable (75+ copies/mL)	2.13	1.25-3.61
CD4 count at diagnosis 6			0.001
≥200	1.00	ref
<200	2.85	1.58-5.15
CD4 nadir prior to cancer diagnosis 6			0.76
≥200 cells/μL	1.00	ref
<200 cells/μL	1.08	0.65-1.80
CD4/CD8 ratio 6			0.008
≥0.43 (median)	1.00	ref
<0.43	2.04	1.20-3.44
Antiretroviral therapy			0.34
No	1.00	ref
Yes	0.73	0.40-1.36
Hepatitis C virus infection history 7			0.59
No	1.00	ref
Yes	1.14	0.70-1.84
Intravenous drug use history 8			0.74
No	1.00	ref
Yes	1.08	0.66-1.80

¹Data available for 69 (78%) of participants

²Data available for 83 (94%) of participants

³Data available for 81 (92%) of participants

⁴Data available for 85 (97%) of participants

⁵Prescription of at least one of 19 molecular therapies for lung cancer listed by the National Cancer Institute as of July 2019(“Targeted Cancer Therapies,”)

⁶All HIV laboratory measurements prior to cancer diagnosis available for 75 (85%) of participants

⁷Positive hepatitis C virus antibody prior to cancer diagnosis

⁸Participant reported injection drug use history at intake in the AIDS Institute Reporting System (“Agency (Client) Intake Forms,”)

Table 4.

The associations of CD4 count, CD4/CD8 ratio, and VL with survival in persons living with HIV and NSCLC in the Lung Cancer Clinical Cohort at Montefiore Medical Center¹

Variable	HR1	95% CI	p
Age (per year)	1.00	0.97-1.04	0.89
Sex			0.23
Male	1.00	Ref
Female	1.47	0.78-2.78
Cancer stage			<0.0001
Local	1.00	Ref
Regional	3.26	1.21-8.79
Distant	8.35	3.04-22.93
Other/unknown	1.48	0.49-4.47
CD4 count 2			0.02
≥200 cells/μL	1.00	Ref
<200 cells/μL	2.37	1.14-4.92
CD4/CD8 ratio 2			0.35
≥0.433	1.00	Ref
<0.43	1.37	0.71-2.62
Viral load 2,4			0.47
Undetectable	1.00	Ref
Detectable	1.29	0.65-2.59

¹Multivariable model that is comprised of the six variables listed in the table. Complete covariate information available for 73/88 (83%) of participants.

²Laboratory counts are based on the nearest measurement prior to lung cancer diagnosis

³Median level

⁴Limit of detection 75 copies/mL

Discussion

We found PLWH with lower CD4, lower CD4/CD8 ratio, and detectable HIV VL at NSCLC diagnosis experienced worse survival than HIV- NSCLC patients, whereas those with higher CD4 count, higher CD4/CD8 ratio, and undetectable VL experienced similar survival. To the best of our knowledge, our study is the first to compare survival after NSCLC diagnosis in PLWH with HIV- when categorizing PLWH by various measures of HIV status. These results are important given that CD4-related values and viral load are potentially modifiable risk factors, via HAART use. In our analysis restricted to PLWH, CD4 count at NSCLC diagnosis was a marker for increased risk of death, independent of viral load and CD4/CD8 ratio at diagnosis. While CD4/CD8 ratio and VL were not associated with survival, this may be due to multi-collinearity among CD4 count, CD4/CD8 ratio, and VL and/or low sample size contributing to insufficient power. To address low sample size, we utilized multiple imputation, yielding similar findings. Further studies with large sample sizes and number of deaths would better be able to evaluate the association of these factors with survival. Even with these caveats, our results suggest that careful co-management of HIV may be important for prognosis of NSCLC.

Prior literature has generally shown that PLWH have worse survival from cancer compared to those without HIV infection(Biggar et al., 2005; K. Sigel et al., 2013), although some studies have not shown a survival difference(Rengan et al., 2012). Our study shows no survival differences when comparing all PLWH with HIV-negative patients. However, PLWH with certain laboratory features have worse survival compared to HIV-, suggesting that the mixed findings in the literature could be related to different populations of PLWH (i.e. different distributions of CD4, CD4/CD8 and viral loads). Thus, lung cancer diagnosis is a critical point for CD4 counts to be monitored in PLWH, to assess prognosis and initiate antiretroviral therapy if not already done. Our findings suggest VL may be an important prognostic marker although findings were not statistically significant, potentially due to low power. When the analysis was re-run without adjustment for treatments, which by definition occur after cancer diagnosis, increased the magnitude of our findings when comparing PLWH with worse markers with HIV- individuals. Treatments could be part of the causal pathway between HIV infection and mortality for PLWH with higher risk markers. We hypothesize that those with poorer HIV markers may experience greater adverse effects from cancer treatment and decreased response from cancer therapy, although this theory is beyond the scope of our study. Other treatment-related barriers may include lack of social support and other issues related to social determinants of health. Among PLWH, stage and CD4 count at cancer diagnosis were prognostic factors, after accounting for age, sex, CD4/CD8, and VL. A previous study of 42 PLWH NSCLC patients in France also found CD4 level, dichotomized at 200 cells/μL, to be associated with survival, as well as with performance status and HAART(Makinson et al., 2011). In the latter cohort, CD4 nadir was not associated with survival, similar to our cohort, where we did not observe an association in univariate analysis (Table 3).

A strength of our study lies in our large sample of lung cancer patients from an established cohort. In the Bronx, MMC treats many PLWH in comprehensive programs, which may enable increased surveillance of these high-risk populations for cancer symptoms and access to cancer treatment. Further, we assessed the role of HIV status in survival when accounting for important clinical and sociodemographic factors. In addition, we evaluated important HIV-related factors such as intravenous drug use, hepatitis C status, antiretroviral therapy use, and laboratory measures of HIV including viral load and immune status. Nevertheless, our findings must be interpreted cautiously given the limited amount of detailed smoking data or clinical information such as performance status and other comorbidities. Furthermore, our results may have been influenced by missing data, particularly unknown HIV status and race/ethnicity. To address the missing HIV status data, we removed patients with unknown HIV status in a sensitivity analysis and found similar results (Table S1). In order to include patients with missing race/ethnicity with otherwise complete covariate information, we conducted an analysis that no longer adjusted for race/ethnicity in our comparison of PLWH with HIV-, and results were similar (Table S2). In addition, while we obtained additional death information using NDI for PLWH, we did not have comparable information for HIV- and thus did not utilize NDI death information in comparisons between PLWH and HIV-. However, given that vital status was assessed in all patients at least annually by the Montefiore/Einstein Cancer Registry, we do not believe this potential bias strongly influenced our findings. Finally, our results are based on all-cause mortality rather than lung cancer-specific mortality. The higher risk of death seen in a subset of the PLWH population perhaps could be attributed to death from AIDS-related complications. However, the cause of death information captured by the NDI showed that known deaths in PLWH were overwhelmingly attributed to lung cancer (78%, Table S5). In our PLWH-only analysis, there were no significant differences in the proportion of deaths attributed to lung cancer among those with low versus high CD4 counts (83 vs 63%, p=0.09). While this analysis cannot lead us to conclude that these HIV-related markers lead to any change in lung cancer natural history, the NDI death information supports that our survival analyses are largely related to lung-cancer specific mortality.

Future research should continue to explore the unique considerations for the prevention, screening for and prognosis of lung cancer in PLWH. We found PLWH to have high smoking rates compared to HIV- with NSCLC. Our findings, combined with the known high prevalence of smoking in the general PLWH population(Park, Hernandez-Ramirez, Silverberg, Crothers, & Dubrow, 2016), demonstrate the need for continued investment in smoking cessation efforts to prevent future lung cancer cases. Similar to previous studies, we showed that PLWH with lung cancer are diagnosed at younger ages(Brock et al., 2006; Shiels et al., 2017). The lung cancer screening guidelines were recently updated to include younger patients and those with a 20 pack-year history(“Lung Cancer: Screening,” 2021), increasing the eligibility of PLWH for screening. In addition, we found a positive correlation between diagnosis year and age of lung cancer diagnosis in PLWH, which may reflect the aging PLWH in the Bronx and increased risk of lung cancer with age. Future studies should also further explore relationships between HIV-related factors and survival in PLWH. It has been hypothesized that HIV-associated immunosuppression directly affects tumor behavior and cancer outcomes(Coghill, Suneja, Rositch, Shiels, & Engels, 2019). Additional research on optimal HAART management to minimize immune suppression in PLWH with NSCLC is warranted. In addition, understanding HIV-related factors and cancer survival has implications for cancer therapies. A recent review suggested immune checkpoint inhibitor therapy may be safe and efficacious for PLWH with advanced cancer(Cook & Kim, 2019) and is an area of active research(“Immunotherapy in Patients With HIV Infection and Advanced Cancer,” 2019; “Treatment With Immunological Checkpoint Inhibitors of HIV-infected Subjects With Cancer (PembroHIV),” 2018). With the myriad of recently developed therapies for NSCLC, the disease is becoming increasingly treatable. Persons living with HIV are an important group of NSCLC patients who should also benefit from these therapies while not losing sight of their HIV management.