Paper
HIV Incidence, Prevalence, and Undiagnosed Infections in U.S. Men Who Have Sex With Men
Published May 15, 2018 · Sonia Singh, R. Song, A. S. Johnson
Annals of Internal Medicine
136
Citations
8
Influential Citations
Abstract
Infection with HIV, as well as its burden, is a major health issue for the United States. Each year, about 40000 persons receive an HIV diagnosis, and more than 1 million are living with the virus (1, 2). Testing for HIV, linkage to and engagement in care, and receipt of antiretroviral therapy are critical to prevent disease progression. Persons who are aware of their HIV infection also may reduce risk behavior and, with successful treatment, achieve viral suppression, which greatly reduces the risk for transmitting the virus (35). Some successes of treatment and prevention efforts are reflected in recent decreases in HIV diagnoses among women and persons who inject drugs (2). However, progress in reducing infections among men who have sex with men (MSM) has been limited. National goals include increasing the percentage of persons who are aware of their HIV infection and reducing the number of new infections. To meet these objectives, targets must be met for all high-risk populations in the United States, particularly gay, bisexual, and other MSM of all racial/ethnic groups (6). Although MSM represent approximately 2% of the U.S. population, they made up 67% of persons who received an HIV diagnosis in 2015 (2, 7). Assessing HIV outcomes nationally as well as among MSM is important for guiding prevention efforts and monitoring progress toward national goals. We updated and extended our model estimating HIV incidence and prevalence and the percentage of undiagnosed HIV infections (8) for additional years2008 to 2015for sex and transmission categories, and for MSM by race/ethnicity and age. Methods Using data from the National HIV Surveillance System (NHSS) of the Centers for Disease Control and Prevention (CDC) on the first CD4 value after HIV diagnosis, we modeled HIV incidence and prevalence and the percentage of undiagnosed HIV infections from 2008 through 2015 for persons aged 13 years and older (812). We used information on persons living with diagnosed HIV infection by the end of 2007 to estimate HIV prevalence and the percentage of undiagnosed HIV infections. All states, the District of Columbia, and 6 U.S. territories report cases of HIV infection and associated demographic and clinical information to the CDC without personal identifiers. Although cases of stage 3 HIV infectionthat is, AIDShave been reportable in all jurisdictions since the early 1980s, implementation of confidential, name-based HIV reporting was staggered over time, with all jurisdictions reporting by 2008. The first CD4 test result after an HIV diagnosis is a required data element on the HIV case report form. In addition, as of December 2016, all but 6 states had implemented mandatory laboratory reporting of all CD4 values, and 37 states and the District of Columbia reported complete data to the NHSS, representing 72% of persons living with an HIV diagnosis in the United States (1). Cases are routinely deduplicated within and between jurisdictions (13). Deaths are ascertained by linking HIV surveillance data to vital records. Death ascertainment for a given year of death typically is completed within 12 to 18 months; therefore, we used data that allowed at least 18 months for reporting each diagnosed case. Using the first CD4 value after HIV diagnosis, we first estimated the time from HIV infection to diagnosis at the individual level on the basis of a well-characterized CD4 depletion model commonly used in the literature (9, 10): where t is the duration of infection at the date of the CD4 test and ai and bi are model parameters specific to the U.S. HIV population groups (sex, age, and transmission category) in which most HIV infections are subtype B (8). The t estimates among persons in an HIV population group may be used to estimate the distribution of diagnosis delay (time from HIV infection to diagnosis) in the group. These estimates in turn may be used to approximate HIV incidence and prevalence and the percentage of undiagnosed HIV infections (812). Because not everyone with an HIV diagnosis had a CD4 test administered or reported, persons with a CD4 test result were assigned a weight to account for those without a result. This weight is the inverse of the probability that a person with an HIV diagnosis has a CD4 test after diagnosis. This probability is estimated by using the proportion of cases with a CD4 test result reported to the NHSS. Considering the heterogeneity in the availability of CD4 data, the population is stratified on the basis of the year of HIV diagnosis, sex, race/ethnicity, transmission category, age at diagnosis, and disease status at the end of the study periodthat is, whether the person had HIV infection never classified as AIDS, died without the infection ever having been classified as AIDS, or had infection that progressed to AIDS. Because we report results for persons aged 13 years and older, if a person had an estimated date of infection before age 13, the date of infection was set to the date the person reached 13 years of age. The distribution of diagnosis delay (time from HIV infection to diagnosis) was then used to estimate annual HIV incidence, which represents persons with diagnosed or undiagnosed HIV infection. The prevalence of HIV, which represents the number of persons with diagnosed or undiagnosed HIV infection who were alive at the end of a given year, was estimated by subtracting the cumulative number of deaths from the cumulative number of infections. The number of persons with undiagnosed HIV infection was estimated by subtracting the cumulative number of diagnoses from the cumulative number of infections. The proportion of undiagnosed HIV infections was estimated by dividing the number of undiagnosed infections by the total HIV prevalence for each year. More details are provided in the Appendix. All analyses were carried out by using SAS/STAT, version 9.4 (SAS Institute). Approximately 30% of HIV infection cases are reported to the CDC without an identified risk factor (2). To provide case counts by transmission categorya summary classification of the single risk factor most likely to have been responsible for transmissionmultiple imputation was used to handle missing values (14). Multiple imputation is a statistical approach in which each missing transmission category is replaced with a set of plausible values that represent the uncertainty about the true, but missing, value. Variables in the imputation model included age at HIV diagnosis, race/ethnicity, birth country origin, stage of disease at HIV diagnosis, type of HIV diagnosis facility, year of HIV diagnosis, and delay between HIV diagnosis and reporting of the case. Multiple imputation was performed separately for males and females because each sex had different numbers of transmission categories. Ten imputation values were generated to achieve a 95% relative efficiency based on the proportion of missing data. Several data sets with imputed values were analyzed by using standard statistical procedures, and the results were combined (15, 16). To account for model uncertainty, results were rounded to the nearest hundred for estimates greater than 1000 and the nearest ten for those less than or equal to 1000. We examined trends during 2008 to 2015 by using the estimated annual percentage change (EAPC) and its associated 95% CIs. We used Poisson regression with a log link function to calculate EAPC. The EAPC for diagnosed proportions was calculated by using the logarithm of the estimated prevalence that served as offset in the Poisson regression model (17). We compared incidence estimates for 2014 and 2015 by using a simple Z test. Rates per 100000 population were calculated for estimates of HIV incidence and prevalence (18). Rates for transmission categories were calculated by using published population size estimates as denominators (7, 19, 20). Role of the Funding Source This study used data collected as part of routine public health surveillance and was not funded. Results United States HIV Incidence In the United States, the estimated annual number of HIV infections, or HIV incidence, decreased 14.8%, from 45200 infections in 2008 to 38500 in 2015 (EAPC, 2.6% [95% CI, 3.2% to 2.1%]) (Figure; for all years, 2008 to 2015, see Appendix Table 1). The incidence of HIV attributed to heterosexual contact, injection drug use, and male-to-male sexual contact and injection drug use decreased by 6.3% (CI, 7.4% to 5.2%), 10.7% (CI, 12.8% to 8.6%), and 4.3% (CI, 6.9% to 1.6%) per year, respectively. The decreasing trend for persons who inject drugs may have been leveling off since 2014, with stable incidence in 2014 and 2015 (P= 0.22). The incidence decreased among both male and female adults and adolescents with infection attributed to heterosexual contact or injection drug use, with the reduction potentially leveling off among both males and females who inject drugs after 2014 (stable incidence in 2014 and 2015; P= 0.32 and P= 0.47, respectively) (Appendix Table 1). The incidence of HIV among MSM remained relatively stable, with 26700 infections in 2008 and 26200 in 2015. Men who have sex with men had the highest annual rates of HIV incidence each year, with a 2015 rate (513.7 [CI, 443.7 to 583.7] per 100000) 16 times that of persons with infection attributed to injection drug use (32.1 [CI, 20.3 to 43.8] per 100000) and 135 times that of persons with infection attributed to heterosexual contact (3.8 [CI, 3.3 to 4.2] per 100000) (Table 1). Figure. Estimated HIV incidence among persons aged 13 years or older, by transmission category (adjusted for missing transmission category), United States, 2008 to 2015. Shown are the estimated annual percentage changes and associated 95% CIs. Appendix Table 1. Estimated HIV Incidence Among Persons Aged 13 Years, by Sex and Transmission Category*United States, 20082015 Table 1. Estimated HIV Incidence and Prevalence and the Percentage of Undiagnosed HIV Infections Among Persons Aged 13 Years or Older, b
HIV incidence and prevalence among men who have sex with men in the U.S. remain high, with undiagnosed infections accounting for 67% of new infections in 2015.
Full text analysis coming soon...