McQuillan, Geraldine M PhD*; Kruszon-Moran, Deanna MS*; Granade, Timothy MS†; Feldman, Jane W BS‡
The Centers for Disease Control and Prevention (CDC) has conducted AIDS surveillance since the epidemic was first recognized in 1981.1 Based on national HIV/AIDS surveillance data, CDC has estimated recently that the number of annual incident HIV infections has remained stable2 but the estimated number of persons now living with HIV (HIV prevalence) has increased.1 Surveillance data also continue to demonstrate that non-Hispanic blacks and for all races, men who have sex with men are disproportionately affected by HIV infection.1
To monitor national HIV seroprevalence, the National Health and Nutrition Examination Surveys (NHANES) have included HIV antibody testing since NHANES III (conducted 1988-1994).3 Because each NHANES survey cycle consists of a nationally representative sample of the civilian noninstitutionalized US population, and not a sample selected for HIV risk behaviors, measurement of HIV seroprevalence in each survey cycle provides an understanding of the background level of infection in the general population. Data from the current NHANES surveys, conducted in 1999-2002 and 2003-2006, provide an opportunity to update the HIV seroprevalence estimate in the general household population and to observe trends in HIV seroprevalence between the previous NHANES III survey and the two 4-year cycles of the current survey. Self-reported data on medication use in the survey provide insight into antiretroviral treatment of HIV-infected individuals in this population. Data on self-reported risk behaviors provide information on the contribution of these behaviors to HIV infection in the general population.
Study Populations and Sample Design
The NHANES are a series of cross-sectional surveys designed to provide national statistics on the health and nutritional status of the general household population of the United States, through household interviews, standardized physical examinations, and the collection of biologic samples in special mobile examination centers (MECs). In 1999, NHANES became a continuous survey with data released every 2 years. The sampling plan for the survey was a stratified, multistage, probability cluster design that provides a sample of the 50 states that is representative of the general household population of the US.4 African Americans, Mexican Americans, adolescents, and low-income persons were sampled at higher frequencies than other persons to provide more precise estimates for these groups. Informed consent was obtained from all participants, and the National Center for Health Statistics, CDC Ethical Review Board approved the protocol.
In NHANES 1999-2006, serum samples were available for testing for HIV antibodies from persons aged 18-49 years. Race and ethnicity were categorized, based on the subjects' self-reported information, as non-Hispanic white, non-Hispanic black, and Mexican American. Subjects who were not classified into one of these categories were classified as “other” (this included all non-Mexican American Hispanics). Because there were only 10 non-Hispanic whites, 7 Mexican Americans, and 4 sample persons in the “other” race/ethnic group who tested HIV reactive during this period, sample sizes were too small to describe associations with potential risk factors for these groups individually. Overall serpositivity is presented for all these groups combined. Detailed analyses are presented only for the total population and non-Hispanic blacks. Other variables analyzed with the NHANES 1999-2006 HIV test results include the following: poverty index [calculated by dividing total family income by the poverty threshold index adjusted for family size at year of the interview and categorized as either below poverty (<1) or at or above poverty (1 or above)]5; education (measured as last year of school completed and grouped into 2 levels: less than high-school graduate compared with high school or more completed); and marital status (grouped into 3 groups: married, widowed, or living as married; divorced or separated; and never married). Sexual behavior and drug-use data were collected in a private room in the MEC using an audio computer-assisted self-interviewing technique. The data collected included use of cocaine and intravenous drugs, age of first intercourse, lifetime number of sexual partners, and history of male-to-male sex. The prescription drug-use data needed to determine antiretroviral use were collected during the household interview. Respondents were asked: “Have you taken or used any medicines for which a doctor's or dentist's prescription is needed, in the past month?” For each medication reported, the interviewer asked to see the medication container to record the product names from the label. If the container was unavailable, the interviewer probed the subject for this information.6
In NHANES III, conducted from 1988 to 1994, 11,203 participants aged 18-59 years were anonymously tested for antibody to HIV. Because testing was conducted anonymously, age was collected only in 2 age groups, 18-39 years and 40-59 years. Therefore, the 18- to 39-year age group is the only age group directly comparable between the current survey and NHANES III.
In NHANES, 1999-2006 serum specimens were tested for the presence of antibodies to both HIV-1 and HIV-2 (Biorad Genetic Systems HIV-1/HIV-2 Peptide EIA, Redmond, WA). In 2004, the assay was changed to the Biorad Genetic Systems HIV-1/HIV-2 plus O EIA. All specimens repeatedly reactive for HIV antibody were confirmed by a western blot [Cambridge Biotech HIV-1 Western blot, Calypte Biomedical Corporation, Rockville, MD (now Maxium Biomedical Products, Rockville, MD)]. NHANES participants who consented to the HIV test but did not have venipuncture or did not have sufficient serum available for the serum antibody test were tested for the presence of HIV antibodies in a urine specimen using a testing algorithm similar to that used for the serum specimens (Calypte HIV-1 Peptide EIA, Calypte Biomedical Corporation, Pleasanton, CA). Specimens reactive for HIV antibodies by the enzyme immunoassay screen were confirmed by HIV western blot (Cambridge Biotech HIV-1 Urine Western blot, Calypte Biomedical Corporation).7 Serum samples were evaluated for the presence of antibody to herpes simplex type 2 (HSV-2) using a type-specific immunodot assay previously described.8 CD4 T-lymphocyte enumeration was performed on HIV positives and age-matched controls using a method by Fiebig et al that has been previously described.9
In NHANES III, serum samples were tested for the presence of antibodies to HIV by an enzyme immunoassay method (Genetic Systems Redmond, WA or Organon-Teknika, Durham, NC). Repeatedly positive specimens were confirmed by western blot (Biotech/Dupont, Rockville, MD).3
HIV seroprevalence estimates were weighted to represent the total civilian noninstitutionalized US household population to account for oversampling and nonresponse to the household interview and physical examination, but not for nonresponse to phlebotomy. The weights were further ratio-adjusted by age, gender, and race/ethnicity to the US population control estimates from the Current Population Survey which was adjusted for undercounts.10 Standard errors were calculated using SUDAAN (Research Triangle Institute, Research Triangle, NC),11 a family of statistical procedures for analysis of data from complex sample surveys. Ninety-five percent confidence intervals (CIs) were estimated by using the logit transformation.12 The exact binary method was used to estimate the 95% CIs for antiretroviral and low CD4 counts due to the small sample size and few degrees of freedom in this subgroup of HIV-reactive participants. To examine possible predictors of HIV seropositivity, differences in prevalence and tests for trends were evaluated by examining P values calculated using a univariate t statistic obtained from a general linear contrast procedure in SUDAAN. Estimates are indicated as unstable when the standard error of the estimate relative to the estimate itself (RSE) was greater than 30%.13 P values from tests of significance were not corrected for multiple comparisons.
A backward stepwise logistic modeling procedure in SUDAAN was used to examine which variables were independently associated with HIV seroprevalence. All variables with significant associations with increased HIV seropositivity from the univariate analysis were entered into the model. Variables with a Satterthwaite-adjusted F statistic P < 0.05 were considered to be significant predictors. The final model included race/ethnicity and age group, along with all other variables significantly associated with increased HIV seroprevalence. Individual models were created to determine which variables were significantly associated with HIV seroprevalence among non-Hispanic blacks alone. The total male population and non-Hispanic black male population were also modeled separately from females to include male-to-male sex in the model.
Of the 14,872 participants aged 18-49 years selected for the survey from 1999 to 2006, 12,917 (86.9%) were interviewed, 12,310 (95.3% of those interviewed) were examined, and 11,928 (96.9% of those examined) were tested for HIV antibody. The lowest response rates for HIV testing were among those who had less than a high-school education (95.3%), those born outside the United States (95.9%), those living below poverty (96.4%), and those never married (96.4%).
The overall HIV seroprevalence for 1999-2006 for those aged 18-49 years was 0.5% (95% CI 0.3-0.6) with a US population estimate of 615,000 (95% CI 440,000-840,000) individuals living with HIV infection. HIV seroprevalence remained stable between the first 4 years and the latter 4 years of the survey (P = 0.57; Table 1).
In NHANES 1999-2006 (Table 1), seroprevalence was significantly higher among those aged 40-49 years [0.6% (95% CI 0.4-1.0)] compared with those aged 18-29 years [0.3% (95% CI 0.1-0.5)], and this age trend was significant for the total population (P = 0.03) and among non-Hispanic blacks (P = 0.01). Males were more likely to be seroreactive than females in the total population (P = 0.001) and among non-Hispanic blacks (P = 0.04). Non-Hispanic blacks were significantly more likely to be reactive compared with the other combined race/ethnic group (P < 0.001), with a seroprevalence of 2.0% (95% CI 1.5-2.7) and a population estimate of 320,000 HIV-infected individuals (95% CI 240,000-430,000).
In both 4-year cycles, non-Hispanic blacks were significantly more likely to be reactive for HIV antibody when compared with all the other combined race/ethnic groups (1999-2002, P < 0.001; 2003-2006, P = 0.001), with the highest seroprevalence observed among non-Hispanic black males aged 40-49 [1999-2002, 4.5% (95% CI 2.2-9.0); 2003-2006, 3.2% (95% CI 1.3-8.0)]; all other age-specific prevalence estimates within gender and race/ethnicity for each 4-year cycle are not presented due to small number of HIV positives and unstable estimates. The trend of increasing reactivity with age observed when analyzing all 8 years together was again significant but only for non-Hispanic blacks and only in 1999-2002 (P = 0.02) not in 2003-2006.
The seroprevalence among those aged 18-39 years in NHANES III [the only age group directly comparable; 0.4% (95% CI 0.2-0.7)] was not significantly different from the seroprevalence in the same age group in the 1999-2002 survey [0.4% (95% CI 0.2-0.8)] and the 2003-2006 survey [0.4% (95% CI 0.3-0.7); P = 0.77 for the linear test for trend across the 3 surveys].
Table 2 presents the seroprevalence of HIV for the period 1999-2006 by demographic characteristics and risk behaviors. Several factors were associated with a higher HIV seroprevalence in the total population: (1) having never been married compared with those widowed, married, or living as married (P = 0.03); (2) having greater than 50 lifetime sexual partners compared with 0-9 partners (P = 0.01); (3) reporting male-to-male sexual behavior and (P = 0.002); and (4) having detectable antibodies to HSV-2 (P < 0.001). Among non-Hispanic blacks, those with any of the following (1) less than a high-school education compared with high school or more (among those aged 20-49; P = 0.02), (2) reporting male-to-male sexual behavior (P = 0.005), (3) reporting having ever used cocaine and P = 0.01), and (4) having detectable antibodies to HSV-2 (P < 0.001) were all significantly associated with a higher HIV seroprevalence.
A logistic regression analysis was performed to assess independent predictors for HIV prevalence in the total population and, among non-Hispanic blacks and among males to include the measurement of risk for men who have sex with men. After adjusting for all significant variables and for age, the presence of HSV-2 antibody was associated with the greatest relative odds for HIV infection in the total population (P < 0.001; Table 3). Additional significant risk factors for infection for this group in order of their relative magnitude of effect are male gender (P < 0.001), having never been married (P < 0.001), and having fifty plus lifetime sexual partners (P = 0.02). Non-Hispanic black race/ethnicity remained in the model but had a much weaker association with HIV seropositivity [OR 2.6 (95% CI 1.1-5.9); P = 0.03] after adjustment for these cofactors when compared with its association with age adjustment alone [OR 8.3 (95% CI 4.3-15.9); P < 0.001]. Among non-Hispanic blacks, antibody to HSV-2 (P < 0.001) and male gender (P = 0.02) remained significant predictors of HIV seroreactivity, but marital status and the number of lifetime sexual partners were no longer associated with increased risk of HIV infection.
Male-to-male sex had the strongest association with HIV infection among the variables in the model within the total male population (P < 0.001; Table 4) followed by the presence of HSV-2 antibody (P < 0.001). Non-Hispanic black race/ethnicity (P = 0.01) and having never been married (P = 0.01) had smaller but still statistically significant associations with HIV infection. Among non-Hispanic black males, male-to-male sex (P < 0.001) and HSV-2 antibody (P = 0.01) had similar relative odds. Divorced or separated (P = 0.003) marital status was the only other variable significantly associated with seroprevalence in the model for this race/ethnic gender subgroup. Because education was not relevant to those aged less than 20 years, it was not included in the models presented. In models of those aged 20-49, less than a high-school education was a significant predictor of HIV seropositivity among both non-Hispanic blacks [OR 3.0 (95% CI 1.5-6.1); P = 0.003] and non-Hispanic black men (OR 4.9 [95% CI (1.4-17.1); P = 0.01].
Reporting of use of antiretroviral therapy (ART) in the last month among HIV-reactive non-Hispanic blacks [16.8% (95% CI 5.5-35.4)] was significantly lower than ART use among HIV-infected individuals in the all other combined race/ethnic group [67.9% (95% CI 47.8-84.0); P < 0.001]. The percent of HIV-infected individuals with CD4 T-lymphocyte counts less that 200 cells/mm3 was similar among non-Hispanic blacks [27.2% (95% CI 15.4-42.0); P = 0.77] and the other combined race/ethnic group [23.5% (95% CI 7.1-49.0)]. Estimates for ART and CD4 counts are based on small sample sizes and few degrees of freedom and should be interpreted with caution.
Data from NHANES 1999-2006 provide an update on HIV seroprevalence in the US household population aged 18-49 years. These data demonstrated no significant trend over time in HIV infection among the 18- to 39-year age group when comparing NHANES III (1988-1994) and each cycle of the current survey NHANES 1999-2002 and NHANES 2003-2006.3,9
In both current 4-year cycles and in NHANES III, non-Hispanic blacks were significantly more likely to be HIV reactive than any of the other race/ethnic groups.3 This higher percentage of seroprevalence among non-Hispanic blacks is consistent with other national surveillance data1,14 and with recent estimates of HIV incidence.2 In the NHANES population, the prevalence estimate for non-Hispanic black males was 6 times higher compared with non-Hispanic white men [0.4% (95% CI 0.2-0.9, data not previously shown)]. This prevalence ratio is similar to the ratio found in the prevalence estimate calculated by extended back calculation.14
Of the self-reported risk behaviors, male-to-male sexual behavior had the strongest association with HIV seroprevalence. Concentration of HIV in this risk group is also consistent with recent estimates.2,14
The presence of HSV-2 antibody was associated with an increase in HIV seroprevalence in all race/ethnic groups, which is consistent with previous reports.15 In all cases, this association remained statistically significant after controlling for age, race/ethnicity, gender, marital status, and the number of lifetime sexual partners, and was the strongest predictor of HIV infection in both the total population and in non-Hispanic blacks.
Among non-Hispanic blacks, a history of ever using cocaine or street drugs was also associated with HIV infection in the univariate analysis. However, this was not true for injection of drugs. Other studies have demonstrated that the use of crack cocaine was an independent risk factor for HIV infection, but these studies did not have laboratory data linked to questionnaire data.16-18 In the NHANES population, this variable did not remain in the model after controlling for demographic variables and the presence of HSV-2 antibody.
In our study, socioeconomic variables had little impact on HIV seroprevalence with the exception of an increased prevalence among those with less than a high-school education among non-Hispanic blacks and non-Hispanic black males.
Results from the multivariate analyses demonstrated that risk factors such as male-to-male sex (among men) and the presence of HSV-2 antibody had the strongest association with HIV seropositivity. In contrast, the relative odds of infection associated with non-Hispanic black race were greatly reduced after controlling for these and other demographic and risk variables. Among non-Hispanic blacks, male sex, and HSV-2 positivity were the only variables that remained in the model.
Disparities in ART use for HIV treatment were also reported among HIV-reactive participants. Non-Hispanic blacks were significantly less likely to report ART compared with the all other combined race/ethnic group. State-based surveillance data have reported race/ethnic disparities in the reduction of HIV mortality along with disparities in the use of HIV therapy.17 Despite this disparity, the percent of infected individuals with CD4 counts less than 200 cell/mm3 was similar among non-Hispanic blacks and all the other combined race/ethnic group.
There are several potential limitations in this study. Only the US household population may participate in NHANES. Prisoners, the homeless, and other institutional populations who may be at greater risk for HIV infection are not included in the sampling. The lack of coverage of these high-risk populations could explain the difference in the NHANES estimate of 615,000 (95% CI 440,000-840,000) infected individuals compared with the recently published estimate of 826,500 HIV-infected adolescents and adults aged 13-49 years based on an extended back-calculation method.14 In 2006, the Bureau of Justice Statistics estimated that 21,980 State and Federal inmates were HIV positive.19 The infection rate among homeless adults in the United States ranges between 10.5% and 21%.20,21 With an estimate of 175,914 chronically homeless persons in the United States,22 this would add between approximately 20,000 and 40,000 HIV-positive individuals to the population totals. In addition, HIV testing in NHANES did not include adolescents aged 13-17 years. Surveillance data from 2006 estimated 1454 HIV/AIDS cases for individuals aged 13-19 years.23 In 1996, Karon et al23 estimated that adjusting the NHANES population estimate for noncovered populations and nonresponse would produce similar confidence limits as those produced with the back-calculation method. Another limitation of the survey was that despite combining 8 years of data, there were only 67 individuals diagnosed with HIV. This small sample size may have limited the power needed to see differences in survey cycles. In addition, more detailed analysis of demographic characteristics and associated risk behaviors could not be conducted especially among nonblack participants.
Despite these limitations, data from NHANES do provide estimates for the general household population. Because NHANES participants are selected irrespective of health status and risk, the calculated estimates are not affected by an individual's knowledge of their HIV status or test-seeking behavior. These data demonstrate that racial/ethnic disparities persist in both HIV seroprevalence and treatment. HIV infection in the household population is significantly higher among the non-Hispanic black population and among men who have sex with men. HIV testing will remain part of the laboratory protocol in future NHANES surveys with the age range in NHANES 2009-2010 expanded to include those aged 50-59 years to continue to monitor HIV seroprevalence in the general US population.
1. Centers for Disease Control and Prevention. HIV/AIDS Surveillance Report
. Vol 18. Atlanta, GA: Department of Health and Human Services, 2006.
2. Hall HI, Song R, Rhodes P, et al. Estimation of HIV incidence in the United States. JAMA
3. McQuillan GM, Khare M, Karon JM, et al. Update on the seroepidemiology of human immunodeficiency virus in the United States household population: NHANES III, 1988-1994. J Acquir Immune Defic Syndr
5. Bureau of the Census. Poverty in the United States: 1990. In: Bureau of the Census, ed. Current Population Reports, Series P-60
. Washington, DC: US Government Printing Office; 1991;Series P60-175:194-196.
7. Berrios DC, Avins Al, Haynes-Sanstad K, et al. Screening for human immunodeficiency virus antibody in urine. Arch Pathol Lab Med
8. Lee FK, Coleman M, Pereira L, et al. Detection of herpes simplex virus type-2-specific antibody with glycoprotein G. J Clin Microbiol
9. McQuillan GM, Kruszon-Moran D, Kotriri BJ, et al. Prevalence of HIV in the US household population: the National Health and Nutrition Examination Surveys, 1988-2002. J Acquir Immune Defic Syndr
10. Mohadjer L, Montaquila J, Waksberg J. National Health and Nutrition Examination Survey III: Weighting and Examination Methodology. Prepared for National Center for Health Statistics
. Hyattsville, MD: Westat Inc; 1996.
11. Shah BV, Barnwell BG, Bieler GS, et al. SUDAAN Users Manual Release 7.0
. Research Triangle Park, NC: Research Triangle Institute; 1996.
12. Wolter K. Introduction to Variance Estimation
. New York, NY: Springer-Verlag Inc; 1985.
14. Center for Disease Control and Prevention. HIV prevalence estimates-United States 2006. MMWR Morb Mortal Wkly Rep
15. Van de Perre P, Segondy M, Foulongne, et al. Herpes simplex virus and HIV-1: deciphering viral synergy. Lancet Infect Dis
16. McCoy CB, Shehghan L, Metsch LR, et al. Injection drug use and crack cocaine smoking: independent and dual risk behaviors for HIV infection. Ann Epidemiol
17. Morin SF, Sengupta S, Cozen M, et al. Responding to racial and ethnic disparities in the use of HIV drugs: analysis of state policies. Public Health Rep
18. Cooke JA, Burke-Miller JK, Cohen MH, et al. Crack cocaine, disease progression, and mortality in a multicenter cohort of HIV-1 positive women. AIDS
20. Reback CJ, Kamien JB, Amass L. Characteristics and HIV risk behaviors of homeless substance-using men who have sex with men. Addict Behav
21. Robertson MJ, Clark RA, Charlebois ED, et al. HIV seroprevalence among homeless and marginally housed adults in San Francisco. Am J Public Health
22. United States Department of Housing and Urban Development. HUD Reports Drop in the Number of Chronically Homeless Persons
. Available at: http://www.hud.gov/news/release.cfm
?content=pr08-113.cfm. Accessed June 2, 2009. News Release No. 08-113.
23. Centers for Disease Control and Prevention. HIV/AIDS Surveillance Report 2007
. Vol 19. Atlanta, GA: US Department of Health and Human Services, Centers for Disease Control, and Prevention; 2009:1-63. Also available at: http://www.cdc.gov/hiv/topics/surveillance/resources/reports/
. Accessed June 4, 2009.
© 2010 Lippincott Williams & Wilkins, Inc.