Using graph learning to understand adverse pregnancy outcomes and stress pathways.

To identify pathways between stress indicators and adverse pregnancy outcomes, we applied a nonparametric graph-learning algorithm, PC-KCI, to data from an observational prospective cohort study. The Measurement of Maternal Stress study (MOMS) followed 744 women with a singleton intrauterine pregnancy recruited between June 2013 and May 2015. Infant adverse pregnancy outcomes were prematurity (<37 weeks' gestation), infant days spent in hospital after birth, and being small for gestational age (percentile gestational weight at birth). Maternal adverse pregnancy outcomes were pre-eclampsia, gestational diabetes, and gestational hypertension. PC-KCI replicated well-established pathways, such as the relationship between gestational weeks and preterm premature rupture of membranes. PC-KCI also identified previously unobserved pathways to adverse pregnancy outcomes, including 1) a link between hair cortisol levels (at 12-21 weeks of pregnancy) and pre-eclampsia; 2) two pathways to preterm birth depending on race, with one linking Hispanic race, pre-gestational diabetes and gestational weeks, and a second pathway linking black race, hair cortisol, preeclampsia, and gestational weeks; and 3) a relationship between maternal childhood trauma, perceived social stress in adulthood, and low weight for gestational age. Our approach confirmed previous findings and identified previously unobserved pathways to adverse pregnancy outcomes. It presents a method for a global assessment of a clinical problem for further study of possible causal pathways.

Age, Race, and At-Risk Drinking in an HIV-infected U.S. Military Cohort.

INTRODUCTION: There is a high prevalence of at-risk drinking in the U.S. military. Among HIV-infected individuals, alcohol abuse confers additional risk for adverse health outcomes. In the military, however, the characteristics of HIV-infected individuals who engage in high-risk drinking are not well defined. The purpose of this study was to assess risk factors associated with at-risk drinking in an HIV-positive longitudinal cohort of DoD beneficiaries. MATERIALS AND METHODS: Annual prevalence of at-risk drinking was calculated for members of the U.S. Military HIV Natural History Study who initiated highly active antiretroviral therapy (HAART) during or after January 2006 through May 2014; each participant completed at least one self-reported alcohol survey within a year of HAART initiation. Univariate and multivariable logistic regression was used to analyze factors associated with at-risk drinking. RESULTS: Sixty-six percent of subjects (495/752) reported at-risk drinking on at least one survey after HAART initiation. At-risk drinkers were more likely to be Active Duty compared to Retired (OR 0.65 95% CI [0.46, 0.92]). In multivariate models, Caucasian race (OR 3.30 95% CI [2.31, 4.71]); Hispanic/other race (OR 2.17 95% CI [1.51, 3.14]) and younger age (OR 0.61 per 10 years older, [95%CI 0.49, 0.75]) were significantly associated with at-risk drinking. Single relationship status (OR 1.51 95% CI [1.08, 2.13]) was also associated with at-risk drinking. CONCLUSIONS: Consistent with general alcohol consumption patterns in the military, we found a high prevalence of at-risk drinking among individuals with HIV infection, which was associated most closely with young, non-African Americans. Targeting interventions toward this group will be important to reduce at-risk drinking and its potential for HIV-related complications.

The Association between Sexually Transmitted Infections, Length of Service and Other Demographic Factors in the U.S. Military.

BACKGROUND: Numerous studies have found higher rates of sexually transmitted infections (STIs) among military personnel than the general population, but the cumulative risk of acquiring STIs throughout an individual's military career has not been described. METHODS: Using ICD-9 diagnosis codes, we analyzed the medical records of 100,005 individuals from all service branches, divided in equal cohorts (n = 6,667) between 1997 and 2011. As women receive frequent STI screening compared to men, these groups were analyzed separately. Incidence rates were calculated for pathogen-specific STIs along with syndromic diagnoses. Descriptive statistics were used to characterize the individuals within each accession year cohort; repeat infections were censored. RESULTS: The total sample included 29,010 females and 70,995 males. The STI incidence rates (per 100 person-years) for women and men, respectively, were as follows: chlamydia (3.5 and 0.7), gonorrhea (1.1 and 0.4), HIV (0.04 and 0.07) and syphilis (0.14 and 0.15). During the study period, 22% of women and 3.3% of men received a pathogen-specific STI diagnosis; inclusion of syndromic diagnoses increased STI prevalence to 41% and 5.5%, respectively. In multivariate analyses, factors associated with etiologic and syndromic STIs among women included African American race, younger age and fewer years of education. In the overall sample, increasing number of years of service was associated with an increased likelihood of an STI diagnosis (p<0.001 for trend). CONCLUSION: In this survey of military personnel, we found very high rates of STI acquisition throughout military service, especially among women, demonstrating that STI-related risk is significant and ongoing throughout military service. Lower STI incidence rates among men may represent under-diagnosis and demonstrate a need for enhancing male-directed screening and diagnostic interventions.

Factors associated with 10 years of continuous viral load suppression on HAART.

BACKGROUND: The principal goal of HAART is sustained viral load (VL) suppression resulting in immune reconstitution and improved HIV outcomes. We studied the factors associated with 10 years of continuous VL suppression on HAART in the US Military HIV Natural History Study. METHODS: Participants with continuous VL suppression (CS, n = 149) were compared to those who did not have continuous viral load suppression (NCS, n = 127) for ≥10 years on HAART. Factors associated with >10 years of VL suppression were evaluated by multivariate logistic regression. Additionally, association between CS and CD4 reconstitution was analyzed with a mixed effects model. RESULTS: Compared to NCS participants, a lower proportion of CS participants started HAART in the early HAART era (66 vs 90 %, for years 1996-1999; p < 0.001) and had less antiretroviral use prior to HAART (37 vs 83 %; p < 0.001). At initial HAART, the median CD4 cell count was higher and VL was lower for CS compared to NCS participants (375 cells/uL [256, 499] vs 261 cells/uL [146, 400]; p < 0.001 and 4.4 log10 copies/mL [3.5, 4.9] vs 4.5 log10 copies/mL [3.8, 5.0]; p = 0.048, respectively). New AIDS events were lower during HAART (5 vs 13 %; p = 0.032) and post-HAART CD4 trajectories were greater for the CS compared to NCS group. Factors negatively associated with ≥10 years of VL suppression included log10 VL at first HAART (OR 0.61, 95 % CI 0.4, 0.92; p = 0.020) and antiretroviral use prior to HAART (OR 0.16, 95 % CI 0.06, 0.38; p < .001). CONCLUSIONS: Sustained VL suppression is a key to long-term health in HIV-infected patients, as demonstrated by the lower proportion of AIDS events observed 10 years after HAART initiation. The current use of more potent and well-tolerated regimens may mitigate the negative factors of pre-HAART VL and prior ARV use encountered by treatment initiated in the early HAART era.

Characterizing the Association Between Alcohol and HIV Virologic Failure in a Military Cohort on Antiretroviral Therapy.

BACKGROUND: The effects of at-risk drinking on HIV infection remain controversial. We investigated the impact of self-reported alcohol consumption on surrogate markers of HIV progression among individuals initiated on highly active antiretroviral therapy (HAART). METHODS: We analyzed individuals who were surveyed on alcohol use within a year of HAART initiation between 2006 and 2014. At-risk drinking was defined as consumption of at least 3 or 4 drinks/d, or 7 and 14 drinks/wk among women and men, respectively. We performed time-updated generalized estimating equation logistic regression to determine the effect of at-risk drinking on virologic failure (VF) and mixed-effects linear regression on CD4 count reconstitution, controlling for potential confounders. RESULTS: Of 801 individuals initiated on HAART, 752 individuals with alcohol survey data were included in the analysis. Of these, 45% (n = 336) met criteria for at-risk drinking at HAART initiation on at least 1 survey. The rates of VF were 4.30 per 100 person-years (95% CI [2.86, 6.21]) for at-risk drinkers and 2.45 per 100 person-years (95% CI [1.57, 3.65]) for individuals without at-risk drinking. At-risk drinking was not significantly associated with VF (OR 1.73, 95% CI [0.92, 3.25]) (p = 0.087) or CD4 reconstitution (CD4 increase 11.4; 95% CI [-19.8, 42.7]) in univariate analyses; however, in our multivariate model, a statistically significant relationship between VF and at-risk drinking was observed (OR 2.28, 95% CI [ 1.01, 5.15]). CONCLUSIONS: We found a high proportion of at-risk drinking in our military cohort, which was predictive of VF in multivariate analysis. Given alcohol's effect on myriad HIV and non-HIV outcomes, interventions to decrease the prevalence of at-risk drinking among HIV-infected individuals are warranted.

Association between hepatitis B vaccine antibody response and CD4 reconstitution after initiation of combination antiretroviral therapy in HIV-infected persons.

BACKGROUND: Hepatitis B virus (HBV) vaccine antibody response has been associated with reduced risk of AIDS or death. However, it is unknown whether HBV vaccine responsiveness is associated with improved immune reconstitution during treatment with combination antiretroviral therapy (cART). We evaluated the relationship between HBV vaccine response status and CD4 reconstitution on cART in the U.S Military HIV Natural History Study. METHODS: Participants with viral load <400 copies/mL within 1 year on initial cART and documented HBV vaccination and surface antibody (anti-HBs) prior to cART were included. Participants were characterized as HBV vaccine responders (anti-HBs ≥10 IU/L) or non-responders (<10 IU/L) and further divided into 2 groups based on vaccine administration before or after HIV diagnosis. Linear mixed regression was used to model CD4 reconstitution during the first year of cART. RESULTS: Of the 307 and 169 participants vaccinated before or after HIV diagnosis, HBV vaccine response occurred in 288 (94%) and 74 (44%), respectively. For those vaccinated before HIV diagnosis, CD4 counts increased by a median 190 [IQR 99-310] cells/mm(3) for responders and 186 [IQR 116-366] cells/mm(3) for non-responders during the first year (P = 0.684). Participants vaccinated after HIV diagnosis had median increases of 185 [IQR 76-270] and 143 [IQR 47-238] cells/mm(3) for responders and non-responders, respectively (P = 0.134). In contrast to those with CD4 > 350 cells/mm(3) at cART initiation, participants with CD4 < 200 and 200-350 cells/mm(3) had significantly reduced CD4 gains in both groups by longitudinal mixed models, but there was no difference in CD4 recovery according to HBV vaccine seroresponse. CONCLUSIONS: Although HBV vaccine responsiveness is associated with a reduction in HIV disease progression, HBV vaccine responders do not achieve greater CD4 gains during the first year of cART. Additional clinical markers are needed to predict the magnitude of post-cART immune recovery.

A single dose of benzathine penicillin G is as effective as multiple doses of benzathine penicillin G for the treatment of HIV-infected persons with early syphilis.

BACKGROUND: Treatment guidelines recommend the use of a single dose of benzathine penicillin G (BPG) for treating early syphilis in human immunodeficiency virus (HIV)-infected persons. However, data supporting this recommendation are limited. We examined the efficacy of single-dose BPG in the US Military HIV Natural History Study. METHODS: Subjects were included if they met serologic criteria for syphilis (ie, a positive nontreponemal test [NTr] confirmed by treponemal testing). Response to treatment was assessed at 13 months and was defined by a ≥4-fold decline in NTr titer. Multivariate Cox proportional hazard regression models were utilized to examine factors associated with treatment response. RESULTS: Three hundred fifty subjects (99% male) contributed 478 cases. Three hundred ninety-three cases were treated exclusively with BPG (141 with 1 dose of BPG). Treatment response was the same among those receiving 1 or >1 dose of BPG (92%). In a multivariate analysis, older age (hazard ratio [HR], 0.82 per 10-year increase; 95% confidence interval [CI], .73-.93) was associated with delayed response to treatment. Higher pretreatment titers (reference NTr titer <1:64; HR, 1.94 [95% CI, 1.58-2.39]) and CD4 counts (HR, 1.07 for every 100-cell increase [95% CI, 1.01-1.12]) were associated with a faster response to treatment. Response was not affected by the number of BPG doses received (reference, 1 dose of BPG; HR, 1.11 [95% CI, .89-1.4]). CONCLUSIONS: In this cohort, additional BPG doses did not affect treatment response. Our data support the current recommendations for the use of a single dose of BPG to treat HIV-infected persons with early syphilis.

Hepatitis B vaccine responsiveness and clinical outcomes in HIV controllers.

BACKGROUND: Hepatitis B virus (HBV) vaccine responsiveness is associated with reduced risk of AIDS or death in HIV-infected individuals. Although HIV controllers (HIC) typically have favorable immunologic and clinical characteristics compared to non-controllers, vaccine responsiveness has not been studied. METHODS AND FINDINGS: In the U.S. Military HIV Natural History Study, HBV vaccine response was defined as antibody to hepatitis B surface antigen (anti-HBs) ≥ 10 IU/L after last vaccination. For determination of vaccine responsiveness, HIC (n = 44) and treatment-naïve non-controllers (n = 476) were not on highly active antiretroviral therapy (HAART) when vaccinated while treated non-controllers (n = 284) received all HBV vaccine doses during viral load (VL)-suppressive HAART. Progression to AIDS or death was also compared for all HIC (n = 143) and non-controllers (n = 1566) with documented anti-HBs regardless of the timing of HBV vaccination. Positive vaccine responses were more common in HIC (65.9%) compared to HAART-naïve non-controllers (36.6%; P<0.001), but similar to non-controllers on HAART (59.9%; P = 0.549). Factors associated with vaccine response for HIC compared to HAART-naïve non-controllers include HIC status (OR 2.65, 95% CI 1.23-5.89; P = 0.014), CD4 count at last vaccination (OR 1.28, 1.15-1.45 for every 100 cells/uL; P<0.001), and number of vaccine doses administered (OR 0.56, 0.35-0.88; P = 0.011). When HIC were compared to non-controllers on HAART, only CD4 count at last vaccination was significant (OR 1.23, 1.1-1.38 for every 100 cells/uL; P<0.001). The rate of AIDS or death per 100 person/years for HIC compared to non-controllers was 0.14 (95% CI 0-0.76) versus 0.98 (95% CI 0.74-1.28) for vaccine responders and 0 (95% CI 0-2.22) versus 4.11 (95% CI 3.38-4.96) for non-responders, respectively. CONCLUSIONS: HIC have improved HBV vaccine responsiveness compared to treatment-naïve non-controllers, but similar to those on VL-suppressive HAART. Progression to AIDS or death can be predicted by HBV vaccine responder status for non-controllers, however these events are rarely observed in HIC.

HIV outcomes in Hepatitis B virus coinfected individuals on HAART.

BACKGROUND: Understanding the impact of hepatitis B virus (HBV) coinfection on HIV outcomes in the highly active antiretroviral therapy (HAART) era continues to be a critical priority given the high prevalence of coinfection and the potential for impaired immunologic, virologic, and clinical recovery. METHODS: Participants from the US Military HIV Natural History Study with an HIV diagnosis on HAART and serologically confirmed HBV infection status at HAART initiation (HI) were classified into 4 HBV infection (HB) groups. HIV virologic, immunologic, and clinical outcomes were evaluated by HB status. RESULTS: Of 2536 HIV-positive HAART recipients, with HBV testing results available to determine HB status in the HI window, HB status at HI was classified as HB negative (n = 1505; 66%), resolved HB (n = 518; 23%), isolated hepatitis B core antigen (n = 139; 6%), or chronic HB (n = 131; 6%). HIV virologic suppression and failure at 6 months or 1 year were not significantly different by HB status. A significantly faster rate of increase in CD4 cell count during the period between 4 and 12 years was observed for chronic HB relative to HB negative. Chronic and resolved HB were associated with an increased risk of AIDS/death compared with HB-negative individuals (chronic HB-hazard ratio = 1.68, 95% confidence interval: 1.05 to 2.68; resolved HB-hazard ratio = 1.61, 95% confidence interval: 1.15 to 2.25). CONCLUSIONS: HB status did not have a significant impact on HIV virologic outcomes, however, CD4 cell count reconstitution after HI and the risk of an AIDS event or death after HI may be associated with HB status.

Delayed-type hypersensitivity (DTH) test anergy does not impact CD4 reconstitution or normalization of DTH responses during antiretroviral therapy.

INTRODUCTION: Delayed-type hypersensitivity (DTH) testing is an in vivo assessment of cell-mediated immunity. Although highly active antiretroviral therapy (HAART) improves immunologic parameters, the relationship between DTH responsiveness and CD4 gains on HAART is not completely understood. We investigated CD4 reconstitution and the change in DTH responses from treatment baseline through 24 months of viral load (VL)-suppressive HAART in the U.S. Military HIV Natural History Study. METHODS: Treatment-naïve subjects with VL <400 copies/mL after ≥24 months on HAART were included (n=302). DTH testing consisted of ≥3 recall antigens, and responses were classified by the number of positive skin tests: anergic (0-1) or non-anergic (≥2). Pre-HAART DTH results were compared for the outcome of CD4 reconstitution at 24 months of HAART. Improvement in DTH responses was also analyzed for those anergic before HAART initiation. RESULTS: Non-anergic responses were observed in 216 (72%) participants, while 86 (28%) individuals were anergic prior to HAART initiation. Demographically there were similar distributions of age at HIV diagnosis and HAART initiation, as well as gender and race or ethnicity. There were no significant differences between non-anergic and anergic participants in pre-HAART CD4 count (409 cells/µL, interquartile range (IQR) 315-517 vs. 373 cells/µL, IQR 228-487; p=0.104) and VL (4.3 log10 copies/mL, IQR 3.4-4.9 vs. 4.4 log10 copies/mL, IQR 3.6-5.0; p=0.292). Median CD4 gains 24 months after HAART initiation were similar between the non-anergic (220 cells/µL, IQR 115-358) and anergic groups (246 cells/µL, IQR 136-358; p=0.498). For individuals anergic before HAART initiation, DTH normalization occurred at 24 months post-HAART in the majority of participants (51 of 86, 59%). Normalization of DTH responses was not associated with CD4 count at HAART initiation (OR 0.73, 95% CI 0.47, 1.09 per 100 cells; p=0.129) nor with AIDS diagnoses prior to HAART (OR 0.34, 95% CI 0.04, 2.51; p=0.283). CONCLUSIONS: DTH responsiveness has been shown to predict HIV disease progression independent of CD4 count in untreated individuals. In the setting of HAART, pre-HAART anergy does not appear to impact CD4 gains or the ability to normalize DTH responses after 24 months of VL-suppressive HAART.

The VACS index predicts mortality in a young, healthy HIV population starting highly active antiretroviral therapy.

BACKGROUND: The Veterans Aging Cohort Study (VACS) index is a weighted combination of age and 8 clinical variables. It has been well correlated with all-cause mortality among HIV-infected patients. The US Military HIV Natural History Study (NHS) cohort provides a different validation population profile, being younger and healthier. A significant portion of the US HIV population is similarly composed; so, evaluation of the VACS index in this population is of great interest. METHODS: NHS subjects have medical history and laboratory data collected at 6-month visits. We performed an external validation of the VACS index in the NHS evaluating correlation, discrimination, and calibration for all-cause mortality after highly active antiretroviral therapy initiation (HI). We then tested whether combining longitudinal VACS index values at different time points improves prediction of mortality. RESULTS: The VACS index at 1 year after HI was well correlated with all-cause mortality (Harrell c statistic 0.78), provided good discrimination (log-rank P < 0.05), and was marginally well calibrated using Brier score. Accounting for VACS index at HI and 6 months after HI significantly improved a standard model, including only the VACS index at 1 year after HI (net reclassification improvement = 25.2%, 95% CI: 10.9% to 48.9%). CONCLUSIONS: The VACS index was well correlated and provided good discrimination with respect to all-cause mortality among highly active antiretroviral therapy initiating subjects in the NHS. Moderate overprediction of mortality in this young, healthy population suggests minor recalibration that could improve fit among similar patients. Considering VACS index at HI and 6 months improved outcome prediction and allowed earlier risk assessment.

Vitamin D deficiency and its association with low bone mineral density, HIV-related factors, hospitalization, and death in a predominantly black HIV-infected cohort.

BACKGROUND: Low bone mineral density (BMD) is common among patients infected with human immunodeficiency virus (HIV) and present in higher rates in black subjects. This study assessed vitamin D levels in HIV cases versus noninfected matched controls to determine if deficiency was associated with BMD and HIV clinical outcomes. METHODS: In total, 271 military beneficiaries with HIV underwent dual energy x-ray absorptiometry (DEXA) screening in 2001-2. Serum 25OH-vitamin D levels were determined using stored serum from the time of DEXA and 6-18 months prior. Two non-HIV-infected controls for each active duty case (n = 205) were matched on age, sex, race, zip code, and season using the Department of Defense Serum Repository (DoDSR). Vitamin D levels <20 ng/mL were considered deficient. HIV-related factors and clinical outcomes were assessed using data collected in the DoD HIV Natural History study. RESULTS: In total, 165 of 205 (80.5%) active duty HIV cases had 2 matched controls available. HIV cases had greater odds of for vitamin D deficiency (VDD) compared with controls (demographics adjusted paired data odds ratio [OR], 1.46, 95% confidence interval [CI], .87-2.45), but this was not statistically significant. Blacks were disproportionately deficient (P <.001) but not relative to HIV status or BMD. Low BMD was associated with typical risk factors (low body mass index and exercise levels, alcohol use); given limited available data the relationship between tenofovir exposure and VDD or low BMD could not be determined. Analysis of HIV-specific factors and outcomes such as exposure to antiretrovirals, HIV progression, hospitalizations, and death revealed no significant associations with vitamin D levels. CONCLUSIONS: VDD was highly prevalent in black HIV- infected persons but did not explain the observed racial disparity in BMD. Vitamin D deficiency was not more common among HIV- infected persons, nor did it seem associated with HIV- related factors/clinical outcomes.