With the advent and subsequent introduction of effective combination antiretroviral therapy (ART) in 1996, the mortality rate for individuals infected with HIV in the United States has decreased by 70%.1 Nonetheless, there were an estimated 37,600 new HIV infections in the United States in 2014 and more than 1.1 million individuals in the United States are living with HIV.2 With reduced HIV-related mortality and a growing and aging HIV-infected population, there is concern that non-AIDS–defining cancers (NADCs) may be a future source of considerable morbidity and mortality.3–6 Indeed, the proportion of all deaths attributable to NADCs has increased, with as many as 20% of deaths among people living with HIV (PLWH)/AIDS due to invasive NADCs.7–9
PLWH have excess risk for certain NADCs compared with the general population. The likely reasons are multifactorial, including increased life expectancy, premature aging, loss of control of oncogenic infections due to HIV-related immune suppression, and increased co-infections with hepatitis and human papillomavirus.10–13 PLWH also have higher rates of tobacco and alcohol use, and use of ART in HIV has been associated with increase in and accumulation of abdominal fat, dyslipidemia, and insulin resistance.14–17 Given this risk factor profile, PLWH may be at especially high risk for esophageal and stomach cancers. However, although some studies have reported increased risks of esophageal and stomach cancers in HIV-infected individuals, others have reported no association.18–22 Because there is substantial etiologic heterogeneity for subcategories of these cancers, it is important to examine risk according to specific histologic or anatomical subtypes of esophageal and stomach cancers. This is especially important given the observation that certain squamous cell cancers are more commonly diagnosed in PLWH compared with adenocarcinomas. No previous study has examined the association between HIV and risks of specific subtypes of esophageal and stomach cancer.
We, therefore, conducted a large retrospective cohort study to examine the risk of esophageal and stomach cancer subtypes among patients with HIV infection seen in the US National Veterans Health Administration (VHA) system. We also examined risk factors for these cancers among HIV-infected veterans.
Study Population and Design
We performed a retrospective cohort study using individual-level patient data from the national VA Corporate Data Warehouse (CDW) and the VA Central Cancer Registry (CCR). The CDW is a comprehensive automated VA database that is updated in real time and includes all inpatient and outpatient encounters, laboratory, pharmacy, and vital status information on all VA users. The VA CCR was initiated in 1995 and serves as a national data repository for over 750,000 VA patients with cancer.23,24 Cancer registrars at each VA medical center manually abstract data, and data are then aggregated into the national cancer registry where cases are merged, and quality assurance checks are conducted. The CCR includes information on patient demographics, date of cancer diagnosis, primary site, histology, grade, tumor size, extension, staging, and treatment.
HIV-Infected Patient Cohort
The HIV-infected cohort included patients aged 18 years or above who fulfilled 2 of the 3 following criteria: (1) laboratory criteria included patients with at least one positive HIV antibody test by Eliza or Western Blot; tested for HIV viral load (any ±/indeterminate); or tested for CD4+ count; (2) treatment criteria included patients with at least one prescription in inpatient or outpatient pharmacy records for HIV ART; and (3) diagnosis criteria included any inpatient or outpatient encounter with an International Classification of Diseases, Ninth Revision (ICD-9) (042 or V08) or ICD-10 code (B20 & Z21) for HIV. This 2 of the 3 criteria definition was highly accurate for identify patients when compared with a gold-standard VA HIV Clinical Case Registry (positive predictive value, 93.5%; sensitivity, 95.2%). We used the earliest date of the HIV diagnostic criteria as the index date of follow-up unless it was before October 1, 1999, (inception date of majority of the CDW data) in which case we assigned October 1, 1999, as the HIV-infected patient's index date.
HIV-Uninfected Patient Cohort
We identified over 10 million unique patients listed in the VA databases during the study period. Potential HIV-uninfected controls had to be at least 18 years old and have at least 2 outpatient visits between October 1, 1999, and December 31, 2016. We excluded patients from serving as a potential control if they had one or more criteria suggestive of HIV or were missing their date of birth or sex in CDW. HIV-uninfected controls were matched to HIV-infected patients based on age at index date (within 2 year), sex, and occurrence of an outpatient visit within 1 month of the HIV-infected patient's index date. We then used random selection with replacement to select 4 matches for each HIV-infected patient from the entire matched HIV-uninfected control cohort. Nearly 92% of the controls were selected only once. Less than 1% of the controls were selected more than twice.
During 12,168 person-years (PY) of follow-up, only 1 HIV-infected women developed esophageal cancer [esophageal squamous cell carcinoma (ESCC)], yielding an incidence rate of 8.22 per 100,000 PY; none developed gastric cancer. No women in the HIV-uninfected control cohort developed esophageal or gastric cancer. Because of the small number of outcomes, we excluded female patients from the Cox regression analyses.
We used a hierarchical approach to define the occurrence of esophageal or stomach cancer in the HIV-infected and HIV-uninfected cohorts. First, we identified all esophageal and stomach cancer cases in the VA CCR based on primary site codes C15 and C16 with histology codes 817XX through 818XX as well as text searches. We then identified patients with any ICD-9 (150.X and 151.X) or ICD-10 code (C15 and C16) for esophageal and stomach cancer from the inpatient and outpatient files of CDW. We examined the discordance of cancer diagnosis in the 2 groups of patients classified as esophageal or stomach cancer based on VA CCR and ICD-9/ICD-10 to identify patients who had an ICD-9 or ICD-10 code but was not identified as esophageal or gastric cancer in the VA CCR data. We conducted a manual review of the electronic medical record for each discordant patient to determine their true cancer status. This hierarchical approach ensured high validity of all the captured esophageal and stomach cancer cases.25 All-cause mortality was identified from the VA Vital Status file. The Vital Status file combines information from Medicare, VA, Social Security, and VA compensation and pension benefits to determine date of death (sensitivity 98.3%; specificity 99.8% relative to National Death Index).26
We extracted demographic data and risk factor information using ICD codes and pharmacy and laboratory data from the CDW, including age at index date, sex, race/ethnicity (White, Black, and other/unknown), history of alcohol abuse (categorized as: yes, no), smoking (ever, never), gastroesophageal reflux disease (GERD; ever, never), and Helicobacter pylori infection (ever, never) (see Table 1, Supplemental Digital Content, http://links.lww.com/QAI/B307 for ICD codes). We defined baseline body mass index (BMI) by using the closest height and weight values in the 2 years before or after index date. For the HIV-infected cohort only, other HIV-specific variables included were CD4 count (cells/μL) and percent of time with undetectable HIV viral load, both as time-updated variables throughout the study period. Calendar year of HIV diagnosis was defined as the earliest HIV diagnosis date in the VA from the VA HIV Clinical Case Registry or VA CDW whichever was earliest.
Cancer Risk in HIV-Infected Patients vs. HIV-Uninfected Controls
We compared baseline characteristics between HIV-infected patients and HIV-uninfected matched controls using Student's t-test for continuous variables and χ2 tests for categorical variables. Because our objective was to examine the risk of incident cancer, we excluded patients with evidence of esophageal or stomach cancer before or within 90 days after their index date. Therefore, follow-up time at risk was calculated from 90 days after their index date (thus, the possible earliest start date of follow-up was January 1, 2000) to the development of esophageal or stomach cancer, death, or December 31, 2016, whichever was earlier. We calculated incidence rates for esophageal and stomach cancer in the HIV-infected and HIV-uninfected cohorts per 100,000 PY overall and in several prespecified subgroups of patients including subgroups defined by histology and anatomical site. Furthermore, we compared incidence rates among HIV-uninfected controls separately with (1) HIV-infected patients with CD4 count ≤200 and (2) HIV-infected patients with CD4 count >200; and separately with (1) HIV-infected patients with percent time with an undetectable viral load <80% and (2) HIV-infected patients with percent time with an undetectable viral load ≥80%. We then estimated adjusted hazard ratios (aHRs) and 95% confidence intervals (CIs) for the association between HIV infection and the risks of esophageal adenocarcinoma (EAC), ESCC, gastric cardia cancer (GCC), and gastric noncardia cancer (GNCC) using Cox proportional hazards regression models.
Risk Factors for Cancer in HIV-Infected Individuals
Using only the HIV-infected cohort, we examined for potential risk factors for developing EAC, ESCC, GCC, and GNCC using Cox proportional hazards regression models. Potential predictors included age, race/ethnicity, alcohol, smoking, BMI, GERD, H. pylori infection, year of HIV diagnosis, nadir CD4 count, and percent of time with undetectable HIV viral load. Variables with P <0.2 in univariate analysis were included in the multivariable model.
Analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC). Statistical significance was determined at α = 0.05, and P values for statistical significance were 2-sided.
We included data from 44,075 HIV-infected patients and 157,705 HIV-uninfected matched controls in the primary analysis. The characteristics of HIV-infected patients and HIV-uninfected controls are shown in Table 1. Because of matching, there was no difference in mean age between the 2 patient groups. The HIV-infected patients were more likely than the HIV-uninfected controls to be Black (52.3% vs. 19.7%), have a history of alcohol abuse (29.2% vs. 17.5%), and have a history of smoking (80.5% vs. 77.3%). By contrast, the HIV-infected patients had, on average, lower baseline BMI than HIV-uninfected controls (25.5 vs. 29.5 kg/m2) and were less likely to have GERD (23.9% vs. 30.4%). Most HIV-infected patients had a calendar year of HIV diagnosis in the early ART (43.2% pre-2001) or recent ART (35.5% 2006–2016) eras. Approximately half of the HIV-infected patients had a nadir CD4 count ≤200 at some point during follow-up, but nearly half of the HIV-infected patients spent more than 80% of their follow-up time with an undetectable viral load.
Cancer Risk in HIV-Infected Patients vs. HIV-Uninfected Matched Controls
The mean duration of follow-up for the HIV-infected patients following HIV index date from the CDW was 8.96 years (SD, 5.66) and 9.96 years (SD, 5.73) for the HIV-uninfected controls from their corresponding index date. Among the HIV-infected patients, 23 developed incident EAC for an incidence rate of 5.83 per 100,000 PY; the rate was 12.3 per 100,000 PY among the HIV-uninfected control cohort (n = 194 incident EAC; Table 2). Conversely, the rate of ESCC among HIV-infected patients (n = 36; incidence rate, 9.12 per 100,000 PY) was higher than that in the HIV-uninfected controls (n = 66; incidence rate, 4.20 per 100,000 PY). Incidence rates for esophageal carcinomas were elevated among HIV-infected patients compared with HIV-uninfected controls for carcinoma of the middle esophagus but reduced for carcinoma of the lower esophagus (Table 2). For stomach carcinoma, there was a higher rate of GNCC among HIV-infected patients (n = 20; incidence rate, 5.07 per 100,000 PY) than HIV-uninfected controls (n = 48; incidence rate, 3.05 per 100,000 PY), but a lower rate of GCC among HIV-infected patients (n = 6; incidence rate, 1.52 per 100,000 PY) vs. HIV-uninfected controls (n = 41; incidence rate, 2.61 per 100,000 PY). The incidence rates among HIV-infected patients varied by CD4 count (see Table 2, Supplemental Digital Content, http://links.lww.com/QAI/B307) and percent time with an undetectable viral load (see Table 3, Supplemental Digital Content, http://links.lww.com/QAI/B307).
In unadjusted Cox models, HIV infection was statistically significantly associated with elevated risk for ESCC [unadjusted hazard ratio (HR), 2.21; 95% CI: 1.47 to 3.13] but associated with statistically significantly lower risk for EAC (unadjusted HR, 0.48; 95% CI: 0.31 to 0.74) (Table 3). After controlling for differences in the distribution of race/ethnicity between HIV-infected patients and HIV-uninfected controls, there remained a statistically significant association between HIV infection and elevated risk for ESCC (age and race aHR, 1.84; 95% CI: 1.20 to 2.84). HIV infection remained associated with risk of ESCC after further adjusting for alcohol abuse and smoking use (HR, 1.58; 95% CI: 1.02 to 2.47), but not after adjusting for BMI (Table 3). The magnitude of the association with HIV was stronger and statistically significant with risk of ESCC when we compared the HIV-uninfected controls with HIV-infected patients with CD4 count ≤200 (w/out BMI HR, 2.20; 95% CI: 1.35 to 3.60) (see Table 4, Supplemental Digital Content, http://links.lww.com/QAI/B307). There remained some evidence for an inverse relationship between HIV infection and the risk of EAC in the fully adjusted model, although the CIs were wide and HRs were no longer statistically significant (fully aHR, 0.85; 95% CI: 0.54 to 1.33).
HIV infection was associated with increased risk for GNCC in the unadjusted model (HR, 1.69; 95% CI: 1.00 to 2.85); however, the associated was attenuated toward the null after controlling for race/ethnicity (HR, 1.12; 95% CI: 0.65 to 1.94) and in the fully adjusted models (w/out BMI HR, 1.06; 95% CI: 0.61 to 1.84; w/BMI HR, 0.84; 95% CI: 0.47 to 1.48). It is unclear whether or not HIV infection is associated with risk of GCC. Although the point estimates are suggestive of an inverse relationship, we had very few cases of GCC in our cohorts resulting in wide CIs.
Risk Factors for Cancer in HIV-Infected Individuals
Along with known risk factors (increasing age, White race, smoking, and GERD), we found that percent time with undetectable viral load was associated with risk of EAC in univariate analysis (Table 4). In the multivariable model, this factor was no longer statistically significant perhaps due to small numbers of EAC. For ESCC, the strongest independent risk factors were increasing age, alcohol use, smoking, and lower BMI. CD4 count ≤200 was associated with 2-fold higher risk of ESCC in the univariate model (HR, 2.15; 95% CI: 1.08 to 4.29). The association was minimally attenuated and the 95% CIs included the null in an adjusted model (aHR, 1.87; 95% CI: 0.93 to 3.74). For GNCC, CD4 count ≤200 was associated with increased risk (aHR, 3.38; 95% CI: 1.21 to 9.48) independent of known risk factors including H. pylori infection (aHR, 7.89; 95% CI: 2.28 to 27.3) and black race (vs. white; aHR, 3.80; 95% CI: 1.09 to 13.3) (Table 5). In sensitivity analyses, the associations with CD4 count were unchanged when we excluded values from the period 12 months preceding cancer/censoring, and when we modeled CD4 count as a fixed exposure taking the value corresponding to the period 12 months preceding cancer/censoring.
In this study, we report higher crude rates of ESCC and GCC among male veterans with HIV infection compared with matched male veterans without HIV infection. Conversely, HIV-infected patients had lower crude rates of EAC and GNCC than matched HIV-uninfected controls. Consistent with characteristics of PLWH seen in other studies, our HIV-infected patients were significantly more likely than matched HIV-uninfected controls to be black, have a history of alcohol abuse, and have ever smoked. After controlling for these differences between our 2 cohorts, HIV infection remained associated with approximately 50% increased risk for ESCC. In addition to established risk factors for esophageal and stomach cancers (ie, GERD for EAC, lower BMI and alcohol abuse for ESCC, and H. pylori infection for GNCC), we found that low CD4 count may be independently associated with increased risks for ESCC and GNCC in HIV-infected patients.
With the aging HIV infection population, NADCs are an increasing public health problem. There is strong evidence that compared with the general population, PLWH are at increased risk for lung cancer, head and neck cancers, anal cancer, hepatocellular carcinoma, and Hodgkin lymphoma.27 The reasons are unclear; however, this may be due to faster cancer progression, distinct etiologies or etiologic heterogeneity, and/or risk factor profiles that are more prone to developing these cancers. Studies of esophageal and stomach cancer among PLWH have been limited to overall cancer risks and a lack of individual-level patient data. Given the known etiologic heterogeneity for subcategories of these cancers, considering them in totality may mask any true relationships with HIV infection. In our study among HIV-infected veterans with comprehensive data on demographics and risk factors, we found HIV infection was associated with 58% increased risk for ESCC independent of other risk factors. Among HIV-infected patients with CD4 count ≤200, risk of ESCC was almost 120% higher than that among controls. This evidence for effect modification by CD4 count has been observed for other cancers, including squamous cancers of the anal canal, cervix, and oropharynx.27 Crude rates of EAC and GCC in our HIV-infected patients were lower than among uninfected patients; however, the strength of these inverse relationships was greatly attenuated in multivariate models.
Use of ART in HIV has been associated with increase in and accumulation of abdominal fat, dyslipidemia, and insulin resistance.17 Recent data show an increase in the prevalence of obesity and weight gain among HIV-infected adults starting ART.14 Intra-abdominal fat is metabolically active, and increased stores may result in hormonal, adipokine, and cytokine changes,28 lead to chronic inflammatory state and metabolic disorders that promote carcinogenesis.29 The interrelationships therefore between HIV infection, obesity, and risk of developing cancers of the esophagus and stomach are highly complex. Compared with the matched HIV-uninfected controls, the HIV-infected patients in our study were significantly less likely to be obese (14.8% vs. 36.4%). When we included BMI in the multivariable model for ESCC, the point estimate (ie, HR) for HIV infection was attenuated considerably toward the null (from 1.58 to 1.07), possibly because low BMI was significantly associated with HIV status. Our results regarding BMI should be interpreted with caution because we only used baseline BMI in our analysis. Other studies have shown that individuals with HIV gain significant amounts of weight after starting ART,30 and because 55% of our cohort was not on ART at the start of follow-up (ie, at 90 days from index date), a large percentage of the cohort could have had changes in BMI over the follow-up that was not captured.
This study has multiple strengths, including the large sample size, extended follow-up period, and individual-level patient data. The Veterans Health Administration is the largest integrated health care system and provider of comprehensive HIV care in the United States. This is the largest study published to date evaluating the association between HIV infection and the risk of esophageal and stomach cancers and the only to examine the specific subtypes of these cancers. We combined the advantages of the comprehensive fully automated nationwide VA clinical, laboratory, and administrative databases (CDW and CCR) with those of direct complementary manual electronic medical records abstractions. We have previously examined the validity of our algorithm for identifying patients with HIV using these automated data, with high positive-predictive value (93.5%) and sensitivity (95.2%), and we used a hierarchical approach that ensured high validity of all the captured esophageal and stomach cancer cases.25
Our study also has several limitations. First, although the HIV algorithm was highly predictive of presence of HIV diagnosis in the administrative databases in our previous study, it is possible that a small proportion of patients with HIV might not have been captured using our definition and therefore was included in the control cohort. It is unclear how this may have affected the findings of our study; however, this misclassification would be minimal given the low prevalence of HIV infection in the general population. Furthermore, some HIV-negative veterans (eg, those with needlestick injuries and biological fluid exposure) may have been included incorrectly as HIV positive. However, when tested, we found that our algorithm has a positive-predictive value for HIV of 93.5%. Thus, the number of false positives, which drives positive predictive value, is likely very low. Second, although many VA patients get their care exclusively at the VA, some patients with dual insurance coverage (eg, Medicare for elderly patients) may get some of their care outside the VA. As a result, the first known HIV diagnosis in the VA system may not accurately reflect the patient's true index diagnosis date should they have been diagnosed with HIV outside the VA. Although we used the fee basis file to capture outside utilization paid for by the VA, and we expect that this percentage is low, this is a limitation of using retrospective clinical data. Third, even in the setting of >40,000 HIV-infected patients, we observed few cases of EAC, ESCC, GCC, and GNCC limiting our power to precisely quantify associations with HIV infection and HIV-related factors. Fourth, we conducted our study among male veterans and approximately 80% were over 40 years old at HIV index date; thus, the generalizability of these data to other populations may be somewhat limited. However, although our results may not be generalizable to women and nonveterans, because esophageal and stomach cancers disproportionately affect older males, our study population is adequate to study risk factors and outcomes of these cancers. Furthermore, the biological processes of disease progression are likely similar in veterans and nonveterans with HIV. Fifth, this was a retrospective cohort study using covariate data extracted from administrative databases by ICD codes, laboratory values, and pharmacy records. Although many of these variables have been validated in other VA studies, some exposures (eg, H. pylori and GERD) may not have been completely captured or may have been misclassified. Nonetheless, we were able to identify associations with these established risk factors for esophageal and stomach cancers in our cohort. Sixth, although we attempted to control for potential confounders, residual confounding by poorly measured or unknown/unmeasured factors could still have influenced our results. Finally, we also used a relatively new measure of HIV viral load control, “percent of time with undetectable HIV viral loads,” which may not adequately describe the true time that individuals had an undetectable HIV viral load. However, we have previously shown that it is a better measure of overall HIV viral control compared with a single HIV viral load measurement.31 Also, because our data were extracted from clinical care, there are potential variations in the frequency of follow-up visits and thus CD4+ and viral load testing.
In conclusion, in this large retrospective cohort study of male veterans, we found that HIV infection was associated with 58% increased risk for ESCC. The risk for ESCC was 2.2-fold higher among HIV-infected patients with CD4 count ≤200 when compared with matched HIV-uninfected controls. Risk factors for these cancers in HIV-infected patients were otherwise similar to those in a HIV-uninfected population. Although our results do not justify overall population-based screening among HIV-infected patients, endoscopic screening may be considered among those at highest risk (eg, those with CD4 count ≤200 and who have a history of alcohol abuse and heavy cigarette smoking). In general, additional efforts to implement tobacco cessation and moderation of alcohol use may have an effect in reducing the risk of ESCC in HIV-infected patients. Further prospective studies in the VA with longer follow-up for large number of these rare cancers, and in the setting of global populations (ie, outside of the United States) with high prevalence of HIV and high incidence rates for esophageal and stomach cancers are needed to better understand the relationship with HIV infection. These studies also need to include women with HIV infection and address the complex interrelationships between HIV infection, low BMI, and risk of ESCC.
The opinions stated here are those of the authors and do not necessarily represent the views of the National Institutes of Health or the US Department of Veterans Affairs.
1. Palella FJ Jr, Delaney KM, Moorman AC, et al. Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection. HIV Outpatient Study Investigators. N Engl J Med. 1998;338:853–860.
2. CDC. HIV in the United States: At A Glance; 2017. Available at https://www.cdc.gov/hiv/statistics/overview/ataglance.html
. Accessed October 1, 2018
3. Shiels MS, Pfeiffer RM, Gail MH, et al. Cancer burden in the HIV-infected population in the United States. J Natl Cancer Inst. 2011;103:753–762.
4. Robbins HA, Shiels MS, Pfeiffer RM, et al. Epidemiologic contributions to recent cancer trends among HIV-infected people in the United States. AIDS. 2014;28:881–890.
5. Morlat P, Roussillon C, Henard S, et al. Causes of death among HIV-infected patients in France in 2010 (national survey): trends since 2000. AIDS. 2014;28:1181–1191.
6. Shiels MS, Cole SR, Kirk GD, et al. A meta-analysis of the incidence of non-AIDS cancers in HIV-infected individuals. J Acquir Immune Defic Syndr. 2009;52:611–622.
7. Goehringer F, Bonnet F, Salmon D, et al. Causes of death in HIV-infected individuals with immunovirologic success in a national prospective survey. AIDS Res Hum Retroviruses. 2017;33:187–193.
8. Simard EP, Pfeiffer RM, Engels EA. Mortality due to cancer among people with AIDS: a novel approach using registry-linkage data and population attributable risk methods. AIDS. 2012;26:1311–1318.
9. Simard EP, Engels EA. Cancer as a cause of death among people with AIDS in the United States. Clin Infect Dis. 2010;51:957–962.
10. Rohner E, Wyss N, Heg Z, et al. HIV and human herpesvirus 8 co-infection across the globe: systematic review and meta-analysis. Int J Cancer. 2016;138:45–54.
11. Koziel MJ, Peters MG. Viral hepatitis in HIV infection. N Engl J Med. 2007;356:1445–1454.
12. Rowhani-Rahbar A, Hawes SE, Sow PS, et al. The impact of HIV status and type on the clearance of human papillomavirus infection among Senegalese women. J Infect Dis. 2007;196:887–894.
13. Ucciferri C, Tamburro M, Falasca K, et al. Prevalence of anal, oral, penile and urethral human papillomavirus in HIV infected and HIV uninfected men who have sex with men. J Med Virol. 2018;90:358–366.
14. Koethe JR, Jenkins CA, Lau B, et al. Rising obesity prevalence and weight gain among adults starting antiretroviral therapy in the United States and Canada. AIDS Res Hum Retroviruses. 2015;32:50–58.
15. Galvan FH, Bing EG, Fleishman JA, et al. The prevalence of alcohol consumption and heavy drinking among people with HIV in the United States: results from the HIV Cost and Services Utilization Study. J Stud Alcohol. 2002;63:179–186.
16. Mdodo R, Frazier EL, Dube SR, et al. Cigarette smoking prevalence among adults with HIV compared with the general adult population in the United States: cross-sectional surveys. Ann Intern Med. 2015;162:335–344.
17. Grinspoon S, Carr A. Cardiovascular risk and body-fat abnormalities in HIV-infected adults. N Engl J Med. 2005;352:48–62.
18. Grulich AE, van Leeuwen MT, Falster MO, et al. Incidence of cancers in people with HIV/AIDS compared with immunosuppressed transplant recipients: a meta-analysis. Lancet. 2007;370:59–67.
19. Patel P, Hanson DL, Sullivan PS, et al. Incidence of types of cancer among HIV-infected persons compared with the general population in the United States, 1992–2003. Ann Intern Med. 2008;148:728–736.
20. Gallagher B, Wang Z, Schymura MJ, et al. Cancer incidence in New York State acquired immunodeficiency syndrome patients. Am J Epidemiol. 2001;154:544–556.
21. Engels EA, Pfeiffer RM, Goedert JJ, et al. Trends in cancer risk among people with AIDS in the United States 1980–2002. AIDS. 2006;20:1645–1654.
22. Frisch M, Biggar RJ, Engels EA, et al. Association of cancer with AIDS-related immunosuppression in adults. JAMA. 2001;285:1736–1745.
23. Zullig LL, Jackson GL, Dorn RA, et al. Cancer incidence among patients of the U.S. veterans affairs health care system. Mil Med. 2012;177:693–701.
24. Jackson GL, Melton LD, Abbott DH, et al. Quality of nonmetastatic colorectal cancer care in the department of veterans affairs. J Clin Oncol. 2010;28:3176–3181.
25. Kanwal F, Kramer JR, Mapakshi S, et al. Risk of hepatocellular cancer in patients with non-alcoholic fatty liver disease. Gastroenterology. 2018;155:1828–1837.
26. Sohn MW, Arnold N, Maynard C, et al. Accuracy and completeness of mortality data in the department of veterans Affairs. Popul Health Metr. 2006;4:2.
27. Thrift AP, Chiao EY. Are non-HIV malignancies increased in the HIV-infected population? Curr Infect Dis Rep. 2018;20:22.
28. Yang X, Smith U. Adipose tissue distribution and risk of metabolic disease: does thiazolidinedione-induced adipose tissue redistribution provide a clue to the answer? Diabetologia. 2007;50:1127–1139.
29. Grivennikov SI, Greten FR, Karin M. Immunity, inflammation, and cancer. Cell. 2010;140:883–899.
30. Hasse B, Iff M, Ledergerber B, et al. Obesity trends and body mass index changes after starting antiretroviral treatment: the Swiss HIV cohort study. Open Forum Infect Dis. 2014;1:ofu040.
31. Kowalkowski MA, Day RS, Du XL, et al. Cumulative HIV viremia and non-AIDS-defining malignancies among a sample of HIV-infected male veterans. J Acquir Immune Defic Syndr. 2014;67:204–211.