Treatment and Outcomes of Oropharyngeal Cancer in People with Human Immunodeficiency Virus

Abstract

HIV-positive people are at increased risk for malignancies associated with human papillomavirus (HPV) infection, including oropharyngeal squamous cell carcinoma (OPSCC). The purpose of this study was to determine whether cancer treatment disparities exist between HIV-positive and HIV-negative people with OPSCC. We conducted a retrospective cohort study comparing OPSCC treatment adequacy and treatment outcomes in HIV-positive and HIV-negative people in the post-antiretroviral therapy era. Treatment adequacy was determined by measuring two primary endpoints associated with OPSCC survival: time to therapy and total radiation dose. Treatment outcomes were assessed by measuring disease-free and overall survival. We identified a total of 37 HIV-positive and 149 HIV-negative people with OPSCC. HIV-positive people experienced a median delay of 10 days from time of OPSCC diagnosis to start of therapy compared with HIV-negative people [hazard ratio (HR) 0.61, 95% confidence interval (CI) 0.38–0.98]. Total post-radiation dose in HIV-positive people was lower than that in HIV-negative people [58.5 Gray (Gy) versus 64.4 Gy, p = .04]. HIV-positive people also experienced greater hazards for disease recurrence (HR 3.43, 95% CI 1.39–8.46) and death (HR 4.21, 95% CI 1.29–13.80) compared with HIV-negative people. In conclusion, we detected a clinically important delay in time to therapy as well as worse disease-free and overall survival in HIV-positive people with OPSCC compared with their HIV-negative counterparts. These findings are relevant to understanding how HIV-positive people are diagnosed and undergo therapy for HPV-associated malignancies and highlight the need to address cancer treatment disparities in this group.

Introduction

Approximately one third of deaths in HIV-positive people are attributed to malignancy.¹ Linkage studies of HIV and cancer registries show that cancer-specific mortality remains elevated in HIV-positive people compared with the general population.^2,3 Delays in initiation of cancer therapy are associated with worse outcomes⁴ and may contribute to treatment disparities in HIV-positive people,⁵ highlighting the need for further understanding of cancer treatment in this group.

HIV-positive people are at particular risk for malignancies associated with human papillomavirus (HPV) infection.^6–8 This includes oropharyngeal squamous cell carcinoma (OPSCC), a type of head and neck cancer (HNC) that is distinct because of a high proportion of HPV-associated tumors.⁹ The incidence of HPV-associated OPSCC in the general population has increased by ∼7.5% per year since 1984 such that ∼70% of OPSCC is now attributed to HPV.⁹ The number of cases of HPV-related OPSCC is expected to double between now and 2020.⁹ The risk is especially high in HIV-positive people; studies show a two- to sixfold increase in risk of OPSCC compared with the general population in this group.^7,8,10

Despite this, little is known to guide treatment of HIV-positive people with OPSCC because they have been historically excluded from most cancer clinical trials.¹¹ To date, retrospective studies of HNCs in HIV-positive people have not provided information on OPSCC specifically,^12–16 even though each HNC type comprises a separate malignancy with type-specific treatment and prognosis.¹⁷ The objective of our study was to determine if cancer treatment disparities exist between HIV-positive and HIV-negative people with OPSCC. For this, we conducted a retrospective cohort study comparing OPSCC treatment adequacy and treatment outcomes in HIV-positive and HIV-negative people.

Methods

Study endpoints

Treatment adequacy was determined by measuring two primary endpoints associated with inferior OPSCC survival: time to therapy and total radiation dose.^18–21 Treatment outcomes were assessed by measuring disease-free and overall survival. Because HPV tumor expression is associated with favorable OPSCC outcome,²² we additionally performed exploratory analyses of overall and disease-free survival by HPV status.

Study population

We obtained data of people treated for OPSCC at four tertiary care centers in the United States: the University of Pennsylvania, Thomas Jefferson University, Temple University, and the University of Alabama at Birmingham (UAB). The eligible population consisted of adults diagnosed with potentially curable (stages I through IVb) OPSCC in the post-antiretroviral therapy (ART) era (start date January 1, 1996).

This study was considered to be exempt research with respect to the requirement for informed consent as assessed by the institutional review boards of the University of Pennsylvania, Thomas Jefferson University, Temple University, and the UAB. All procedures performed in this study were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.

Data collection

To create our cohort, we performed individual-level chart review following preliminary identification of potential participants from billing records and/or data warehouses. The same prespecified criteria were used to standardize OPSCC diagnosis across study sites. OPSCC diagnosis was based on the presence of tissue biopsy or surgical specimen documenting invasive squamous cell carcinoma in the tonsils, base of tongue, soft palate or side, and posterior pharyngeal walls.¹⁷ The presence of invasive disease was necessary; specimens with squamous cell carcinoma in situ were insufficient to establish a diagnosis of OPSCC. Lymph node biopsies and cytology specimens were similarly not diagnostic of OPSCC.

OPSCC staging was determined per Tumor, Node, Metastasis guidelines of the American Joint Committee on Cancer.²³ When available, HPV tumor status was determined from pathology reports. We furthermore completed HPV testing on available stored samples that predated the 2011 National Comprehensive Cancer Network guidance for routine testing of all OPSCC samples.¹⁷ For this, we used immunohistochemistry staining for p16, an established and well-validated surrogate marker of HPV expression.²⁴ Positive staining was defined as the presence of strong and diffuse nuclear and cytoplasmic staining in >70% of tumor cells.²⁴ Samples positive for p16 then underwent HPV immunohistochemistry to enhance specificity.²⁵

To measure OPSCC treatment adequacy and treatment outcomes, we defined time to therapy as the difference in days between diagnosis and initiation of therapy. Date of diagnosis was defined as the day when a biopsy or surgical specimen that documented the presence of invasive squamous cell carcinoma was obtained. Because oropharyngeal cancer (OPSCC) is often locally advanced on presentation,²⁶ treatment often involves more than one modality, with options including definitive combined chemotherapy and radiation, surgical resection followed by postoperative combined chemotherapy and radiation, and induction chemotherapy followed by surgery or combined chemotherapy and radiation. We defined initiation of therapy as the first day an individual received any therapy. Total radiation dose was measured in Gray (Gy). Overall survival was defined as the number of days from date of diagnosis to death from any cause or last known follow-up, and disease-free survival was defined as the number of days from diagnosis to disease recurrence or last known disease-free follow-up.

Statistical analysis

To compare basic demographics and clinical characteristics of HIV-positive and HIV-negative people with OPSCC, we used unpaired t-tests for normally distributed continuous variables and the Wilcoxon rank-sum test for nonnormally distributed continuous variables. For categorical variables, we used chi-squared tests or Fisher's exact tests when expected values were small. We used Cox proportional hazards models to compare time to therapy, overall survival, and disease-free survival, and linear regression to compare total amount of radiation. For overall survival, therapy was analyzed as a time-dependent covariate to measure the effect of therapy itself on survival. We used Schoenfeld residuals to test the proportionality of hazards assumption. Analysis of total radiation dose was stratified by the presence of concomitant surgical resection (yes/no) to account for differences in total recommended radiation dose for people undergoing definitive versus postoperative radiation.¹⁷ Model fit was assessed by examining residual plots. Hazard ratios (HRs) and odds ratios for which the 95% confidence intervals (CIs) did not include 1.0 were considered statistically significant. While we recognize the presence of several distinct prespecified endpoints, we chose to focus on the magnitude of clinically relevant effects rather than adjust for multiple comparisons.

Confounding

We matched HIV-positive and HIV-negative people by institution and year of cancer diagnosis to account for treatment and survival differences due to study site and advances in therapy with time. Potential confounders included age, sex, race, oropharyngeal cancer stage, and medical comorbidities. We selected the modified Cumulative Illness Rating Scale (CIRS) to measure medical comorbidities because it captures a wide range of conditions (including psychiatric illness, substance abuse, and tobacco use) without including HIV infection.²⁷ The effect of tobacco was therefore captured by the CIRS score in our analysis. For our treatment adequacy endpoints (time to therapy and radiation dose), medical comorbidities were not included as potential confounders because of concern that they were in the causal pathway between HIV infection and delays in starting and/or tolerating cancer therapy. Insurance type was similarly not analyzed as a potential confounder because differential health care access may lie within the causal pathway between HIV infection and cancer therapy as well as HIV infection and OPSCC survival. Potential confounders were included in multivariable analysis based on clinical relevance or if univariable analysis showed evidence of confounding (≥10% change in the measure of association) and/or predictive effects (p < .20).

Results

Baseline patient characteristics

We identified a total of 37 HIV-positive and 149 HIV-negative people with OPSCC (Table 1). Most cases were from the University of Pennsylvania. Both HIV-positive and HIV-negative cohorts were composed primarily of male and Caucasian participants, and all were insured with either Medicare, Medicaid, or private insurance. OPSCC staging was similar between HIV-positive and HIV-negative groups, with the majority of people presenting with locally advanced disease (stages III–IVb). HPV tumor status was determined from pathology reports in 75 people and from additional p16 testing on available stored specimens in another 49 people.

Table 1.

Baseline Characteristics of HIV-Positive and HIV-Negative Groups

	HIV-positive (N = 37)	HIV-negative (N = 149)	p
Study center, n (%)
University of Pennsylvania	24 (65)	111 (75)
Temple University	4 (11)	15 (10)
Thomas Jefferson University	2 (5)	8 (5)
University of Alabama at Birmingham	7 (19)	15 (10)
Age in years, mean (SD)	52 (10.9)	59 (8.8)	<.01
Sex, n (%)
Male	34 (92)	110 (74)	.02
Female	3 (8)	39 (26)
Race, n (%)
White or Caucasian	19 (51)	124 (83)	<.01
Black or African American	13 (35)	17 (12)
Other	0 (0)	2 (1)
Not available	5 (14)	6 (4)
Insurance, n (%)
Medicaid or Medicare	23 (62)	36 (28)	<.01
Private	14 (38)	95 (72)
None	0 (0)	0 (0)
Not available	0 (0)	18 (12)
Oropharyngeal cancer stage, n (%)
Stages I–II	4 (11)	14 (9)	.07
Stages III–IVb	30 (81)	133 (89)
Not available	3 (8)	2 (13)
p16 Status, n (%)
Positive	14 (38)	81 (54)	.05
Negative	9 (24)	20 (14)
Not available	14 (38)	48 (32)
Cumulative Illness Rating Scale, median (IQR)	8 (5–10)	5 (3–7)	<.01
CD4 cells per mm3, median (IQR)	341 (65–500), n = 27
Viral load
Undetectable on antiretroviral therapy (<400 copies/mL), n (%)	14 (58%)
Median copies/mL, log10 (if detectable; IQR)	4.38 (3.80–5.37)
	n = 24

Notable differences included a younger age at time of OPSCC diagnosis, a higher proportion of male and African American people, and a higher median CIRS score in our HIV-positive group. HIV-positive people were also more likely to have Medicaid or Medicare rather than private insurance and to be active smokers at time of OPSCC diagnosis.

Oropharyngeal cancer treatment

Median and total follow-up times were 30 and 4,953 days, respectively (Fig. 1). HIV-positive people had a lower hazard of starting therapy and a median delay of 10 days from pathological diagnosis to start of therapy compared with HIV-negative people (Table 2). There was no evidence of violation of the proportionality of hazards assumption using Schoenfeld residuals (p = .97).

FIG. 1.

Kaplan–Meier estimates of time to therapy in patients with OPSCC, by HIV status. OPSCC, oropharyngeal squamous cell carcinoma.

Table 2.

Oropharyngeal Cancer Treatment and Outcomes

	HIV-positive, median days (IQR)a	HIV-negative, median days (IQR)a	Hazard ratio (95% CI)	Adjusted hazard ratio (95% CI)b
Time to therapy	38 (28–73) n = 25	28 (15–46) n = 96	0.56 (0.36–0.88)	0.61 (0.38–0.98)
Time to recurrence	128 (28–287) n = 25	301 (213–853) n = 123	3.00 (1.42–6.38)	3.43 (1.39–8.46)
Time to death	128 (66–1,317) n = 30	905 (413–1,536) n = 116	3.11 (1.28–7.52)	4.21 (1.29–13.80)
Time to death (therapy as time-dependent covariate)	188 (401–1,536) n = 22	905 (401–1,536) n = 83	3.88 (1.49–10.10)	4.15 (1.14–15.12)

^aModel-based median days.

^bTime to therapy adjusted for age, sex, race, and oropharyngeal cancer stage. All other variables adjusted for age, sex, race, oropharyngeal cancer stage, and comorbidities.

CI, confidence interval.

The total unadjusted radiation dose for individuals receiving definitive radiation was 64.2 Gy for HIV-positive people (n = 11, 95% CI 59.1–69.4) and 69.3 Gy for HIV-negative people (n = 52, 95% CI 67.0–71.7) (p = .08). The total unadjusted radiation dose for individuals receiving postoperative radiation was 59.2 Gy for HIV-positive people (n = 3, 95% CI 54.5–64.0) and 63.3 Gy for HIV-negative people (n = 39, 95% CI 62.0–64.7) (p = .10). After adjustment for age, sex, race, and oropharyngeal cancer stage, a statistically significant difference was detected in the postoperative radiation group only. Total adjusted radiation dose for individuals receiving definitive radiation was 59.9 Gy for HIV-positive people (95% CI 46.8–72.3) and 64.7 Gy for HIV-negative people (95% CI 49.5–79.9) (p = .14), whereas total adjusted radiation dose for individuals receiving postoperative radiation was 58.5 Gy for HIV-positive people (95% CI 48.7–68.4) and 64.4 Gy for HIV-negative people (95% CI 54.9–73.9) (p = .04).

Disease-free and overall survival

The median and total follow-up times were 2.5 and 644 years, respectively (Fig. 2). HIV-positive people were more likely to experience both disease recurrence and death compared with HIV-negative people (Table 2). There was no evidence of violation of the proportionality of the hazards assumption for either the disease-free or overall survival (p = .73 and p = .73, respectively).

FIG. 2.

Kaplan–Meier estimates of (A) disease-free survival and (B) overall survival in patients with OPSCC, by HIV status.

A separate analysis with therapy as a time-dependent covariate evaluated whether delays in initiating OPSCC therapy might be responsible for the greater hazard for death in the HIV-positive group. We noted no change in the HR with therapy as a time-dependent covariate among people for whom both treatment and survival data were available (Table 2).

Discussion

Our study detected a median 10-day delay in initiation of therapy in HIV-positive people with OPSCC, a delay that is known to be associated with worse clinical outcomes,¹⁹ as well as inferior disease-free and overall survival. Small decreases in cumulative radiation dose were also noted in people undergoing either definitive or postoperative radiation, although the results were only statistically significant in our adjusted postoperative radiation analysis. The relevance of this small difference is unclear. While older studies of HNC treatment show that a difference of 10 Gy is associated with worse outcomes,^20,21 more recent work suggests that even smaller differences are important.²⁸ Our findings indicate the presence of significant cancer treatment disparities in HIV-positive people with OPSCC.

Previous studies of HIV-positive people did not analyze OPSCC specifically and instead pooled data from series of people with different head and neck malignancies, which are inherently heterogeneous with different etiologies.^12–16 Furthermore, direct comparison groups were often missing. For example, one relatively recent study found worse locoregional control and survival following chemotherapy and radiation in people with HNC compared with the observed locoregional control and survival rates from their overall institutional experience.¹⁴ A second study found no survival difference in people with HIV compared with Surveillance, Epidemiology, and End Results (SEER) data from the general population.¹² Our study specifically analyzed one type of HNC (OPSCC) and focused on the outcomes in HIV-positive people compared with the HIV-negative cohort to allow for direct comparison of outcomes.

Delays in therapy are likely due to a combination of patient-driven, provider-driven, and health systems factors.²⁹ Patient-driven factors may include greater propensity for medical comorbidities that affect the ability of HIV-positive people to receive cancer care. Providers in turn may hesitate to offer therapy to HIV-positive people before further workup due to concerns regarding their ability to tolerate treatment. Although we do not have data on the reasons for cancer treatment delays, the higher comorbidity scores in our HIV cohort make it plausible that comorbidities contributed.³⁰ Personal and health system factors may have further contributed to the observed delay. Black race and the absence of private insurance are associated with lack of or delays in obtaining cancer treatment,^31,32 and our HIV-positive group did have a significantly higher proportion of African American individuals and people on either Medicaid or Medicare. Our study, however, was not designed to determine whether delays in time to therapy were related to race or insurance status or to elucidate the exact factors leading to delays of therapy. These questions represent important lines of future work.

The lower survival in our HIV-positive group is consistent with previous population-based studies that evaluated cancer-specific mortality in HIV-positive people,^2,3,33 although none specifically demonstrated lower survival for OPSCC. In our study, analysis of therapy as a time-dependent covariate in individuals for whom therapy and survival data were available showed no difference in the HR for overall survival before and after the start of cancer therapy. This suggests that delays in initiation of cancer therapy are not responsible for the worse overall survival in our HIV-positive group. However, our analysis without therapy as a time-dependent covariate did show worse disease-free survival in HIV-positive people, suggesting that at least some of the decrease in overall survival is cancer-specific, making it difficult to entirely dismiss an effect of timely therapy on overall survival. For a disease like locoregionally advanced OPSCC, the short time from diagnosis to start of cancer therapy may not have been sufficient time for death to occur, such that a difference in the HR before and after start of cancer therapy was not detected.

Only 58% of HIV-positive people in our study were on effective ART and had undetectable viral loads at time of diagnosis. Current practice is to begin ART soon after HIV diagnosis regardless of immunologic status given the benefits of early initiation, including improved survival.³⁴ Prompt initiation of or maintenance of ART is also favored for HIV-positive people with cancer. Although data are limited for most malignancies, HIV-positive people with lymphoma respond better to chemotherapy and have improved survival when on ART.^35–37 Thus, the worse survival in our HIV-positive cohort may be related to the high proportion of people with uncontrolled HIV infection, which may have made it harder for people to complete treatment or even affected treatment response itself.

Our study has several limitations in addition to those inherent to retrospective cohorts. The potential for diagnostic misclassification is inherent to all retrospective multicenter studies. Hence, we used individual chart review on all potential participants and the same prespecified diagnostic criteria across study sites to standardize the diagnosis of OPSCC. While we identified a relatively large number of HIV-positive people with OPSCC, we ultimately did not reach our targeted sample size. This precluded us from making definitive conclusions regarding the impact of radiation dose and, although we collected existing data on HPV tumor status and performed p16 testing on available older specimens, our sample size was ultimately too small to adequately stratify survival analysis by HPV status.²² Additionally, we were unable to obtain cancer treatment data on all individuals for whom survival information was available, thus limiting our ability to fully control for the effects of therapy on survival.

Finally, generalizability should be considered. The majority of our people presented with locally advanced disease, which is consistent with the presentation of OPSCC in the general population.³⁸ However, results may not be as applicable to people with early (stages I–II) disease given their limited representation in our cohort. Our findings are not pertinent to people with distant metastases at diagnosis. This group was excluded due to different management and goals of care compared with people with curable disease. All HIV-positive people in our study were insured, even though HIV-positive people with cancer are more likely to be uninsured or underinsured compared with the general population.³⁹ Hence, the delays in initiation of cancer therapy in HIV-positive people with OPSCC in the general population may be greater than what we describe in our study.

Our study provides important data to understand how HIV-positive people are diagnosed and undergo therapy for an increasingly common HPV-associated malignancy. The presence of a clinically important delay in time to therapy, coupled with worse disease-free and overall survival in our HIV-positive group, highlights the need to understand the factors driving this disparity with the eventual goal of designing programs that provide additional support for HIV-positive people to access cancer treatment, additional information to oncologists on how to approach therapy for this group of people, or both.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This work was supported by a pilot grant from the University of Pennsylvania Abramson Cancer Center and Penn Center for AIDS Research (P30-AI-045008-15). Dr. C.E.B. was furthermore supported by the National Institutes of Health (NIH) T32 training grant (CA009679-22-S1).