HIV Rising Among Young Black MSM, Stable Overall
Prejean J, et al "Estimated HIV incidence in the United States, 2006-2009" PLoS ONE 2011; 6(3): e17502.
Published: August 03, 2011
The rate of new HIV infections in the U.S. is roughly stable at about 50,000 a year, but there are alarming increases among some groups, the CDC reported Wednesday.
In the four years from 2006 through 2009, the annual number of new infections ranged from 48,100 to 56,000, not a significant increase, according to Joseph Prejean, PhD, and colleagues at the agency.
But during the same period, the rate of new infections among those ages 13 through 29 rose 21%, driven by a 34% increase among men in that age group who have sex with men, the researchers reported online in PLoS ONE.
That increase, in turn, was driven largely by a 48% increase among blacks in the group of young men who have sex with men -- from 4,400 new infections in 2006 to 6,500 new cases in 2009, Prejean told reporters during a telephone press conference.
* Explain that the rate of new HIV infections in the U.S. is roughly stable at about 50,000 a year, but there are alarming increases among some groups, especially black men ages 13 through 29 who have sex with men.
* Note that men who have sex with men accounted for more than half of all new infections every year.
The overall stability is reassuring but not satisfactory, CDC director Thomas Frieden, MD, told reporters.
"We're glad it's not increasing, but it's not good enough," he said.
Frieden reiterated the agency's suggestion that all Americans should have an HIV test at least once, and more often if they are at elevated risk of infection.
Knowing one's HIV status becomes even more important, he said, in the light of recent data showing that treatment of HIV almost completely prevents transmission of the virus.
The new figures are the first multiyear estimates using newer and more refined methods to capture new infections, Prejean and colleagues reported.
They are based on surveillance data collected in 16 states and two cities where at least 15% of blood samples from newly diagnosed patients were tested with an assay that distinguishes recent from long-standing infection.
In fact, the lowest level of samples tested using the so-called BED HIV-1 Capture Enzyme Immunoassay was 17% in 2006, 20% in 2007, 27% in 2008 and 22% in 2009, the researchers reported.
Those data were extrapolated to yield national estimates of new infections in each year, including a revised estimate for 2006. Specifically:
* In 2006, estimated HIV incidence overall among those ages 13 or older was 48,600, down from the previous estimate of 56,300.
* In 2007, it was 56,000, dropping to 47,800 in 2008 and rising again in 2009 to 48,100.
* The estimated rate of new infections was 19.8 per 100,000 individuals in 2006, 22.5 per 100,000 in 2007, and 19 per 100,000 in both 2008 and 2009.
In each of the four years, men had the most new infections, accounting respectively for 75%, 76%, 75%, and 77% of all new cases.
Men who have sex with men accounted for 56%, 58%, 56%, and 61% of all new infections in the four years.
And blacks were also disproportionately represented, accounting for 44%, 42%, 46%, and 44% of new infections, although they represent just 14% of the U.S. population.
Over the four years, the rate of new infections among blacks and Hispanics averaged 7.65 and 2.95 times higher, respectively, than the rate among whites.
The striking difference in the rate of new infections among young black men who have sex with men is worrisome, Frieden said. "We are very concerned about that group," he said.
Prejean said it's clear that "individual risk factors do not account for" the disparity.
In fact, he said, men in that group have fewer sex partners than others, are less likely to use illegal drugs that increase the risk of HIV, and are no more likely to engage in unprotected receptive anal sex.
Instead, the explanation is probably more complicated, he said, involving such things as higher rates of sexually transmitted disease, lower rates of HIV diagnoses, less access to healthcare, and "deeply rooted stigma" concerning HIV.
The researchers cautioned that the estimates are based on data from only 16 states and two cities, and numbers for the U.S. are extrapolations. Some high morbidity areas - such as the District of Columbia and California - did not contribute data to the original estimates.
The estimates are also based on the assumption that HIV testing behavior did not change over the four years, they noted.
Estimated HIV Incidence in the United States, 2006-2009
The only population with a change in HIV incidence over the entire four-year period was 13-29 year olds, and within that age group, the only risk group experiencing increases was MSM. Among young MSM the estimated number of new infections increased significantly from 2006-2009; the increase in incidence in this group was largely driven by a statistically significant increase in new HIV infections of 48% (12.2% annually) in young, black/African American MSM.
Overall, HIV incidence in the United States was relatively stable 2006-2009; however, among young MSM, particularly black/African American MSM, incidence increased. HIV continues to be a major public health burden, disproportionately affecting several populations in the United States, especially MSM and racial and ethnic minorities. Expanded, improved, and targeted prevention is necessary to reduce HIV incidence.....there was a 21% (95% CI:1.9%-39.8%; p = 0.017) increase in incidence for people aged 13-29 years, driven by a 34% (95% CI: 8.4%-60.4%) increase in young men who have sex with men (MSM). There was a 48% increase among young black/African American MSM (12.3%-83.0%; p<0.001). Among people aged 13-29, only MSM experienced significant increases in incidence, and among 13-29 year-old MSM, incidence increased significantly among young, black/African American MSM. In 2009, MSM accounted for 61% of new infections, heterosexual contact 27%, injection drug use (IDU) 9%, and MSM/IDU 3%.
Joseph Prejean1*, Ruiguang Song1, Angela Hernandez1, Rebecca Ziebell2, Timothy Green1, Frances Walker1, Lillian S. Lin1, Qian An1, Jonathan Mermin1, Amy Lansky1, H. Irene Hall1, for the HIV Incidence Surveillance Group
1 Division of HIV/AIDS Prevention, National Center for HIV, Viral Hepatitis, STD and TB Prevention, Centers for Disease Control and Prevention, Atlanta, Georgia, United States of America, 2 The Ginn Group, Peachtree City, Georgia, United States of America
The estimated number of new HIV infections in the United States reflects the leading edge of the epidemic. Previously, CDC estimated HIV incidence in the United States in 2006 as 56,300 (95% CI: 48,200-64,500). We updated the 2006 estimate and calculated incidence for 2007-2009 using improved methodology.
We estimated incidence using incidence surveillance data from 16 states and 2 cities and a modification of our previously described stratified extrapolation method based on a sample survey approach with multiple imputation, stratification, and extrapolation to account for missing data and heterogeneity of HIV testing behavior among population groups.
Estimated HIV incidence among persons aged 13 years and older was 48,600 (95% CI: 42,400-54,700) in 2006, 56,000 (95% CI: 49,100-62,900) in 2007, 47,800 (95% CI: 41,800-53,800) in 2008 and 48,100 (95% CI: 42,200-54,000) in 2009. From 2006 to 2009 incidence did not change significantly overall or among specific race/ethnicity or risk groups. However, there was a 21% (95% CI:1.9%-39.8%; p = 0.017) increase in incidence for people aged 13-29 years, driven by a 34% (95% CI: 8.4%-60.4%) increase in young men who have sex with men (MSM). There was a 48% increase among young black/African American MSM (12.3%-83.0%; p<0.001). Among people aged 13-29, only MSM experienced significant increases in incidence, and among 13-29 year-old MSM, incidence increased significantly among young, black/African American MSM. In 2009, MSM accounted for 61% of new infections, heterosexual contact 27%, injection drug use (IDU) 9%, and MSM/IDU 3%.
Overall, HIV incidence in the United States was relatively stable 2006-2009; however, among young MSM, particularly black/African American MSM, incidence increased. HIV continues to be a major public health burden, disproportionately affecting several populations in the United States, especially MSM and racial and ethnic minorities. Expanded, improved, and targeted prevention is necessary to reduce HIV incidence.
The Centers for Disease Control and Prevention (CDC) maintains an HIV surveillance system in which all states and U.S. territories submit data on reported diagnoses of HIV. The data are de-duplicated at CDC both within and across states, and the reported case counts are adjusted for reporting delay to estimate the number of new diagnoses of HIV infection and AIDS, annually . These estimates have been used to track the HIV epidemic in the United States, but because HIV diagnosis can occur at any point during the long latency between infection and symptom development in HIV disease the estimates have been limited by their reliance on HIV testing and reporting practices. In response to these limitations, CDC, in conjunction with a number of state and local health departments, implemented national HIV incidence surveillance, using a serologic marker of recent HIV infection to classify new diagnoses of HIV infection as either recent or long-standing . Additional data on history of HIV testing and antiretroviral use are collected to determine sampling weights for estimation of the number of new HIV infections in the United States, both diagnosed and undiagnosed . Based on data from HIV incidence surveillance, Hall and colleagues developed the first national HIV incidence estimate based on a direct measure of recency of infection . It was estimated that in 2006 approximately 56,300 (95% confidence interval [CI]: 48,200-64,500) individuals were infected with HIV.
HIV surveillance is a dynamic system with additional data continually reported to state surveillance systems, and estimates of HIV diagnoses and incidence are updated to reflect these newly available data, science, and programmatic considerations. For example, the incidence estimation model is sensitive to sudden changes in HIV testing patterns which could influence estimates of HIV incidence  if recommendations for routine HIV testing  are fully implemented. In addition, the continued refinement of the method for estimating the mean recency period for the BED HIV-1 Capture Enzyme Immunoassay (BED; the assay that currently serves as the serologic marker of recent infection)  and for modeling the recency period distribution  has allowed for improvement of the method of modeling HIV incidence . We updated the earlier estimate of HIV incidence for 2006 based on additional data and methodological refinements, and extended the previous results with estimates for the years 2007, 2008, and 2009.
The total number of persons aged 13 years and older diagnosed with HIV infection in the surveillance areas for the years 2006-2009 and reported through June 2010 was 29,279, 29,943, 28,831, and 27,040, respectively. Adjustment for reporting delay brought the totals to 30,702, 31,883, 31,357, and 31,162 respectively. Among individuals not diagnosed with AIDS within 6 months of HIV diagnosis, the number with BED results by year was 6,096 (31%) for 2006, 7,615 (37%) for 2007, 8,863 (44%) for 2008 and 9,615 (50%) for 2009. Among individuals who had a remnant diagnostic specimen tested with the BED in 2006, a higher percentage were black/African American, men who have sex with men (MSM) and in the youngest age group when compared to the distribution of new diagnoses. In 2007 a higher percentage were women, black/African American and in the youngest age group, and a lower percentage were injection drug users. In 2008 and 2009, a higher percentage were black/African American and in the youngest age group (Table S3). Of those without a diagnosis of AIDS at or within six months of HIV diagnosis, after imputation, the percent classified recent by year was 31%, 33%, 31%, and 30%. By year, 10,954 (37%), 13,322 (44%), 14,031 (49%), and 14,805 (55%) individuals had information on whether they had had a previous HIV-negative test, and after imputation 17,033 (58%), 16,533 (55%), 16,465 (57%), and 15,237 (56%) were classified as repeat testers, respectively. Among repeat testers, the percentage with T less than or equal to 12 months was 43%, 41%, 42%, and 43% respectively, by year. The proportion of cases with concurrent HIV and AIDS diagnoses was 21%, 19%, 19% and 18%, respectively.
Based on the revised stratified extrapolation approach with a recalculated mean STARHS recency period using the BED of 162 days and using new diagnoses of HIV infection reported through June 2010, an estimated 48,600 individuals aged 13 years or older in the United States were infected with HIV in 2006 (95% CI: 42,400-54,700), with an additional 56,000 (95% CI: 49,100-62,900), 47,800 (95% CI: 41,800-53,800) and 48,100 (95% CI: 42,200-54,000) infected in 2007, 2008 and 2009, respectively. In each year, the most new infections occurred in males (accounting for 75%, 76%, 75%, and 77% of new infections respectively), MSM (56%, 58%, 56%, and 61%), and blacks/African Americans (44%, 42%, 46%, and 44%). The rate of new infections overall for 2006 through 2009 was estimated to be 19.8 (95% CI: 17.3-22.2), 22.5 (95% CI: 19.7-25.3), 19.0 (95% CI: 16.6-21.4), and 19.0 (95% CI: 16.6-21.3) per 100,000 individuals, respectively. Blacks/African Americans and Hispanics/Latinos experienced the heaviest impact of the epidemic with rates that were 7.4 and 2.8 times the rate in whites respectively in 2006, 7.1 and 3.0 times the rate in whites in 2007, 8.4 and 3.0 times the rate in whites in 2008, and 7.7 and 2.9 times the rate in whites in 2009 (Table 1). Each year blacks/African American males had the highest rate of new infections, and among women, black/African American women also experienced the highest HIV incidence rates (Figure 1).
From 2006 to 2009 there was no significant change in HIV incidence overall and there was no significant change in incidence in any race/ethnicity group or risk group overall. There was an overall significant increase in HIV incidence from 2006 to 2007 (15%, 95% CI: 3.6%-26.8%; p = 0.006) with increases in men (17%, 95% CI: 3.0%-31.1%; p = 0.01), Hispanics/Latinos (24%, 95%CI: 0.2%-49.7%; p = 0.027), young people 13-29 years old (29%, 95% CI: 8.6%-49.7%; p = 0.002), and MSM (20%, 95% CI: 2.7%-36.8%; p = 0.013). In all of these groups, except young people, the estimated HIV incidence decreased significantly between 2007 and 2008, in each case falling below 2006 levels. In young people aged 13-29 years the estimated incidence of HIV infection decreased in 2008 as compared to 2007, but remained higher than in 2006, and this group was the only group to evidence a statistically significant increase in HIV incidence between 2006 and 2009 (Table 1). Within the youngest age group only black/African American males experienced a statistically significant increase in HIV incidence from 5,300 (95% CI: 4,200-6,400) in 2006 to 7,600 (95% CI: 6,300-8,900) in 2009, a 43% increase (95% CI: 11.6%-75.2%; p = 0.001) (Table 2). HIV incidence was essentially unchanged 2006-2009 in Hispanic/Latino, and white males aged 13-29 (Tables 3 and 4).
Among the 13-29 year age group, by year, MSM made up 62%, 64%, 65%, and 69% of new infections, including 59%, 58%, 62%, and 66% of new infections among blacks/African Americans, 63%, 68% 65%, and 72% of new infections among Hispanics/Latinos, and 65%, 70%, 71%, and 72% of new infections among whites. Within MSM there were racial/ethnic differences in the age distribution of new infections. Among black/African American and Hispanic/Latino MSM, most new infections occurred in the youngest MSM, with MSM 13-29 accounting for 49%, 57%, 62%, and 60% of new infections by year among blacks/African Americans and 40%, 43%, 46%, and 45% of new infections by year among Hispanics/Latinos, compared with 23%, 25%, 28%, and 28% among whites (Table 5). Although there was a significant increase in new infections from 2006 to 2009 (34%, 95%CI: 8.4%-60.4%; p = 0.002) among young MSM overall (EAPC = 8.1%, 95% CI: 1.9%-14.9%; p = 0.01), the only significant increase in any MSM subgroup 2006-2009 was among young, black/African American MSM (48%, 95% CI: 12.3%-83.0%; p = 0.001), with EAPC 12.2% (95% CI: 4.2%-20.9%; p = 0.002). The EAPC among other young MSM was not significant (Figure 2). Among white MSM, the group most affected 2006-2008 was men 30-39 years of age, who accounted for 35%, 34%, and 31% of new infections by year and in 2009, men 40-49 years of age, who accounted 30% of new infections (Table 5).
Rather than expand the analysis to include in each year's analysis all areas that met the inclusion criteria for that year, we chose to limit the surveillance areas contributing data for analysis to those that met the inclusion criteria for all analysis years in order to ensure greater comparability across the analysis period. If we had expanded the analysis, the HIV incidence estimate for 2006 would have been higher by approximately 1.4% new infections and the estimates for 2007-2009 would have been lower than those presented by 2.3%%, 1.5%, and 3.1%respectively. Consistent with the analysis from 16 states and 2 cities, in the expanded analysis the only MSM subgroup to show a significant increase 2006-2009 was young, black/African American MSM (data not shown).
Based on the revised stratified extrapolation approach for estimating HIV incidence, the number of new infections in the United States remained relatively stable between 2006 and 2009. Our analysis examines HIV incidence over a four-year period to provide the clearest picture of the current status of trends in incidence. The only population with a change in HIV incidence over the entire four-year period was 13-29 year olds, and within that age group, the only risk group experiencing increases was MSM. Among young MSM the estimated number of new infections increased significantly from 2006-2009; the increase in incidence in this group was largely driven by a statistically significant increase in new HIV infections of 48% (12.2% annually) in young, black/African American MSM.
The point estimate of the number of new HIV infections in 2006 presented here is somewhat lower than the previous estimate but within the confidence interval of that estimate . The difference is attributed to additional years of data including removal of duplicate case reports between states, and several improvements in the method, including calculating the probability of testing within the STARHS recency period using the observed testing frequency rather than the mean recency period which allowed us to mitigate the impact of artificially limiting the probability for repeat testers who test frequently. Additionally, using newly available information on the length and distribution of the STARHS recency period would be expected to decrease the estimate by approximately 4% even if no other revisions to the model had been made because the updated mean recency period is 4% higher than the mean recency period used for the previous estimate. Finally, uncertainty related to reporting delay adjustment may impact estimates. As additional years' data become available we expect that the HIV incidence estimates presented here will be revised further, consistent with estimates based on surveillance data which are subject to reporting delay. Because of anticipated adjustments to these estimates over time, it is important to focus primarily on the overall population impact and trends, rather than the precise point estimates, when drawing conclusions from these analyses.
Consistent with the higher rates of HIV diagnoses among MSM in general , and of both HIV diagnoses and HIV incidence in African American and Hispanic/Latino men and women -, the 2006-2009 HIV incidence estimates continue to demonstrate the disproportionate impact of HIV disease in these groups. Although the results did not demonstrate a significant increase in HIV incidence among MSM overall, they indicate an increase among young people aged 13-29 that is largely driven by increases in incidence among young MSM, particularly young black/African American MSM. The increase in HIV incidence in young MSM is in line with the increase seen in new diagnoses in MSM in recent years in the United States  and internationally  as well as with increases in HIV incidence seen in the United States using an extended back-calculation model  and with international trends in incidence in MSM -.
These estimates are subject to several limitations. First, in order to maintain consistency across the analysis years, we limited our analysis to only those states that consistently met the inclusion criteria for all analysis years. Therefore, the estimates are based on data from 16 states and 2 cities. The included areas represented 61% of reported cases of AIDS in the United States for the years 2006 and 2009, 62% for the years 2007 and 2008.
Because data were only available for a limited number of surveillance areas, we extrapolated our HIV incidence estimates from the included areas to the rest of the United States by applying the ratio of HIV incidence to AIDS incidence in the included areas to the AIDS incidence in the rest of the United States. To evaluate the validity of this extrapolation we compared the ratio of HIV diagnoses (as a proxy for incidence) to AIDS diagnoses in the included areas to that ratio in the areas to which we extrapolated. By extrapolating from the same surveillance areas each year, we may have underestimated HIV incidence overall by approximately 4%; using different areas each year, we may have underestimated HIV incidence by about 3%. Additionally, while the represented areas included several jurisdictions with very high morbidity, others-including the District of Columbia and the state of California-were not included in the estimate. Because HIV incidence in an area is driven by both risk behavior and HIV prevalence, the HIV incidence estimate may have been higher if these areas had been included. However, our method of extrapolating to the United States as a whole using the same 68 strata used for estimation likely compensates for the lack of inclusion of some cities with high HIV morbidity. It also limits overestimation of HIV incidence due to the differing ratio of HIV diagnoses to AIDS diagnoses in the incidence versus non-incidence areas.
Next, an assumption of the model is that HIV testing behavior has not changed over several years . Comparing by year, the percentage of persons diagnosed with HIV and AIDS concurrently (within the same month), the percentage of persons classified as repeat testers, and among the repeat testers, the average time since last negative HIV test, the data indicate that HIV testing behavior remained stable over the analysis period. As a result of the apparent stability in these indicators of HIV testing behavior we did not adjust the model to account for a change in HIV testing behavior. CDC has recently funded HIV testing initiatives to reach individuals whose behavior puts them at high risk for HIV transmission and, in 2006, published recommendations for routine, opt-out testing for all individuals 13-64 seeking medical care . While there are no data to indicate the level of uptake of the recommendations, evidence suggests that implementation of opt-out testing does increase HIV testing . The data on history of HIV testing and antiretroviral use collected through HIV incidence surveillance may not have been collected over a long enough period of time to reflect recent changes in HIV testing behavior. If HIV testing has increased recently we may have overestimated HIV incidence in recent years. CDC will continue to track these indicators of testing behavior and adjust the estimation model as needed to reflect any changes detected in future years.
Additionally, concerns about using the BED assay within STARHS for cross-sectional estimation of HIV incidence have been raised for some HIV subtypes due to the misclassification of long-standing infections as recent , however, these issues are less relevant in the United States because of the predominance of HIV subtype B and as a result of the integration of STARHS results and supplemental data on history of HIV testing and antiretroviral use into national HIV surveillance . This integration ensures that cases diagnosed with AIDS, at or within 6 months of HIV diagnosis, can be correctly classified as long-standing irrespective of STARHS result and allows for the removal and subsequent imputation of STARHS classification for cases in which antiretroviral use occurred prior to diagnosis. However, some HIV infections that were diagnosed late in the course of HIV disease, but not close to the time of AIDS diagnosis could have been misclassified as recent infections, adding to the uncertainty in calculation of HIV incidence and in comparison of HIV incidence across years.
In our previous work we noted that the HIV incidence estimate for 2006 could have been an overestimate if we had underestimated the likelihood of testing for HIV within one year of infection. The revised model for incidence estimation presented here addresses this concern by using the entire distribution of the STARHS recency period which limits the impact on the weights assigned to repeat testers who test more frequently than once per year, thus limiting the amount of overestimation bias inherent in the previous model. It is still possible, however that we could have overestimated incidence if a significant number of individuals were motivated to be tested for HIV by a real or perceived recent exposure, as this motivation was not addressed in the estimation model. We previously estimated that we may have overestimated HIV incidence by as much as 7% as a result of excluding motivation from the model to calculate incidence .
Finally, a number of additional assumptions of the model have been previously described . These assumptions include that HIV incidence has been stable in recent years, that the likelihood of HIV testing prior to a diagnosis of AIDS is constant both with respect to time and to duration of HIV infection, that all HIV infections will eventually be diagnosed (either through testing or through death), and that individuals accurately report the HIV testing and treatment information, especially the date of their last negative HIV test. In addition, an assumption of using multiple imputation for missing data on BED result and HIV testing group (repeat or new testers) is that these data are missing at random, though not necessarily missing completely at random. These assumptions were addressed in the previous description of the model, and their effects were determined to be minimal, or to counterbalance one another.
The estimates for 2006-2009 continue to underscore the disproportionate toll that the HIV epidemic has taken on several populations in the United States including racial/ethnic and sexual minorities and injection drug users with 95% of new infections 2006-2009 estimated to have occurred in individuals in one or more of these groups. Though transmission rates have decreased substantially since the beginning of the epidemic , public health programs are presented with new challenges. There is a need to address the prevention needs both of people at risk for HIV infection as well as of those living with HIV disease-those who are aware of their HIV status and those who are not. With an estimated 21% of people living with HIV unaware of their HIV status , and the majority of new HIV infections transmitted by these individuals , it is important to expand testing to those people most at risk and provide them with care and prevention services. Adequate funding and services should be directed to individuals at greatest risk for acquiring and transmitting HIV infection, if we are to make a further impact on the HIV epidemic in the United States.