The development of type-specific serological tests has greatly aided studies of the descriptive epidemiology of herpes simplex virus type 1 (HSV-1) and type 2 (HSV-2) infection. In recent years, commercially available type-specific enzyme-linked immunosorbent assay (ELISA) has proven to be highly sensitive and specific as well as an economical and practical method to determine HSV-1 and HSV-2 serostatus in larger samples. Since the early 1990s, a number of studies have been conducted in various settings to determine HSV-1 and HSV-2 seroprevalence among adults and young adults in the United States and worldwide.1,2
Malkin2 notes, however, that despite the National Health and Nutrition Examination Surveys (NHANES) national prevalence data, there are no national incidence data in the United States. Instead, the national incidence of HSV infection has been estimated using statistical modeling of data from different subjects in different time periods.3 The disadvantage of such estimates, however, is that the generally used mathematical model estimates a linear increase in incidence with time, assuming that the shape of age-specific incidence curve remains constant. Although studies with serial specimens avoid such statistical limitation, they are relatively rare and have been typically limited to special populations,4–7 including studies of adolescent women followed for the acquisition of sexually transmitted disease,8–11 studies of pregnant women,12,13 overseas seroincidence studies of cohorts of factory workers,14,15 as well as vaccine16 and clinical trials.17,18
We had access to a convenience sample of young, geographically diverse, active-duty US personnel with serial serum specimens archived in the Department of Defense (DoD) Serum Repository, which, approximately, since 1990 contains at least one specimen for every military member. Using these data, we analyzed these serologic data to estimate HSV-1 and HSV-2 seroincidence among healthy young, military adults, as well as to identify demographic characteristics associated with seroconversion. We further determined the sex-, age-, and race-specific seroprevalences of HSV-1 and HSV-2, adjusted to the US Census 2000 population, to compare these estimates with national seroprevalence data from NHANES.19,20
MATERIALS AND METHODS
Military Sample Study Design
The study population was composed of active-duty service members who were originally selected as controls for the Military New Onset Psychosis Project (MNOPP), a series of nested case-control studies exploring associations of schizophrenia and bipolar disorder with selected biomarkers in US military personnel.21 We included all controls from the bipolar disorder study based on the availability of HSV-1 and HSV-2 serologic test data, except that we excluded those aged 30 years or older because of small numbers, producing a cohort of 1094 military personnel aged 18 to 30 years who served between 1989 and 2005. In the MNOPP study, controls were military personnel randomly selected from the Defense Military Surveillance System (DMSS), an executive information system operated by the Armed Forces Health Surveillance Center, whose database contains up-to-date and historical data on diseases and medical events (e.g., hospitalizations, ambulatory visits, reportable diseases, HIV tests, acute respiratory diseases, and health risk appraisals) and longitudinal data on personnel and deployments.21 Through the Defense Military Surveillance System, the Armed Forces Health Surveillance Center provides the sole link between the DoD Serum Repository and other databases. This repository contains over 50 million frozen serum specimens and is the largest of its kind in the world. Controls were matched to bipolar disorder cases on selected demographic characteristics (e.g., age, sex, and race) as well as availability of serum specimens drawn close in time to the case samples. Data elements included in our analyses were age, sex, race/ethnicity, era of military service, family income, and region of residence (the last 2 coming from zip code data; see statistical analysis section below for details). For comparison to published National Health and Nutrition Examination Survey (NHANES) data from 2 survey periods, we divided our HSV1/2 seroprevalence estimates into 2 similar periods, 1989–1998 and 1999–2005. Information was not available on history of prior sexually transmitted infections (STIs), other medical conditions, sexual behavior, or marital status.
Serum Retrieval From DoD Serum Repository
Controls had 2 specimens retrieved from the DoD Serum Repository, drawn between 1989 and 2005. The average time between the first and second blood draw was 40.7 months (standard deviation ± 33.1 month). Procedures to catalogue and store serum samples archived in the DoD Serum Repository have been described in detail previously.22 Samples obtained from the repository were kept frozen at (−30°C) until tested.
HSV-1 and HSV-2 serum antibodies were detected using a commercially available indirect ELISA (Herpes Select, Focus Technologies, Cypress, CA), shown to have very high sensitivities (91.2% and 96.1%, respectively) as well as relatively high specificities (92.3% and 97.0%, respectively).23 The ELISA consists of binding serum to solid phase antigens (gG-1 antigen for HSV-1 and gG-2 antigen for HSV-2) and subsequent reactions with enzyme-labeled antihuman IgG and enzyme substrate. The opacity generated by the enzyme-substrate reaction was measured in optical density (OD) units by means of a microplate colorimeter. Seropositivity was defined by comparison with the standard sera following the manufacturer's protocol: negative <0.9; positive >1.1; and equivocal ≥0.9 and ≤1.1.
Seroprevalence was defined as the number of individuals who tested positive for HSV-1 or HSV-2 antibody on the initial serum sample, divided by the number tested (excluding equivocal results, which numbered only 23 initial specimens and 20 second specimens for HSV-1, and 7 initial specimens and 14 second specimens for HSV-2; none of these excluded specimens were retested. This yielded prevalence sample sizes of 1071 (i.e., 1094 − 23 = 1071) for HSV-1 and 1087 (i.e., 1094 − 7) for HSV-2. Further, only individuals with an initial seronegative sample (susceptibles) were used for the seroincidence calculations. The seroincidence rate was defined as the number of seroconversions divided by person-years of observation among the susceptible (i.e., initially seronegative) population. Date of seroconversion was assumed to be having occurred midway between the time of the initial negative and subsequent positive specimens, as the actual date of seroconversion was not known. Among seroconverters, person-years of observation time accrued from the first test until the assumed date of seroconversion. For those who did not seroconvert, person-years of observation were defined as the total time between the first sample and the second sample. There was an extremely wide range of times between initial and subsequent specimens (from 3 days to 14 years); therefore, we calculated seroincidence rates using only individuals whose time between specimens was 3 months to 5 years. This approach allowed for the necessary time to develop IgG antibodies in response to a newly acquired infection and minimized the influence of outliers on the seroincidence rates.
Data were analyzed using SAS software, version 9.2 (Cary, NC). As the study population for this investigation was generated for a different purpose,21 it was not representative of the US military active-duty population or the US young adult population. We therefore standardized seroprevalences and seroincidence rates to the US population, using weighted survey descriptive analyses (see Statistical Appendix for details), with sampling weights calculated from the age, race, and sex distributions of the US population.24 The data used to calculate these weights came from the 2000 US census,25,26 and we recalculated some weights for sparse strata (see Statistical Appendix for details).
We linked the zip code of an individual's home of record to 2000 US Census data to determine the region of the country from which the individual entered military service. This allowed us to obtain and control for premilitary entry factors that might have affected HSV infection rates, such as geographical region (a proxy for the rate of “background” infection) and income, a general proxy for social class.27 Income data were divided into categories of “low” and “high” using the sample median as the cut-point ($38,000). Era of service was defined based on the MNOPP study reference date,21 generally within 2 years of the date of first specimen. Covariates in the logistic regression analyses included age, race, sex, era of service, region, and income. Weighted survey logistic regression analyses used HSV-1 or HSV-2 seroconversion as their outcomes. This study was approved by the Walter Reed Army Institute of Research Institutional Review Board.
Military Personnel Characteristics and HSV-1 and HSV-2 Seroprevalences
Basic demographic characteristics of military personnel in our sample along with crude seroprevalence estimates for HSV-1 and HSV-2 and estimates weighted to the US population are shown in Tables 1 and 2. Compared with the US population of age 18 to 29 years, our sample had relatively more 18 to 19 year olds and fewer 25 to 29 year olds. For example, in the US population of 18 to 29 years, white males make up 8.2%, 30.2%, and 16.6%, respectively, in the 3 age groups; for our sample, the corresponding figures are 23.2%, 23.5%, and 5.4%. For white females, the US population percentages in the 3 age groups are 1.4%, 4.5%, and 2.2% olds versus 7.4%, 8.5%, and 1.9%, respectively (see Statistical Appendix, Table A1, for additional data). All seroprevalence estimates are based on initial serum samples, drawn either before or at military enlistment, with equivocal results omitted from the calculations. Our estimated HSV1/2 seroprevalences of 40.7% (95% CI: 37.8%–43.7%) for HSV-1 and 7.5% (95% CI: 6.0%–9.1%) were comparable to, albeit generally lower than, US national data provided by NHANES serosurveys; the overall sample HSV-1/HSV-2 coinfection rate (data not shown) was 4.8%.
Comparison of HSV-1 and HSV-2 Seroprevalence Estimates With Prevalence Estimates Provided by NHANES Serosurveys 1988–1994 and 1999–2004
Table 3 shows HSV-1/2 national seroprevalence estimates that were derived from the military sample and national estimates from NHANES20 using the published NHANES age, sex, and race categories. For HSV-1, many of our seroprevalence estimates decreased little, if at all, and some actually increased between the early and late periods, compared with the uniform decline in early versus late seroprevalence estimates from NHANES; for HSV-2, our seroprevalence estimates did decrease uniformly between early and late periods, as did the estimates from NHANES. In addition, our HSV-1 seroprevalence estimates for non-Hispanic blacks and for Hispanics are notably lower than the NHANES estimates, and there are no statistically significant differences in HSV-2 prevalences, although the confidence limits for our estimates are fairly large, probably due to small sample size.
National Seroincidence Rates by Demographic Characteristics for HSV-1 and HSV-2
Table 4 shows sample and national estimates of seroincidence rates for HSV-1 and HSV-2 per 100 person-years, by several demographic characteristics, except that there were too few data to analyze seroconversion rates for Hispanics and Northeast region for HSV-2. These calculations were based on 407 susceptibles for HSV-1 and 650 susceptibles for HSV-2. The overall rate of HSV-1 seroconversion was 9.1 per 100 person-years (CI: 4.6–13.5), whereas the rate of HSV-2 seroconversion was 6.2 per 100 person-years (CI: 3.1–9.3). As noted in the methods, these estimates were all calculated by restricting the analysis to individuals with a follow-up period from 3 months (n = 348) to 5 years (n = 20). There were no statistically significant differences in adjusted HSV-1 or HSV-2 annualized rates by any of the individual demographic characteristics, including family income (based on zip code) and region of country (data not shown). The seroreversion rate for both HSV-1 and HSV-2 was approximately 1%.
Risk Factors for HSV-1 and HSV-2 Seroconversion
Table 5 shows the multivariable analysis results of seroconversion rates for all risk factors. For HSV-1, there were no significant associations, except for era of service. For HSV-2, we found a nearly 4-fold higher odds ratio for female subjects compared with males and a significant 4-fold higher odds ratio for non-Hispanic blacks compared with non-Hispanic whites. Neither family income (based on zip code) nor region of country were significantly associated with seroincidence (data not shown).
To generate national estimates of HSV-1 and HSV-2 seroprevalence and seroincidence for the young adult US military population, we used data from a selected group of active-duty military personnel (aged 18–29 years at accession) who had archived serial serum samples from the DoD Serum Repository. Our military sample was largely composed of young (18–24 years old) adult males and was mostly non-Hispanic white. For HSV-1, the overall national seroincidence rate was 9.1 per 100 person-years (95% CI: 4.6–13.5); for HSV-2, the rate was 6.2 (95% CI: 3.1–9.3). We found no statistically significant differences in HSV-1 seroincidence rates by demographic characteristics, whereas females and non-Hispanic blacks had significantly higher HSV-2 seroincidence rates.
Our US 2000 Census-weighted national estimates of HSV-1 seroincidence are unique, so far as we know, as they are population-based and have been derived from a study population with serial serum samples. Other published seroincidence data are, however, available from other, selected populations. Cherpes et al6 found an HSV-2 seroincidence rate of 5.1 per 100 per 100 woman-years among sexually active women 18 to 30 years of age, whereas Stanberry et al4 found a rate of HSV-2 incidence of 4.4 per 100 person-years among young adolescent girls aged 12 to 15 years; the 95% confidence limits around our national estimate for young females (5.2–18.7 per 100 person-years) barely exclude both of these other point estimates. The HSV-2 incidence rates among homeless adolescent (female mean age 17.7 years; male mean age 18.8 years) found by Noell et al5—11.7% for males and 25.4% for females—were somewhat higher than ours. Risk factors for HSV-2 seroincidence in these same studies included female sex,5,15,16,18 black race,6,10,18 and age.7–9,15 Compared with the seroprevalence data from these same studies, our HSV-1 and HSV-2 crude seroprevalence estimates (40.7% and 7.5%, respectively) are low. We note that although our sample was, strictly speaking, a convenience sample, it is much more demographically and geographically broad and diverse than samples taken from other “special populations,” such as sexually transmitted disease or pregnancy clinics in a particular city. Moreover, use of weights from the 2000 US Census ensures that our weighted estimates exactly reflect the distributions of race, sex, and age in the 18- to 29-year-old US population.
Our HSV-1 prevalence estimates were fairly stable over time in contrast to the NHANES estimates. HSV-2 seroprevalence in both our US population estimates and NHANES data showed decreases in prevalence over time. Overall, our seroprevalence estimates are not statistically different from those of NHANES, except for some demographic subgroups (e.g., HSV prevalence for non-Hispanic blacks in the “late” era).
This study has several limitations. The study population used to generate national HSV1/2 estimates in young adults was a nonrandom sample of active-duty service members and was mostly white males less than age 30 years with a range of military service from 1989 to 2005. Although we limited our seroincidence analyses to those individuals whose time between samples was 3 months to 5 years, there may still be some inherent bias resulting from such a wide range of follow-up periods. It is not a true representative, population sample, and therefore our statistical analyses adjusted for any differences in the age, race, and sex distribution between our sample and the US population. There were, moreover, relatively small numbers of seroconverters, which undoubtedly limited our ability to find statistically significant differences across demographic groups. Preinduction factors that might influence seroprevalence include the fact that military personnel must pass a general medical screening before entry into military service, nearly all are high school graduates, and very few have been convicted of a serious crime; such factors might argue for lower HSV-1/2 infection rates. Additionally, military risk factors could influence seroincidences. For example, HSV-1 seroincidence could be higher than in the comparable civilian population, given the potentially close living conditions of military personnel. Because HSV-2 seroincidence is highly correlated with sexual behavior, HSV-2 seroprevalence may be a surrogate marker of sexual behavior patterns. High prevalence of STIs has been documented in military populations, and studies on STI risk factors have found that US military personnel, especially deployed personnel, frequently engage in high-risk sexual behavior.28–30 Unfortunately, we were unable to examine directly behavioral correlates of HSV1/2 infection, as reliable data were unavailable. In sum, it is not known exactly how the above differences could affect HSV seroprevalence and seroincidence estimates, but the comparison between our estimates and earlier NHANES data suggest that, at the least, HSV1/2 prevalence rates are lower in the military.
Because laboratory tests with even relatively high specificity and sensitivity can generate sufficient numbers of false-positive and false-negative results, which could affect seroincidence estimates, we calculated these potential effects in a simulated sample of 1000 subjects. For an HSV-1 test with reported prevalence of 40%, sensitivity of 91.2%, and specificity of 92.3%, the true prevalence would have been 41.1%. A reported seroincidence of 10% led to a true seroincidence of 16.6%. For HSV-2, a reported prevalence of 8% led to a true prevalence of 10.5%, whereas a reported seroincidence of 5% led to a true seroincidence of 7.7%. Thus, the estimates we calculated based on lab results all underestimate somewhat both the true seroprevalence and seroincidence of HSV-1 and HSV-2.
In summary, we believe that our national estimates of HSV-1 and HSV-2 seroincidence in young adults provide valuable and perhaps unique population data on young adults, taking particular advantage of the availability of a large number of serial serum specimens from the DoD Serum Repository. Our seroprevalence and seroincidence estimates for HSV-1 and HSV-2 were, as expected, generally lower than estimates reported in other studies from selected samples of high-risk groups and comparable with those from NHANES serosurveys. In addition, the multivariable analyses of demographic characteristics associated with HSV-1 and HSV-2 seroconversion identified demographic risk factors (female sex and black race for HSV-2) that are similar to those found in other seroincidence studies.
Our findings can be used to inform the planning of HSV-1 and HSV-2 prevention measures in the general young adult US population. While sexually transmitted illnesses have been considered a particularly challenging problem to military leaders,29 our national estimates do not suggest that HSV infection rates are any higher in the young adult military population than among civilians. Moreover, intervention programs to control STIs in the military,29 such as ease of access to STI screening and annual periodic health assessments that include signs and symptoms of STI, may be directly applicable or easily modifiable for use among US young adult civilian groups.
1. Smith JS, Robinson NJ. Age-specific prevalence of infection with herpes simplex virus types 2 and 1: A global review. J Infect Dis 2002; 186(suppl):S3–S28.
2. Malkin JE. Epidemiology of genital herpes simplex virus infection in developed countries. Herpes 2004; 11(suppl 1):2A–23A.
3. Armstrong GL, Schillinger J, Markowitz L, et al.. Incidence of herpes simplex virus 2 infection in the United States. Am J Epidemiol 2001; 153:912–920.
4. Stanberry LE, Rosenthal SL, Mills L, et al.. Longitudinal risk of herpes simplex virus (HSV) type 1, HSV type 2, and cytomegalovirus infections among young adolescent girls Clin Infect Dis 2004; 39:1433–1438.
5. Noell J, Rohde J, Ochs L, et al.. Incidence and prevalence of Chlamydia, herpes, and viral hepatitis in a homeless adolescent population. Sex Transm Dis 2001; 28:4–10.
6. Cherpes TL, Meyn LA, Krohn MA, et al.. Association between acquisition of herpes simplex virus type 2 in women and bacterial vaginosis. Clin Infect Dis 2003; 37:319–325.
7. Dickson N, Van Roode T, Herbison P, et al.. Risk of herpes simplex virus type 2 acquisition increases over early adulthood: Evidence from a cohort study. Sex Transm Infect 2006; 83:87–90.
8. Fife KH, Fortenberry JD, Ofner S, et al.. Incidence and prevalence of herpes simplex virus infections in adolescent women. Sex Transm Dis 2006; 33:441–444.
9. Gallo MF, Warner L, Macaluso M, et al.. Risk factors for incident herpes simplex type 2 virus infection among women attending a sexually transmitted disease clinic. Sex Transm Dis 2008; 35:679–685.
10. Moss NJ, Harper CC, Ahrens K, et al.. Predictors of incident herpes simplex virus type 2 infections in young women at risk for unintended pregnancy in San Francisco. BMC Infect Dis 2007; 7:113.
11. Brown JM, Wald A, Hubbard A, et al.. Incident and prevalent herpes simplex virus type 2 infection increases risk of HIV acquisition among women in Uganda and Zimbabwe. AIDS 2007; 21:1515–1523.
12. Brown ZA, Selke S, Zeh J, et al.. The acquisition of herpes simplex virus during pregnancy. N Engl J Med 1997; 337:509–515.
13. Mindel A, Taylor J, Tideman RL, et al.. Neonatal herpes prevention: a minor public health problem is some communities. Sex Transm Infect 2000; 76:287–291.
14. McFarland W, Gwanzura L, Bassett MT, et al.. Prevalence and incidence of herpes simplex virus type 2 infection among male Zimbabwean factory workers. J Infect Dis 1999; 180:1459–1465.
15. Kebede Y, Dorigo-Zetsma W, Mengistu Y, et al.. Transmission of herpes simplex virus type 2 among factory workers in Ethiopia. J Infect Dis 2004; 190:365–372.
16. Langenberg AG, Corey L, Ashley RL, et al.. A prospective study of new infections with herpes simplex virus type 1 and type 2. Chiron HSV Vaccine Study Group. N Engl J Med 1999; 341:1432–1438.
17. Tobian AA, Ssempijja V, Kigozi G, et al.. Incident HIV and herpes simplex virus type 2 infection among men in Rakai, Uganda. AIDS 2009; 23:1589–1594.
18. Gottlieb SL, Douglas JM, Foster M, et al.. Incidence of herpes simplex virus type 2 infection in 5 sexually transmitted disease (STD) clinics and the effect of HIV/STD risk-reduction counseling. J Infect Dis 2004; 190:1059–1067.
19. Xu F, Schillinger JA, Sternberg MR, et al.. Seroprevalence and coinfection with herpes simplex virus type 1 and type 2 in the United States, 1988–1994. J Infect Dis 2002; 185:1019–1024.
20. Xu F, Sternberg MR, Kottiri BJ, et al.. Trends in herpes simplex virus type 1 and type 2 seroprevalence in the United States. JAMA 2006; 296:964–973.
21. Niebuhr DW, Millikan AM, Cowan DN, et al.. Selected Infectious Agents and Risk of Schizophrenia Among U.S. Military Personnel Am J Psychiatry 2008; 165:99–106.
22. Rubertone MV, Brundage JF. The Defense Medical Surveillance System and the Department of Defense serum repository: glimpses of the future of public health surveillance. Am J Public Health 2002; 92:1900–1904.
23. HerpeSelect 2 ELISA IgG 2005. Cypress, CA: Focus Technologies; 2005.
24. Chambers RL, Skinner CJ. Analysis of Survey Data. Chapter 14. Chichester, United Kingdom: Wiley, 2003:205–219.
27. United States Zip Code Database Software. 2009. Available at: http://www.zip-codes.com/zip-code-database.asp
28. Cowan FM, Johnson AM, Ashley R, et al.. Antibody to herpes simplex virus type 2 as serological marker of sexual lifestyle in populations. BMJ 1994; 309:1325–1329.
29. Gaydos CA, Quinn TC, Gaydos JC. The challenge of sexually transmitted diseases for the military: What has changed? Clin Infect Dis 2000; 30:719–722.
30. Niebuhr D, Tobler SK, Jordan NN, et al.. Sexually transmitted infections among military recruits. In: Textbook of Military Medicine: Recruit Medicine. Office of the Surgeon General, Department of the Army, The Borden Institute, 2007:253–273.
Statistical Appendix for National Estimates of Seroincidence and Seroprevalence for Herpes Simplex Virus Type 1 and Type 2 Among US Military Adults Aged 18–29 Years
Detailed Material on Statistical Methods
Calculating weighted estimates.
Typically, when we conduct a survey, it is important to have a representative population sample. But sometimes, there is an oversample or undersample of some types of people. This type of sample will introduce bias into estimates. With regard to HSV, the study data are a convenience sample and do not truly represent the proportionality between population and sample in each sex, race, and age strata. We can correct for these biases mathematically with poststratification weights. To calculate a poststratification weight, we need an auxiliary data set about the population to which we can compare our sample data. We obtained the true proportionalities among all subpopulations of sex, race, and age strata from 2000 US Census. The poststratification weights are then computed by dividing those sample proportions from the population proportions.24 Based on the previous 2 reasons, we decided to use survey procedures (PROC SURVEYFREQ, PROC SURVEYLOGISTIC) to estimate the seroincidence and seroprevalence of HSV-1 and HSV-2. SURVEYLOGISTIC fits linear logistic regression models for discrete response survey data by the method of maximum likelihood. In the analyses, this procedure incorporates complex survey sample designs, including designs with stratification and unequal weighting. As the descriptive analysis, we also used the PROC SURVEYFREQ, which can be used to produce one-way to n-way frequency and cross tabulation tables from complex multistage survey designs with stratification and unequal weighting.
Calculating weights in sparse cells.
The poststratification weights set the sample proportions in the denominators, so if a stratum has no event, which would result in a zero proportion, then we will get an undefined weight for that particular strata. There are 2 types of zero cells in a contingency table: structural and sampling. A structural zero cell has an expected value of zero, whereas a sampling zero cell can have nonzero expected value and can be estimable. In our sample data, we had no seroincidence in some strata. We know our zero cells were caused by sampling. To calculate their poststratification weights, we fit a log-linear model to the response frequencies (using PROC CATMOD), which uses maximum likelihood estimation of parameters for the model to predict the values of frequencies within each strata. These predicted frequencies are then used to produce the sample proportions within each stratum.
Appendix Table A1 compares the age, sex, and race distribution of the 18- to 29-year-old US population, based on Census figures, with the age, sex, and race distribution of the military sample we used in our analyses. Tables A2 and A3 show calculations and report true prevalence and incidence rates for HSV-1 and HSV-2, respectively. Starting with an assumed sample of 1000 persons and reported prevalence and incidence rates of 40% and 10% for HSV-1 and 8% and 5% for HSV-2, we use the manufacturer's reported sensitivity and sensitivity values to calculate the true prevalence and incidence values, compared to the reported ones.