Trends of HIV Seroconversion Young Adults in the US Army, 1985 to 1989 Among John G. McNeil, MD; John F. Brundage, MD; Lytt I. Gardner, PhD; Z. Frank Wann, MS; Philip O. Renzullo, MPH; Robert R. Redfield, MD; Donald S. Burke, MD; Richard N. Miller, MD; and the US Army Retrovirus Research Group Because soldiers in the US Army are recurrently tested for the presence of antibody to the human immunodeficiency virus (HIV), HIV seroconversion rates can be directly measured. From November 1985 through October 1989,429 HIV seroconversions were detected among 718 780 soldiers who contributed 1 088 447 person-years of follow-up time (HIV seroconversion rate, 0.39 per 1000 person-years). Period-specific seroconversion rates declined significantly, from 0.49 per 1000 person-years (November 1985 through October 1987) to 0.33 per 1000 person-years (November 1987 through October 1988) to 0.29 per 1000 person-years (November 1988 through October 1989). The HIV seroconversion risk among active-duty soldiers was significantly associated with race/ ethnic group, age, gender, and marital status. Based on these trends, we estimate that approximately 220 soldiers (95% confidence interval, 160 to 297 soldiers) were infected with HIV during 1989 and 1990, with potentially fewer in future years. (JAMA. 1991;265:1709-1714) SINCE 1981, surveillance of the ac¬ quired immunodeficiency syndrome (AIDS) has been a keystone of the na¬ tional effort to monitor the human im¬ munodeficiency virus (HIV) epidemic in the United States and to provide nearterm forecasts of future AIDS cases.1,2 However, over the past decade it is quite likely that the rate and, potential¬ ly, the determinants of HIV infection have changed. Documenting and track¬ ing an evolving infection epidemic has proved difficult because of reliance on a surveillance end point (ie, AIDS) repre¬ senting the clinical end stage of HIV infections that occurred over a long and variable period in the past. The tempo¬ ral trend of AIDS incidence reflects his¬ torical changes in HIV epidemic dynam¬ ics, and, through methods of backcalculation, AIDS case reports have been used to estimate the occurrence of HIV infection at specific times in the past.3"0 However, these methods pro¬ vide insufficient information about the From the Divisions of Preventive Medicine (Drs McNeil, Brundage, Gardner, and Miller and Mr Ren- zullo) and Retrovirology (Drs Redfield and Burke), Walter Reed Army Institute of Research, Washington, DC; and SRA Technologies, Alexandria, Va (Mr Wann). The opinions or assertions contained herein are the private views of the authors and are not to be construed as official or as reflecting the views of the Department of the Army or the Department of Defense. Reprint requests to Division of Preventive Medicine, Department of Epidemiology, Walter Reed Army Institute of Research, Washington, DC 20307-5100 (Dr McNeil) occurrence of new HIV infections over the last 2 to 3 years, a period of critical importance for evaluating changes in the epidemic related to current inter¬ ventions and for making more accurate projections of future HIV-associated morbidity. Direct measurement of the incidence of new HIV infection greatly enhances the ability to track the evolving HIV infection epidemic and substantially im¬ proves the accuracy of epidemic fore¬ casting. Unfortunately, HIV infection incidence is difficult to directly measure and can only be observed in groups that undergo repeated testing for antibody to HIV (HIV Ab). Currently, these groups include aging cohorts of volun¬ teers recruited because of increased risk of HIV infection,6^ repeat blood do¬ nors,9 and US military personnel.1011 The incidence of HIV seroconversion in the US Army and Navy has been re¬ ported for the first 2 years of routine HIV Ab testing programs.1011 The pre¬ sent report analyzes the temporal trend and demographic correlates of incident HIV seroconversion among young adults in the US Army during the period from November 1985 to October 1989. METHODS The US Army is composed of approxi¬ mately 760 000 volunteers, mostly young adults (mean age, 24.5 years; 99.8% are between 17 and 50 years), about 160 000 of whom are new to the Army each year. Soldiers continuously enter the Army from every state, the District of Columbia, several overseas territories, and virtually every US county. Table 1 compares demographic characteristics of soldiers in the Army vs 17- to 50-year-olds in the general US population. Compared with the US pop¬ ulation at large, the Army is composed of substantially greater proportions of males, 17- to 30-year-olds, and persons of minority race/ethnicity. Since November 1985, soldiers on ac¬ tive duty in the US Army have been routinely tested for the presence of HI V Ab. Also since that time, all applicants for military service have been required to have test results negative for HIV Ab prior to induction into a uniformed ser¬ vice. Soldiers who remain on active duty in the Army are required to undergo retesting for HIV Ab at least every 2 years. During the past 2 years most soldiers have been retested, with a mean time from the previous test of about 16 months. Since the inception of the Department of Defense HIV screening programs, the US Army HIV Data System and retrovirus reference laboratory, located at the Walter Reed Army Institute of Research, Washing¬ ton, DC, have received, processed, and cataloged all HIV antibody test results and serological specimens from US Army soldiers and civilian applicants for military service. Descriptions of the US Army's testing program, proce¬ dures, screening algorithms, and as¬ sessments of screening test perfor¬ mance have been reported in detail elsewhere.12"14 During the period we analyzed, No¬ vember 1, 1985, through October 31, 1989, four laboratories conducted all HIV Ab testing: Damon Laboratory, Irving, Tex; the US Army 10th Medical Laboratory, Landstuhl, Federal Re¬ public of Germany; Tripler US Army Medical Center, Honolulu, Hawaii; and the US Army Retrovirus Reference Laboratory, Rockville, Md. AU sera were initially tested by a commercially available enzyme-linked immunosor- Downloaded from jama.ama-assn.org at Florida State University on November 14, 2011 Table 1. Demographic Characteristics of US Pop¬ ulation (17 to 50 Years)* and US Army Population Table 2. —Interval-Specific Human Immunodeficiency Virus Seroconversion Rates — US _(n Age, y US Population = 126 612 000),% 17-20 12 21-30 31-40 41-50 Gender M F 33 32 23 Army Population 757 (n 000),% = 18 53 23 5 49 51 91 W 84 62 B 12 4 30 Race Other *1980 US Census data 9 8 projected to 1989. bent assay (ELISA).lD Reactive sera immediately retested in duplicate by ELISA, and those found repeatedly reactive were further tested using the Western blot method.1617 Diagnostic Western blots were routinely confirmed by Western blot of a second indepen¬ dent serum specimen. Quality control of testing was assured by stringent per¬ formance requirements.1819 Test results were entered into an electronic data¬ base, transferred to tape, and forward¬ ed to the US Army HIV Data System for analysis and archiving. Negative test results were defined by nonreactive ELISA or reactive ELISA and nondiagnostic Western blot results. Positive test results were defined by ELISA reactivity and a diagnostic Western blot (in duplicate) and by clini¬ cal confirmation and disease staging.20 Laboratory procedures and diagnostic criteria for test positivity were consis¬ tently applied throughout the survey period, except for the definition of a diagnostic Western blot, which changed in May 1987 from the presence of anti¬ bodies to HIV-encoded proteins p24 and p55 and/or gp41 to the presence of at least two of the following bands: gpl20/160, gp41, or p24. This change in blot interpretive criteria rarely affected the final diagnosis.21 The HIV Ab testing programs in the Army are considered either "routine" or "adjunct." Routine programs include screening of applicants for military ser¬ vice, periodic birth month screening of all soldiers on active duty (at least every 2 years), and screening during periodic physical examinations (every 2 to 5 years). Adjunct testing is usually asso¬ ciated with medical evaluation and care and includes such programs as the test¬ ing of patients undergoing evaluation for sexually transmitted disease, test¬ ing of soldiers enrolled in drug rehabili¬ tation programs, testing of sex partners of HlV-infected persons, testing per¬ formed because HIV infection is sus¬ pected based on clinical signs, and test¬ ing specifically requested by a soldier were Interval Dates No. of Seroconverters Incidence Rate, Person-Years of Follow-up per 1000 Person-Years Relative Rate* (95% Confidence Interval) 1_11/85-10/87_239_489 322_049_1.00 2_11/87-10/88_140_428 962_033_0.67 (0.54-0.82) 3 11/88-10/89 50 170 163 0.29 0.60(0.44-0.82) *Rate relative to interval 1 who perceives his or her risk of HIV infection to be high. During the period we studied, inclu¬ sion in the incidence analysis required documentation of an initial negative HIV Ab test result from any routine or adjunct program occurring on or after November 1, 1985, and followed by at least one subsequent HIV Ab test prior to October 31, 1989. Follow-up time commenced at the date of an individual's earliest negative test result and accrued until an end point was reached. For this analysis, an end point was defined as either detection of HIV Ab seroconver¬ sion or the date of the latest HIV Ab negative test result to occur before Oc¬ tober 31, 1989. The actual date of HIV Ab seroconversion was estimated to be the midpoint date between an individ¬ ual's latest negative and earliest diag¬ nostic HIV Ab test result. The HIV seroconversion rates were calculated for three time intervals: November 1, 1985, to October 31, 1987 (interval 1); November 1,1987, to October 31, 1988, (interval 2); and November 1, 1988, to October 31, 1989 (interval 3). Interval 1 spans a 2-year period because relatively few HIV Ab tests were conducted be¬ tween November 1985 and October 1986. Follow-up time during the initial 2-year interval closely approximates observed follow-up times for subse¬ quent 1-year periods. Seroconversion rates are not presented for November 1985 through October 1986 because they are based on very short follow-up times and yield highly unstable estimates. STATISTICAL ANALYSES Seroconversion rates were computed as interval-specific incidence densities (IDs) among soldiers who contributed person-time during an interval, condi¬ tional on an initial negative result of an HIV Ab test performed at least 60 days prior to an end point test date. This 60day requirement was established to minimize misclassification as initially negative because of specimen labeling and handling error or data management error; these errors were generally rec¬ ognized and corrected within a few weeks of occurrence. All rates are re¬ ported as the number of HIV Ab seroconversions occurring per 1000 person- years of follow-up within each interval. The Incidence density ratios (IDRs) were calculated by the formula IDR ID/ID2, where the subscript numbers represent various demographic and pe¬ riod-specific strata. Poisson regression was used to calculate the independent summary relative risk estimates for variables that contributed to the under¬ lying hazard rate. Using the Poisson method and a completely categorical model, the expected number of HIV seroconversions per stratum of covariate was weighted by the stratum-spe¬ cific follow-up time (in person-years). A single regression model was fitted to the entire data set, producing summary rel¬ ative risk estimates adjusted for all oth¬ er variables in the model. The final con¬ tribution of each stratum of covariate was modeled with a log link. Actual model fits were carried out with the EGRET statistical package.22 This mod¬ el produced adjusted stratum-specific relative risks (RRs) expressed as RRk exp(ßk), where k indicates the stratum of the covariate.23 Confidence intervals were calculated using SEs of the estimate of ßk. = = RESULTS Between November 1, 1985, and Oc¬ tober 31, 1989, 429 HIV seroconver¬ sions were detected among 718 780 per¬ sons who were tested two or more times for the presence of HIV Ab. These 718 780 soldiers contributed 1088447 person-years of follow-up time, with a resulting overall crude rate of 0.39 (95% confidence interval, 0.36 to 0.43) HIV seroconversions per 1000 person-years of follow-up during the first 4 years of testing within the US Army. Table 2 summarizes interval-specific seroconversion rates over the 4 years analyzed. Relative rates in Table 2 refer to the HIV seroconversion rate in an interval relative to interval 1. A signifi¬ cant rate reduction occurred from inter¬ val 1 to interval 2 (P<.001) but not be¬ tween interval 2 and interval 3 (P= .453), indicating a slowing deceler¬ ation of the epidemic curve. The HIV seroconversion rates were consistently higher when the end point test source was an adjunct program vs a routine testing program. The HIV sero¬ conversions were most efficiently de- Downloaded from jama.ama-assn.org at Florida State University on November 14, 2011 tected in testing programs oriented to¬ ward individuals at presumed high risk (clinical suspicion of infection or dis¬ ease, sex partner of infected persons, soldier-initiated testing), followed by testing of populations at presumed high risk (sexually transmitted disease clin¬ ics, drug and alcohol treatment pro¬ grams), followed by routine testing pro¬ grams, such as birth month testing. Between November 1988 and October 1989, adjunct tests accounted for 5% of all HIV Ab tests performed in the Army, while accounting for nearly 20% of detected seroconversions. The pro¬ portion of all HIV Ab test samples sub¬ mitted from adjunct programs re¬ mained relatively constant from year to year. Adjunct test-associated serocon¬ version rates declined significantly, from 1.03 per 1000 person-years in in¬ terval 1 to 0.61 per 1000 person-years in interval 2 (P<.01), increasing to 0.78 per 1000 person-years in interval 3 (P>.05). Routine test-associated sero¬ conversion rates fell from 0.44 per 1000 person-years in interval 1 to 0.29 per 1000 person-years in interval 2 (P<.001), declining again to 0.22 per ¡2« ftif 1.25 1.00 0.75 0.50 5* 0.25 8§ Interval 1 Interval 2 D Interval 3 S 0.00 Hispanic Black White Fig 1 .—Temporal trend of human immunodeficien¬ cy virus seroconversion by race/ethnic group. Fig 2.—Temporal trend of human 1000 person-years in interval 3 (P>.05). Figures 1 through 4 depict temporal trends of HIV seroconversion for vari¬ ous demographic groups. Trend ana¬ lyses for Hispanic soldiers were limited, because relatively small numbers pro¬ duced unstable rate estimates. Time- dependent demographic characteris¬ tics, such as age, duration of Army service, and marital status, were based each person's status in interval. on Age Unadjusted infection incidence rates did not vary substantially by age, ex¬ cept for a significantly lower rate among soldiers 35 years of age 3). immunodeficiency virus Interval-specific seroconversion by 2.0 older (Table seroconversion or Fig 3.—Temporal race and gender. of HIV seroconversion occurred only among 25- to 29-year-olds. Among black teenage soldiers, HIV incidence in¬ creased over twofold from interval 1 to interval 3. During the three intervals, regardless of race/ethnicity, soldiers who were 25 through 29 years of age had the highest infection rate but also had the greatest rate reduction over time. On the other hand, soldiers with the lowest overall infection rate (eg, &35 years old) had a higher rate of HIV sero¬ conversion during interval 3 than dur¬ ing interval 1. Gender During the first 4 years of HIV screening in the Army, 407 males (0.42 per 1000 person-years) and 22 females (0.15 per 1000 person-years) were de¬ tected as newly infected with HIV. The trend of human immunodeficiency virus seroconversion by Interval 1 Interval 2 D Interval 3 0 Interval D given Race or Ethnic Group Figure 1 illustrates interval-specific HIV seroconversion rates for white, black, and Hispanic soldiers. For each race/ethnic group, seroconversion rates were significantly greater during inter¬ val 1 than interval 2. Among white and Hispanic soldiers, the trend of reduc¬ tion in HIV seroconversion rates con¬ tinued between the last two intervals. Such a reduction was not observed among black soldiers, and seroconver¬ sion rates remained stable at about 0.73 per 1000 person-years during the last 2 years. During the most recent period surveyed, November 1988 through Oc¬ tober 1989, black soldiers had over a sixfold increased crude rate of new HIV infection compared with all other race/ ethnic groups in the US Army. This extremely elevated relative infection rate among blacks was adjusted down¬ ward in the multivariate analysis from 6.08 to 3.80. age and race. 1-5H a rates for each age group, further strati¬ fied for blacks and whites, are shown in Fig 2. Age-specific seroconversion rates were significantly higher among black soldiers of all ages than among white counterparts. Among white sol¬ diers, there was continuous reduction in interval-specific seroconversion rates at all ages, except for soldiers 35 years of age or older. The rate of HIV serocon¬ version declined most significantly among 25- to 29-year-old white soldiers, from 0.43 per 1000 person-years during interval 1 (25 seroconversions) to 0.14 per 1000 person-years during interval 2 (eight seroconversions) to no new infec¬ tions during 58 290 person-years of fol¬ low-up in interval 3. White soldiers who were 35 years of age or older experi¬ enced consecutive increases in their rate of seroconversion from interval 1 through interval 3. Among black sol¬ diers, substantial reduction in the rate 1 Interval 2 Interval 3 s. 10-i 8 g || 0.5 0.0 wmkJHWl <20 20-24 25-29 30-34 >35 Age, y White Soldiers >—i—r- <20 20-24 25-29 30-34 >35 Age, y Black Soldiers Black Overall White Male Soldiers K%1.H,n,ftl, Black Overall White Female Soldiers Downloaded from jama.ama-assn.org at Florida State University on November 14, 2011 Table 3.—Age-Specific Human October 1989 Interval 1 Interval 2 Interval 3 Immunodeficiency Virus Seroconversion Rates for November 1985 Through Incidence Rate No. of Age Group, y Person-Years of Follow-up Seroconverters <20 20-24 25-29 per 1000 Person-Years 0.36 (0.26-0.47) 0.43 (0.37-0.50) 143 577 30-34 Unmarried (95% Confidence Interval), 239 099 152 703 60 0.46 0.39 0.22 (0.38-0.56) (0.30-0.51) (0.16-0.31) Married Table 4. -Poisson Ratios Fig 4.—Temporal trend of human immunodeficien¬ cy virus seroconversion by marital status. Regression Model of Adjusted Human Immunodeficiency Virus Seroconversion Rate Adjusted Rate Ratio (95% Confidence Interval) Variable Age, y (vs <25 y) overall crude male-female IDR was 2.8. The adjusted male-female IDR in¬ creased to 3.47 in the multivariate anal¬ ysis. Among black soldiers, males (0.90 per 1000 person-years) were at signifi¬ cantly higher risk for acquiring HIV in¬ fection than females (0.24 per 1000 per¬ son-years); the male-female IDR was 3.7 (95% confidence interval, 2.1 to 6.5). Among white soldiers, the male sero¬ conversion rate (0.23 per 1000 personyears) was also significantly higher than the female rate (0.08 per 1000 personyears); the male-female IDR was 2.9 (95% confidence interval, 1.4 to 4.8). Among males, the black-white IDR was 3.9, while, among women, the blackwhite IDR was 3.0. These crude rate ratios were similar to comparable ratios reported for prevalent HIV infections in the US Seroconversion trends are summa¬ rized in Fig 3. Among male soldiers, seroconversion rates declined consecu¬ tively overall and among whites but de¬ clined among blacks only from interval 1 to interval 2. Among female soldiers, regardless of race, the rate of new HIV infection during interval 3 did not differ significantly from that during interval 1. During the overall period, the rate among black women (0.24 per 1000 per¬ son-years) was similar to that among white men (0.23 per 1000 person-years). Army.24 Marital Status Overall, soldiers who were not mar¬ ried had a significantly higher crude rate of new infection (0.59 per 1000 per¬ son-years) than married soldiers (0.27 per 1000 person-years); the IDR was 2.2 (95% confidence interval, 1.8 to 2.6). Figure 4 indicates seroconversion trends for married and unmarried sol¬ diers. Among unmarried soldiers, infec¬ tion rates declined from interval 1 to interval 3, while, among married sol¬ diers, rates remained relatively stable. During the most recent interval, the rates among married and unmarried soldiers did not significantly differ. 25-34 =35 Race/ethnicity (vs white) Black Hispanic Male gender (vs female) Unmarried (vs married) Duration of active-duty service, y (vs 4-9 210 Interval 1 s3 y) (vs interval 2) 3 Multivariate Analysis (Poisson Regression) To control for the simultaneous ef¬ fects of potentially confounding demo¬ graphic factors, a Poisson regression model22 was constructed that included the variables listed in Table 4. Table 4 provides adjusted IDRs and 95% confi¬ dence intervals for each adjusted rate ratio. Significant independent determi¬ nants of HIV seroconversion included minority race/ethnicity, age 25 through 34 years, male sex, and unmarried. In¬ dependent of all other factors in the model, the seroconversion rate was sig¬ nificantly higher between November 1, 1985, and October 31, 1987, than in ei¬ ther subsequent period. Rates during intervals 2 and 3 were not significantly different from each other. COMMENT Because soldiers are periodically screened for HIV Ab, it is possible to directly measure and track the inci¬ dence of HIV seroconversion in this large, well-characterized population. Over the surveyed period, HIV inci¬ dence rates declined significantly over¬ all and for many demographic groups, but, disturbingly, remained largely un¬ changed or increased among black men and women in many age groups. Most notably, the seroconversion risk nearly doubled among black teenagers during the 4 years analyzed. 1.40 0.76 (1.10-1.80) (0.49-1.17) 3.80 1.85 <.001 3.47 2.64 (3.05-4.64) (1.27-2.71) (2.23-5.40) (2.11-3.31) 1.18 0.94 (0.90-1.56) (0.69-1.29) .218 1.57 (1.27-1.94) (0.68-1.25) 0.88 .210 .001 <001 .453 Independent of other demographic factors, 25- to 34-year-old soldiers were at significantly higher risk for acquiring HIV infection than other age groups. This finding reflects HIV risk behaviors (eg, risky sex, drug, or partner selec¬ tion practices) that were more likely to occur in this age group. On the other hand, the most dramatic downward trend in HIV seroconversion rates was among 25- to 29-year-old soldiers. Behavioral determinants of HIV ser¬ oconversion have been under intensive investigation by collaborators from the US Army, the Centers for Disease Con¬ trol, and state and territorial health agencies since 1989. Preliminary results of this ongoing case-control study indi¬ cate that most men on active duty ac¬ quired HIV infection via sex with other men and that nearly 75% of these men also had sex with women (who were not known to be at increased risk for HIV infection). In this study ofincident seroconverters, intravenous drug use, al¬ though reported by uninfected controls, has not been reported by any seroconverter to date. The HIV seroconverter case-patients were also much more like¬ ly to report female sex partners at known increased risk for HIV infection (usually because of intravenous drug use) and anonymous female partners who had nonmenstrual bleeding during sex or from whom they acquired a sexu¬ ally transmitted disease.25 The fact that HIV incidence rates fell Downloaded from jama.ama-assn.org at Florida State University on November 14, 2011 among some soldiers but not others might indicate that programs of educa¬ tion and risk reduction had selective ef¬ fects or simply that outmigration pat¬ terns varied among different groups. Over the period analyzed, demographic characteristics of soldiers who left the Army did not change. However, be¬ cause HIV seroconversion is relatively rare among soldiers, demographic char¬ acteristics of those who leave are proba¬ bly not a sensitive indicator of selective outmigration of high-risk individuals. An in-depth qualitative assessment of the potential impact of outmigration patterns on observed incidence trends has recently been completed. This anal¬ ysis involved evaluation of medical in¬ formation for over 750 000 soldiers who left the Army during the survey period. The analysis revealed that soldiers who left the Army were no more likely than demographically comparable soldiers who remained on active duty to experi¬ ence morbidity strongly correlated with subsequent HIV infection diagnosis (McNeil et al26 and P.O.R., J.G.M., and F.W., unpublished results, October 1990). This finding suggests that the observed incidence trends were not heavily influenced by selective outmi¬ gration practices during the survey period. Decelerating infection incidence over time might be explained by a military induction system that effectively and progressively halted inmigration of high-risk persons into the Army popula¬ tion. Figure 5 depicts age-adjusted inci¬ dence rate trends for soldiers with 3 years or less of continuous active-duty service compared with those with more than 3 years of service. Three years was chosen as a cutoff point because it was assumed that most soldiers with 3 years ofservice or less entered the Army after it was widely recognized that HIV Ab screening was routinely performed dur¬ ing the preinduction medical examina¬ tion and within the Army. The figure demonstrates that absolute values and trends of seroconversion were nearly identical for the two groups, and, there¬ fore, selective inmigration of lower-risk individuals does not appear to account for the decelerating incidence trend we report. The overall declining incidence of HIV infection among soldiers may indi¬ cate that education and prevention ef¬ forts in the Army have produced reduc¬ tions in HIV-associated risk behavior. At a minimum, every soldier on active duty is required to receive at least 4 hours of HIV prevention-oriented edu¬ cation each year. The presentation for¬ mat may vary from didactic instruction to interactive video and role playing. Educator proficiency can range from someone who is knowledgeable, charis¬ matic, and experienced to someone who is a relative novice. Most instructional programs focus on health awareness and strategies appropriate for preven¬ tion of sexually transmitted diseases in general. Concurrent with the observed reduction in HIV incidence, substantial reductions in the rates of many other sexually transmitted diseases have been reported among large populations of active-duty soldiers in the United States and the Republic of Korea (Wil¬ liam Ryan, PhD, Walter Reed Army Institute of Research, Washington, DC, oral communication, January 8, 1991). As with virtually all surveillance data, the incidence data we report are potentially limited by sources of bias that require careful consideration. Point estimates of seroconversion risk might be invalid because of selective misclassification, the effects of self-de¬ ferral from testing, or imprecision asso¬ ciated with small numbers and relative¬ ly rare events. Incidence trends could be biased by inconsistent screening test performance, by self-deferral practices that change over time, and by precision constraints. It is highly unlikely that poor or in¬ consistent program performance signif¬ icantly biased the data presented here¬ in, because, from the inception, Army screening programs have operated with strict quality assurances and with ex¬ tremely high specificity.19 Test sensitiv¬ ity has been more difficult to assess, but operating characteristics of the screen¬ ing test panel should not have changed remarkably over the period we surveyed. Seroconversion rates and trends indistinguishable between sol¬ diers with a single negative test result before a positive or negative end point vs soldiers with multiple negative test results before a positive or negative end point. It is highly unlikely that a nega¬ tive HIV Ab test result confirmed by more than one subsequent negative test result was initially misclassified. Thus, the similarity of results for soldiers ini¬ tially classified as HIV-negative based on a single test result compared with soldiers who were persistently HIVnegative is reassuring. Data reported in this analysis would be biased if soldiers at high risk for HIV infection were effectively avoiding test¬ ing or if the frequency of self-deferral were changing over time. Since nearly all routine HIV Ab tests in the Army are conducted as administrative ac¬ tions, a soldier could conceivably post¬ pone testing by simply ignoring the were 1.00 e 5> 0.75 25 Time on Active Duty -ei-<3y - ->3y s e ¿%° 0.25 S. 0.00 i i Interval 1 Interval 2 i Interval 3 Fig 5.—Temporal trend of age-adjusted human im¬ munodeficiency virus seroconversion by time on active duty. computer-generated reminder sent ev¬ ery other year during the birth month calling for a current HIV Ab test. Even though it is fairly easy to temporarily self-defer from testing, over the past 2 V2 years, a relatively constant propor¬ tion of the Army population (30% to 32%) has been out of compliance with current testing policy (ie, >24 months since the last HIV Ab test). Moreover, among soldiers who had gone longer than 24 months since their most recent negative test result, seroconversion rates and temporal trends were indis¬ tinguishable from those of soldiers who were last tested within 24 months. This similarity persisted in an analysis strat¬ ified by age, race/ethnicity, gender, and marital status. Among 15 657 soldiers on active duty when HIV Ab screening began in the Army who were not initial¬ ly tested until after November 1, 1988, HIV infection prevalences were virtual¬ ly identical to demographically similar soldiers who were initially tested be¬ tween November 1, 1985, and October 31, 1988. Although these data do not disprove that soldiers at high risk for HIV infection avoid screening, they at least suggest that self-deferral was not an overwhelming source of bias in this analysis. The size and demographic character¬ istics of the Army population and the method used to compute interval-spe¬ cific IDs can result in somewhat impre¬ cise rate estimates. Imprecision is espe¬ cially troublesome during interval 3, and parameter estimates during that period must be viewed with some cau¬ tion. Precision for interval 3 will im¬ prove dramatically when these data are reanalyzed in 1991. We previously reported that during the first 2 years of screening in the US Army, the HIV seroconversion rate was at least 0.74 per 1000 person10 years. The rate we currently report for that same period is substantially lower (0.49 per 1000 person-years). In our ini¬ tial analysis, conducted in December 1987, soldiers were required to have Downloaded from jama.ama-assn.org at Florida State University on November 14, 2011 > been tested for HIV Ab at least two times between November 1, 1985, and October 31,1987. In the present report, soldiers had to receive only a single neg¬ ative test result during interval 1 to contribute follow-up time to that inter¬ val. Soldiers who were tested two or more times within the first 2 years of Army screening, a period during which the goal was to test all soldiers only once, were at substantially higher over¬ all risk of infection than soldiers who were tested only once during that same period. Moreover, the different analyti¬ cal approaches taken in our original re¬ port and in the present analysis do not account for the observed trend of decel¬ erating infection incidence we currently report. If we apply our initial analytical criteria to the period from November 1987 through October 1989 (ie, two or more tests in this 2-year period), the resulting HIV seroconversion rate is 0.41 per 1000 person-years (107 seroconverters, 258 774 person-years of fol¬ low-up time). This rate is significantly lower (P<.001) than the rate (0.77 per 1000 person-years) we previously re¬ ported for the first 2 years of HlV Ab screening in the Army.10 The validity of the data presented in this analysis should not be confused with their generalizability. Seroconver¬ sion rates and trends reported herein do not necessarily reflect the course of the HIV infection epidemic outside the US Army. Institutional and self-selection factors probably result in an Army pop¬ ulation that imperfectly represents de¬ mographically similar individuals not serving in the Armed Forces. This may be especially true for certain behavioral and life-style characteristics, such as in¬ travenous drug use and homosexuality. In addition, most soldiers live and work in geographic locations that are some¬ what remote from epicenters of the HIV/AIDS epidemic, so potentially risky drug use and sex practices might not result in the same degree of infec¬ tion risk as when they occur in hyperendemic settings. Nonetheless, the HIV incidence data we currently report are remarkably similar to data in other populations, such as civilian applicants for military service,27 and are consistent with ob¬ served reductions in the rate of HIV incidence observed among homosexual men.28'29 Human immunodeficiency virus sero¬ conversions among soldiers have been detected at nearly every location that conducts screening worldwide. Because the active-duty population is highly mo¬ bile and relatively transient, serocon¬ versions diagnosed at a particular loca¬ tion do not necessarily imply that infection occurred there. Seroconver¬ sion rates among soldiers with current or overseas duty assignments consistently fallen within the recent have range of rates observed for soldiers as¬ signed to duty locations within the con¬ tinental United States. Moreover, the number of soldiers assigned outside the continental United States, as well as overall assignment patterns, remained constant over the 4 years surveyed. Since soldiers are not routinely tested for HIV Ab on reassignment in the United States, location-specific sero¬ conversion rates cannot be precisely measured. Detailed risk assessments of newly infected soldiers25 will better de¬ fine infection risk by location for the Army population. Tracking the incidence of HIV infec¬ tion over time within the US Army pre¬ sents a unique opportunity to evaluate the epidemic dynamics of HIV infec¬ tion. Follow-up of this population will continue for the foreseeable future. We acknowledge the contributions of Chet Rob¬ erts, PhD (Walter Reed Army Institute of Re¬ search), for assistance with analysis and interpreta¬ tion of serological specimens; Mitchell Gail, MD, of the National Cancer Institute, Bethesda, Md, for advice regarding the Poisson analysis; and Bill Ryan, PhD (Walter Reed Army Institute of Re¬ search), for providing US Army venereal disease data. References 1. HIV/AIDS Surveillance. Atlanta, Ga: Center for Infectious Diseases, Centers for Disease Control; 1983 to present. 2. Centers for Disease Control. Estimates of HIV prevalence and projected AIDS cases: summary of a workshop, October 31-November 1, 1989. MMWR. 1990;39:110-118. 3. Centers for Disease Control. Update: acquired immunodeficiency syndrome\p=m-\United States, 1989. MMWR. 1990;39;81-86. 4. Gail MH, Brookmeyer R. Methods for projecting course of acquired immunodeficiency syndrome epidemic. J Natl Cancer Inst. 1988;80:900-911. 5. Brookmeyer R, Damiano A. Statistical methods for short term projections of AIDS incidence. Stat Med. 1989;8:23-34. 6. Detels R, English P, Visscher BR, Doll LS, Jaffe HW, Rutherford GW. Seroconversion, sexual activity, and condom use among 2915 seronegative men followed for up to 2 years. J Acquir Immune Defic Syndr. 1989;2:77-83. 7. Hessol NA, Lifson AR, O'Malley PM, Doll LS, Jaffe HW, Rutherford GW. Prevalence, incidence, and progression of human immunodeficiency virus infection in homosexual and bisexual men in hepatitis B vaccine trials, 1978-1988. Am J Epidemiol. 1989;6:1167-1175. 8. Goedert JJ, Kessler CM, Aledort LM, et al. A prospective study of human immunodeficiency virus type 1 infection and the development of AIDS in subjects with hemophilia. N Engl J Med. 1989; 17:1141-1148. 9. Dodd RY, Connelly J, Cumming P. Incidence and prevalence of HIV infection in a low-risk population in the United States. In: Abstracts of the Fourth International Conference on AIDS. Stockholm, Sweden: Swedish Ministry of Health and Social Affairs; 1988;2:347. 10. McNeil JG, Brundage JF, Wann ZF, Burke DS, Miller RN, and the Walter Reed Retrovirus Research Group. Direct measurement of human immunodeficiency virus seroconversions in a serial- ly tested population of young adults in the United States Army, October 1985 to October 1987. N Engl J Med. 1989;320:1581-1585. 11. Garland FC, Mayers DL, Hickey TM, et al. Incidence of human immunodeficiency virus seroconversion in US Navy and Marine Corps personnel, 1986 through 1988. JAMA. 1989;262:3161\x=req-\ 3165. 12. Herbold JR. AIDS policy development within the Department of Defense. Milit Med. 1986;151: 623-630. 13. Kelley PW, Takafuji ET, Tramont EC, et al. The importance of HIV infection for the military. In: Wormser GP, Stahl RE, Bottone EJ, eds. AIDS and Other Manifestations of HIV Infection. Park Ridge, NJ: Noyes Publications; 1987:67-85. 14. Burke DS, Brundage JF, Herbold JR. Human immunodeficiency virus infection among civilian applicants for United States military service, October 1985 to March 1986. N Engl J Med. 1987; 317:132-136. 15. Petricciani JC. Licensed tests for antibody to human T-lymphotropic virus type III: sensitivity and specificity. Ann Intern Med. 1985;103:726-729. 16. Towbin H, Staehelin T, Gordon J. Electrophoretic transfer of proteins from polyacrylamide gels to nitrocellulose sheets: procedure and some applicant. ProcNatlAcadSci U S A. 1979;76:4350-4354. 17. Sarngadharan MG, Popovic M, Bruch L, Sch\l=u"\pbachJ, Gallo RC. Antibodies reactive with human T-lymphotropic retrovirus (HTLV-III) in the serum of patients with AIDS. Science. 1984;224:506-508. 18. Burke DS, Redfield RR. False-positive Western blot tests for antibodies to HTLV-III. JAMA. 1986;256:347. 19. Burke DS, Brundage JF, Redfield RR, et al. Measurement of the false positive rate in a screening program for human immunodeficiency virus infections. N Engl J Med. 1988;319:961-964. 20. Redfield RR, Wright DC, Tramont EC. The Walter Reed staging classification for HTLV-III/ LAV infection. N Engl J Med. 1986;314:131-132. 21. Burke DS. Laboratory diagnosis of human immunodeficiency virus infection. Clin Lab Med. 1989;9:369-392. 22. EGRET Version .25. Seattle, Wash: Statistics and Epidemiology Research Corp; 1990. 23. Breslow NE, Day NE. Statistical Methods in Cancer Research. Lyons, France: International Agency for Research on Cancer; 1987;2:133. 24. Kelley PW, Miller RN, Pomerantz R, Wann FW, Brundage JF, Burke DS. Human immunodeficiency virus seropositivity among members of the active duty US Army 1985-89. Am J Public Health. 1990;80:405-410. JG, Peterman TA, Renzullo PO, Lasley-Bibbs V, Levin LI, and the HIV Seroconverter 25. McNeil Risk Factor Study Group. Recent HIV infection in men in the Army: a case-control study. In: Abstracts of the Sixth International Conference on AIDS. San Francisco, Calif: University of California, San Francisco; 1990;1:265. 26. McNeil JG, Renzullo PO, Brundage JF. A retrospective follow-up study of morbidity among HIV infected young adults prior to a diagnosis of HIV infection: a 'true' natural history study. In: Abstracts of the Fifth International Conference on AIDS. Ottawa, Ontario: International Development Research Center; 1989;1:158. 27. Brundage JF, Burke DS, Gardner LI, et al. Tracking the spread of the HIV infection epidemic among young adults in the US: results of the first four years of screening among civilian applicants for US military service. J Acquir Immune Defic Syndr. 1990;3:1168-1180. 28. Winkelstein W Jr, Wiley JA, Padian NS, et al. The San Francisco Men's Health Study: continued decline in HIV seroconversion rates among homosexual/bisexual men. Am J Public Health. 1988; 78:1472-1474. 29. Van Griensven GJP, De Vroome EMM, Goudsmit J, Coutinho RA. Changes in sexual behavior and the fall in incidence of HIV infection among homosexual men. BMJ. 1989;298:218-221. Downloaded from jama.ama-assn.org at Florida State University on November 14, 2011
© Copyright 2026 Paperzz