Systemic inflammation in pregnant women with HIV: relationship with HIV treatment regimen and preterm delivery Rupak Shivakotia, Mark J. Gigantib, Michael M. Ledermanc, Rachel Ketchumb, Sean Brummelb, Daniela Moisic, Sufia Dadabhaid, Dhayendre Moodleye,f, Avy Violarig, Lameck Chinulah, Maxensia Owori, Amita Guptaj, Judith S. Currierk, Taha E. Tahad, Mary Glenn Fowlerl, for the PROMISE study team aDepartment of Ep AIDS Research, H University, Unive Bloomberg Schoo Medicine, Univers HIV Research Un Obstetrics & Gyne University Researc Medicine, The Joh School of Medicin University School Correspondence t Tel: +1 212 305 7 Received: 27 Oct DOI:10.1097/QAD ISSN 0269-9370 Cop terms of the Creative and share the work p Objective: HIV treatment regimen during pregnancy was associated with preterm delivery (PTD) in the PROMISE 1077 BF trial. Systemic inflammation among pregnant women with HIV could help explain differences in PTD by treatment regimen. We assessed associations between inflammation, treatment regimen, and PTD. Design/methods: A nested 1 : 1 case–control study (N¼362) was conducted within a multicountry randomized trial comparing three HIV regimens in pregnant women: zidovudine alone, or combination antiretroviral therapy (ART) with lopinavir/ritonavir and either zidovudine or tenofovir. Cases were women with PTD (<37weeks of gestational age). The following inflammatory biomarkers were measured in plasma samples using immunoassays: soluble CD14 (sCD14) and sCD163, intestinal fatty acid- binding protein, interleukin (IL)-6, interferon g, and tumor necrosis factor a. We fit regression models to assess associations between second trimester biomarkers (mea- sured before ART initiation at 13–23weeks of gestational age and 4weeks later), treatment regimen, and PTD.We also assessed whether inflammation was amediator in the relationship between ART regimen and PTD. Results: Persistently high interleukin-6 was associated with increased PTD. Compared with zidovudine alone, the difference in biomarker concentration between week 0 and week 4 was significantly higher (P<0.05) for both protease inhibitor-based regimens. However, the estimated proportion of the ART effect on increased PTD mediated by persistently high biomarker levels was 5% or less for all biomarkers. Conclusion: Persistently high IL-6 during pregnancy was associated with PTD. Although protease inhibitor-based ART was associated with increases in inflammation, factors other than inflammation likely explain the increased PTD in ART-based regimens compared with zidovudine alone. Copyright © 2024 The Author(s). Published by Wolters Kluwer Health, Inc. AIDS 2024, 38:1111–1119 idemiology, Columbia University Mailman School of Public Health, New York, NY, bCenter for Biostatistics in arvard T.H. Chan School of Public Health, Boston, MA, cDepartment of Medicine, Case Western Reserve rsity Hospitals Cleveland Medical Center, Cleveland, OH, dDepartment of Epidemiology, Johns Hopkins l of Public Health, Baltimore, MD, USA, eDepartment of Obstetrics and Gynaecology, School of Clinical ity of KwaZulu Natal, fCentre for the Program of AIDS Research in South Africa (CAPRISA), Durban, gPerinatal it, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa, hDepartment of cology, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA, iMakerere University-John Hopkins h Collaboration (MUJHU CARE LTD) CRS, Kampala, Uganda, jDivision of Infectious Diseases, Department of ns Hopkins University School of Medicine, Baltimore, MD, kDivision of Infectious Diseases, David Geffen e, University of California, Los Angeles, Los Angeles, CA, and lDepartment of Pathology, The Johns Hopkins of Medicine, Baltimore, MD, USA. o Rupak Shivakoti, 722 W 168th St., New York, NY 10032, USA. 232; e-mail: rs3895@cumc.columbia.edu ober 2023; revised: 29 January 2024; accepted: 12 February 2024. .0000000000003877 yright Q 2024 The Author(s). Published by Wolters Kluwer Health, Inc. This is an open access article distributed under the Commons Attribution-Non Commercial-No Derivatives License 4.0 (CCBY-NC-ND), where it is permissible to download rovided it is properly cited. The work cannot be changed in any way or used commercially without permission from the journal. 1111 mailto:rs3895@cumc.columbia.edu http://dx.doi.org/10.1097/QAD.0000000000003877 1112 AIDS 2024, Vol 38 No 8 Keywords: gut integrity, immune activation, mediation analysis, monocyte activation, prematurity, preterm birth Introduction Preterm birth (PTB) is the leading cause of global childhood morbidity and mortality [1]. Data from various settings show that pregnant women with HIV (WHIV) have higher rates of preterm delivery (PTD) than mothers without HIV [2,3]. Although various risk factors such as low CD4þ T-cell count, high viral load, co-infections, and comorbidities likely contribute to this relationship [4–6], higher systemic inflammation is thought to be another contributing factor [7]. Among pregnant WHIV, general inflammatory markers (e.g. acute phase proteins and interleukins) as well as markers of gut dysfunction [e.g. intestinal fatty acid- binding protein (I-FABP)], microbial translocation (e.g. lipopolysaccharide-binding protein) and monocyte acti- vation [e.g. soluble CD14 (sCD14) and sCD163] have been studied in the context of PTD [8–13]. There are conflicting data, with some studies showing an association for specific markers with PTD [8,9] whereas others did not observe an association [10–13]. However, these studies differed by important characteristics [e.g. antire- troviral treatment (ART) regimen and duration, time of assessment during pregnancy, study setting, and sample size]. Larger studies that account for these important factors, especially randomized controlled studies com- paring various ART regimens to better understand the relationship of inflammation and PTD, are needed to better understand these discrepant results. Studies have shown differences in PTD rates by maternal antiretroviral therapy regimen [3,14,15]. For example, data from the multicountry PROMISE (1077BF/FF) trial showed antenatal zidovudine (ZDV) alone was associated with lower PTD than were two protease inhibitor-based ART regimens [14] during the antepartum (1077AP) drug regimen randomization of the trial. Although the biological mechanisms that explain these findings are not clear, we hypothesized that specific maternal ART regimens (i.e. protease inhibitor-based ART regimens vs. ZDV alone) may have a differential impact on inflammation, which may partly explain the increased PTD risk with certain regimens. In other words, we hypothesized that the effect of maternal ARTon PTD is partially mediated through inflammation. To address these research gaps and to test our hypothesis, we conducted a nested case–control study within the IMPAACT 1077BF/1077FF ‘PROMISE’ trial, which was conducted in 14 sites in East and Southern Africa and 1 site in India. After accounting for timing of sample collection (all in second trimester) and HIV character- istics (ART regimen, duration on ART and CD4þ cell count) in our study design, we assessed the relationship of inflammation at pre-ART and 4-weeks post-ART initiation with PTD. Leveraging the randomized design, we also studied the effect of ART regimen on inflammation levels and the potential mediating role of inflammation in the observed relationship between ART regimen and PTD. Methods Study design and population The PROMISE 1077BF study [14] was an open-label, randomized, multisite, multicomponent clinical trial conducted in seven countries (India, Malawi, South Africa, Tanzania, Uganda, Zambia and Zimbabwe). In the antepartum component (PROMISE 1077AP) conducted between 2011–2014, pregnant women were randomized to one of three treatment arms: zidovudine and single-dose nevirapine plus a 1-to-2weekpostpartum ‘tail’ of tenofovir and emtricitabine (ZDVþsdNVPþTRV tail, zidovudine alone) to reduce the risk of transmission, or combination ARTwith either zidovudine, lamivudine, and lopinavir- ritonavir (3TC-ZDV/LPV-RTV, zidovudine-based ART); or tenofovir, emtricitabine, and lopinavir-ritonavir (FTC-TDF/LPV-RTV, tenofovir-based ART). The pri- mary efficacy outcomes were HIV transmission at 1week of infant age (range 7–14 days after birth), along with maternal and infant safety (i.e. adverse events and adverse birth outcomes). Eligibility criteria for this study are detailed elsewhere [14], but included pregnancy of at least 13weeks of gestational age and not in labor, CD4þ cell count at least 350 cells/ml, no prior triple-drug ART, and no serious pregnancy complications. Women with active tuberculosis (TB), receiving TB or hepatitis B virus (HBV) treatment within 30 days of enrollment, with a fetus with serious congenital malformation were excluded from PROMISE [14]. The primary objectives of this study were to determine the association of inflammation during the second trimester of pregnancy with PTD in WHIV and to determine the effect of HIV treatment regimen with changes in maternal inflammation. We conducted a nested 1 : 1 case–control (N¼ 362; 181 cases and 181 controls) study. Cases were defined as all eligible women with a PTD (<37weeks of gestational age) who were enrolled in PROMISE 1077AP. Controls were women from PROMISE 1077AP without PTD and were Inflammation, preterm delivery and HIV treatment Shivakoti et al. 1113 randomly selected after stratifying by country of enrollment and gestational age categories. As this study was focused on second trimester inflammation (i.e. well before labor, a pro-inflammatory state, in the third trimester), women enrolled after 23weeks’ gestation were excluded; the rationale for the 23-week cutoff was because we wanted a second sample from 4weeks (þ/�7 days) after intervention to also be within the second trimester. Additional eligibility criteria for this analysis were that participants were HBV-negative, had a singleton live birth delivery, and had samples collected before and 4weeks after intervention. Participants from Tanzania were excluded because of challenges with sample shipments out of country. Ethics statement Informed consent was obtained from all PROMISE 1077AP participants and US Department of Health and Human Services guidelines for human studies were followed. This study was approved by the institutional review boards of each participating institution. Data collection Detailed study procedures for PROMISE 1077AP are explained elsewhere [14]. In brief, after enrollment in PROMISE 1077AP, women were randomized into the three arms and the treatment continued for many participants through the postpartum period. Maternal study visits were at enrollment, week 4, and then every 4weeks through delivery; and at labor and delivery. Data on sociodemographic information (e.g. age, education level), anthropometrics (i.e. BMI) and HIV data (e.g. WHO clinical stage) and pregnancy history (e.g. parity, history of previous preterm birth) were collected at the initial and relevant follow-up visits. Laboratory values for complete blood count, CD4þ cell count, HIV viral load and hemoglobin were also obtained. At the labor and delivery visit, infant gestational age was obtained. PTD was defined as a delivery before 37weeks of gestation. Gestational age at delivery was determined by Ballard examination done by a pediatrician trained on the Ballard [16]; if Ballard score was not available, gestational age was determined by obstetrical estimate (if available) or date of last menstrual period in a hierarchical approach. Laboratory procedures Blood samples from study participants were collected in ethylenediamine tetraacetic acid tubes at select visits, including at entry and week 4 postrandomization. Plasma samples were isolated using standardized procedures at each study site and stored at �80 8C until further use. Inflammatory immune markers (i.e. biomarkers) for this study were measured using immunoassays of plasma samples. Using stored PROMISE plasma samples available at the NIAID BRI repository in Rockville MD, MSD multiplex ELISAs (U-Plex, Meso Scale Discovery, Rockville, USA) measured levels in duplicates of interleukin (IL)-6, interferon g (IFNg) and tumor necrosis factor a (TNFa). Single-plex ELISAs (Quanti- kine; R&D Systems, Minneapolis, MN, USA) measured levels,with 20% run induplicates, ofmarkers for monocyte activation (sCD14 and sCD163) and gut barrier function (I-FABP). The immunoassays were conducted at the Lederman laboratory at Case Western Reserve University. Statistical analysis Study population baseline characteristicswere summarized by PTD status. Associations between immunemarkers and PTD were evaluated using multivariable conditional logistic regression models, stratified by enrolment gesta- tional age categories and country. Separate analyses were conducted for each immune marker assessed using three different continuous measures: levels at baseline (i.e. enrolment/week 0/ART initiation), levels at 4weeks (i.e. 4weeks postenrollment/post-ART initiation), and change in inflammatory markers between the two time points. Inflammatory marker levels below the lower limit of detection (LLOD) were set to one-half the LLOD and levels above the upper limit of detection (ULOD) were set to twice the recorded measurement. As prespecified supplemental analyses for PTD outcomes, inflammatory markers were also analyzed as categorical variables. At baseline and 4weeks, inflammation levels were classified as ‘high’ inflammation for values in the highest quartile or ‘not high’ (alternatively, ‘low’) for values in the lower three quartiles. Enrollment and postenroll- ment measurements were pooled to determine the quartiles. For analysis involving both time points, we created a categorical variable with four levels: persistently low (i.e. lower three quartiles at both times), persistently high, increasing (i.e. increases from lower three quartiles at baseline to highest quartile at 4weeks postenrollment), and decreasing. Multivariable models adjusted for the follow- ing baseline variables that were selected based on directed acyclic graphs: treatment arm, maternal age, BMI, CD4þ cell count, CD8þ cell count, parity, HIV-1 viral load, hemoglobin levels, education, and history of PTD.Models involving the change in immunemarkers over 4weeks also adjusted for immune marker levels at baseline. No adjustments were made for multiple comparisons. To evaluate the association between treatment regimen and change in inflammatorymarker fromweek 0 toweek 4,we fitweightedmultivariable parametric regressionmodels for interval-censored data assuming a normally distributed error term. This approach was chosen to account for censoring of measurements below the LLOD and above the ULOD. Models adjusted for baseline inflammation, age, BMI, CD4þ cell count, CD8þ cell count, country, education level, gestational age at entry,HIV viral load, and hemoglobin levels. Inverse-probability weight methods were used to account for the oversampling of cases because of the case–control design. We also conducted a causal 1114 AIDS 2024, Vol 38 No 8 mediation analysis to estimate the indirect effects of maternal ART regimen on PTD via maternal inflamma- tion.Themediator variablewas a binary variable indicating persistently high inflammatory marker levels across both time points. Additional details regarding the mediation analysis are included in the Supplement, http://links.lww. com/QAD/D138. In all multivariable regression analyses, a multiple imputation chained equation procedure was used to impute missing baseline variables (other than biomark- ers). All analyses were conducted in SAS version 9. Results Study population characteristics In our nested study of 362 pregnant women from 6 countries, the majority of participants were enrolled from South Africa (43%) andMalawi (35%). Median [quartile 1 (Q1), quartile 3 (Q3)] age of all selected women was 25.5 Table 1. Study population characteristics by preterm delivery status. Preterm d Characteristic Yes (N¼181) Country India 8 (4%) Malawi 64 (35%) South Africa 78 (43%) Uganda 18 (10%) Zambia 8 (4%) Zimbabwe 5 (3%) Randomization arm Triple ARV (3TC-ZDV/LPV-RTV) 102 (56%) Triple ARV (FTC-TDF/LPV-RTV) 22 (12%) ZDVþsdNVPþTRV tail 57 (31%) WHO HIV stage Clinical stage I 176 (97%) Clinical stage II 3 (2%) Clinical stage III 2 (1%) CD4þ cell count (cells/ml) 521 (425, 664) CD8þ cell count (cells/ml) 827 (608, 1,098) HIV viral load (log10-copies/ml) 4.0 (3.3, 4.6) Age (years) 26.3 (22.0, 29.3) BMI (kg/m2) 25.2 (22.7, 28.3) Gestational age (weeks) 19.1 (16.9, 21.3) Parity 1 (1, 2) 0 45 (25%) 1 or more 136 (75%) Hemoglobin (g/dl) 11.1 (10.2, 11.8) Education Primary school or less 59 (61%) Secondary school or above 37 (39%) Missing 85 History of PTD Yes 9 (5%) No 134 (74%) Unknown 38 (21%) Study population characteristics are shown overall and by preterm delivery (P Numbers (N) and percentage (%) are shown for categorical variables, where variables. CD8þ cell count measurements were missing for 146 participant RTV, zidovudine, lamivudine, and lopinavir-ritonavir (zidovudine-based AR (tenofovir-based ART); ZDVþsdNVPþTRV tail, tenofovir and emtricitabine (21.9, 28.7) years (Table 1). For HIV treatment, 176 (49%) participants were randomly assigned to zidovu- dine-based ART, 145 (40%) to zidovudine alone, and 41 participants (11%) to tenofovir-based ART. At enroll- ment, most women were at WHO clinical stage I (97%). Median (Q1, Q3) CD4þ cell count was 530 cells/ml (423, 643) and median HIV-1 RNA was 4.0 log10 copies/ml (3.4, 4.5) (Table 1). Most participants had at least one previous pregnancy (75%), with only 4% having a previous preterm delivery. Additional baseline demo- graphics are summarized in Table 1. Although most characteristics were similar by cases and controls, there were modest imbalances with respect to HIV treatment and age. The proportion assigned to zidovudine-based triple ART was higher among parti- cipants with PTD (56%) compared with participants without PTD (41%), consistent with our previous findings from this study [14]. The median age was higher among participants with PTD (26.3 years) compared with participants without PTD (24.2 years) (Table 1). elivery (<37 weeks) No (N¼181) Total (N¼362) 8 (4%) 16 (4%) 64 (35%) 128 (35%) 78 (43%) 156 (43%) 18 (10%) 36 (10%) 8 (4%) 16 (4%) 5 (3%) 10 (3%) 74 (41%) 176 (49%) 19 (10%) 41 (11%) 88 (49%) 145 (40%) 175 (97%) 351 (97%) 6 (3%) 9 (2%) 0 (0%) 2 (1%) 543 (423, 623) 530 (423, 643) 854 (639, 1,083) 839 (622, 1,087) 3.9 (3.4, 4.4) 4.0 (3.4, 4.5) 24.2 (21.9, 28.3) 25.5 (21.9, 28.7) 24.9 (22.8, 29.2) 25.1 (22.7, 28.7) 19.1 (17.0, 21.3) 19.1 (17.0, 21.3) 1 (0, 2) 1 (0, 2) 46 (25%) 91 (25%) 135 (75%) 271 (75%) 11.2 (10.5, 11.8) 11.1 (10.4, 11.8) 67 (59%) 126 (60%) 46 (41%) 83 (40%) 68 153 7 (4%) 16 (4%) 133 (73%) 267 (74%) 41 (23%) 79 (22%) TD) status, defined as delivery at less than 37weeks of gestational age. as median and interquartile range (Q1, Q3) are shown for continuous s (73 PTD, 73 non-PTD). ART, antiretroviral therapy; 3TC-ZDV/LPV- T); FTC-TDF/LPV-RTV tenofovir, emtricitabine, and lopinavir-ritonavir (zidovudine alone). http://links.lww.com/QAD/D138 http://links.lww.com/QAD/D138 Inflammation, preterm delivery and HIV treatment Shivakoti et al. 1115 Association of maternal inflammation with preterm delivery Inflammatory marker levels at baseline and week 4 are summarized in Supplementary Table 1, http://links.lww. com/QAD/D138. In the primary analyses on the association of immune marker concentrations with PTD, there were no statistically significant associations with PTD for any of the inflammatory markers at baseline, 4weeks, or change from baseline to 4weeks (Table 2), but some notable odds ratios were observed. These included those for baseline TNFa [adjusted odds ratio [OR]: 2.15 per 1 log10 unit increase; 95% CI 0.87– 5.28] and IL-6 (OR: 1.61; 95% CI 0.71–3.64), and week 4 TNFa (OR: 2.29; 95% CI 0.86–6.11) and IL-6 (OR: 1.85; 95% CI 0.77–4.44) (Table 2). In supplementary analyses of the association of immune marker categories with PTD using conditional logistic regression, there were similar findings of increased odds of PTD with several immune markers at baseline or week 4 (Supplementary Table 2, http://links.lww.com/QAD/ D138). However, these associations were not statistically significant. In the categorical analysis involving both time points of baseline and week 4, odds of PTD for participants with persistently high levels of IL-6 resulted in a statistically significant 2.47-fold increase in the odds of PTD compared with those with persistently low levels of IL-6 (95% CI 1.16–5.29; Table 3). Of note, 17.4% of those with PTD had persistently high IL-6 while only 7.9% of those without PTD had persistently high IL-6. Although persistently high levels of the other markers (except for sCD14) were also associated with increased odds of PTD, these relationships were not statistically significant (Table 3). Treatment regimen and inflammation Median immune marker concentrations were higher at week 4 compared with week 0 for all six markers among participants on tenofovir-based ART and four markers (I-FABP, IFNg, IL-6, and sCD14) among participants on zidovudine-based ART (Supplementary Table 3, http://links.lww.com/QAD/D138). For Table 2. Adjusted odds ratios for preterm delivery by immune marker co Week 0 Week Immune marker OR (95% CI) P value OR (9 I-FABP 0.96 (0.40–2.31) 0.92 1.06 (0. IFNg 1.08 (0.58–2.02) 0.81 1.56 (0. IL-6 1.61 (0.71–3.64) 0.25 1.85 (0. TNFa 2.15 (0.87–5.28) 0.097 2.29 (0. sCD14 0.96 (0.14–6.46) 0.96 1.50 (0. sCD163 0.90 (0.29–2.85) 0.86 1.87 (0. Odds ratios (ORs) greater than 1 indicate higher odds of PTD for a 1-log10-un log10-fg/ml, for monocyte activation biomarkers are log10-ng/ml, and for I-FA by enrollment gestational age and country and adjusted for treatment, age, cell count, education level, and CD4þ cell count. A multiple imputation p CD8þ cell count. participants on zidovudine alone, median immune marker concentrations were lower at week 4 for all markers (Supplementary Table 3, http://links.lww.com/QAD/ D138). Maternal ART regimen was associated with changes in immunemarker concentrations for five of the six immune markers (I-FABP, IFNg, IL-6, TNFa, and sCD14; Fig. 1). Compared with zidovudine alone, the adjusted mean difference in concentration between week 0 and week 4 was significantly higher in the tenofovir-based ART arm for concentrations of I-FABP (0.25 log10 pg/ ml; 95% CI 0.19–0.31), IFNg (0.13 log10 fg/ml; 95% CI 0.06–0.21), TNFa (0.04 log10 fg/ml; 95% CI 0.01– 0.06), and sCD14 (0.09 log10 ng/ml; 95% CI 0.07–0.11) (Fig. 1). Compared with zidovudine alone, those on zidovudine-based ART had a higher adjusted mean difference in concentrations between week 0 and week 4 for I-FABP (0.11 log10 pg/ml; 95% CI 0.07–0.14), IFNg (0.12 log10 fg/ml; 95% CI 0.08–0.17), IL-6 (0.05 log10 fg/ml; 95% CI 0.02–0.07), and sCD14 (0.08 log10 ng/ml; 95% CI 0.07–0.09) (Fig. 1). In formal causal mediation analyses, the estimated proportion of the ARTeffect on increased PTDmediated by persistently high biomarker levels was 5% or lower for all biomarkers (Supplementary Table 4, http://links.lww. com/QAD/D138). This suggests that there is insufficient evidence that any of the immune markers, including persistently high IL-6, mediates the relationship between maternal ART regimen and PTD. Discussion In this multicountry study of pregnant WHIV initiating HIV treatment, we examined the association of second trimester maternal inflammation with PTD. Among various systemic inflammation and gut function markers tested as continuous and categorical variables, we observed that only persistently high IL-6 at treatment ncentration. 4 post-ART initiation Change (week 4-week 0) 5% CI) P value OR (95% CI) P value 50–2.24) 0.88 1.03 (0.45–2.37) 0.95 86–2.82) 0.15 1.68 (0.88–3.21) 0.11 77–4.44) 0.17 1.75 (0.62–4.98) 0.29 86–6.11) 0.098 1.94 (0.39–9.73) 0.42 25–8.84) 0.65 2.70 (0.19–38.11) 0.46 49–7.15) 0.36 4.40 (0.63–30.91) 0.14 it increase in immunemarker concentration. Units for all cytokines are BP are log10-pg/ml.Multivariable conditional logisticmodels stratified BMI, parity, previous preterm delivery, hemoglobin, viral load, CD8þ rocedure was used to impute missing values for education level and http://links.lww.com/QAD/D138 http://links.lww.com/QAD/D138 http://links.lww.com/QAD/D138 http://links.lww.com/QAD/D138 http://links.lww.com/QAD/D138 http://links.lww.com/QAD/D138 http://links.lww.com/QAD/D138 http://links.lww.com/QAD/D138 http://links.lww.com/QAD/D138 1116 AIDS 2024, Vol 38 No 8 Table 3. Adjusted odds ratios for preterm delivery by categorized immune marker levels at week 0 and week 4. Immune Marker Categories OR (95% CI) P value I-FABP Persistently low Ref. Increasing 1.49 (0.81–2.77) 0.20 Decreasing 0.73 (0.33–1.62) 0.44 Persistently high 1.05 (0.50–2.21) 0.89 IFNg Persistently low Ref. Increasing 0.83 (0.42–1.63) 0.59 Decreasing 0.77 (0.39–1.54) 0.46 Persistently high 1.72 (0.79–3.72) 0.17 IL-6 Persistently low Ref. Increasing 0.96 (0.45–2.05) 0.91 Decreasing 0.96 (0.49–1.89) 0.91 Persistently high 2.47 (1.16–5.29) 0.020 TNFa Persistently low Ref. 0.63 Increasing 0.78 (0.28–2.16) Decreasing 2.32 (0.92–5.88) 0.076 Persistently high 1.32 (0.69–2.50) 0.40 sCD14 Persistently low Ref. Increasing 1.16 (0.62–2.17) 0.65 Decreasing 1.38 (0.46–4.19) 0.57 Persistently high 0.94 (0.48–1.83) 0.85 sCD163 Persistently low Ref. Increasing 0.90 (0.29–2.78) 0.85 Decreasing 1.18 (0.46–3.02) 0.73 Persistently high 1.45 (0.79–2.67) 0.23 ORs greater than 1 indicate higher odds of PTD for each category vs. persistently low inflammation. Multivariable conditional logistic models stratified by enrolled country and gestational age and adjusted for treatment, age, BMI, parity, previous preterm delivery, hemoglobin, viral load, CD8þ cell count, education level, and CD4þ cell count. A multiple imputation procedure was used to impute missing values for education level and CD8þ cell count. Categories are: ’ Persistently low’ if concentration levels at week 0 andweek 4were in the lower three quartiles, ’Increasing’ if the concentration level went from the lower three quartiles at week 0 to the uppermost quartile at week 4, ’Decreasing’ if the concentration level went from the uppermost quartile at week 0 to the lower three quartiles at week 4, and ’ Persistently High’ if the concentration level stayed in the highest quartile at week 0 and week 4. initiation and 4weeks postinitiation was strongly associ- ated with increased odds of PTD. We also observed differences in inflammation levels by ART regimen, with higher levels of posttreatment in the maternal ART (tenofovir-based or zidovudine-based) arms as compared with zidovudine only. However, this increased inflam- mation did not explain our earlier observed relationship between maternal ART regimen and PTD in formal mediation analyses. Overall, our results suggest that select inflammatory markers (i.e. persistently high IL-6) in the second trimester are associated with PTD, and maternal ART regimen are linked to increases in levels of various inflammatory markers; however, inflammation does not mediate the effect of ART regimen on PTD. In our study, only persistently high IL-6was associatedwith PTD. Other studies in nonpregnant adults with HIV also suggest an important role for IL-6, where higher levels at time of early HIV infection was one of the strongest predictors of adverse clinical outcomes [17]. Relevant to pregnancy, we also observed an association of IL-6 (third trimester)withPTD in another study conducted inwomen with andwithoutHIV from India [10], consistentwith data from other studies in the general populations without HIV [18]. Biologically, it is known that IL-6 is involved in pro- inflammatory parturition processes, including by recruiting leukocytes that lead to delivery, and it is hypothesized that high levels earlier in pregnancy might result in PTD [18]. Other studies in pregnantWHIV, however, did not observe an association of IL-6 with PTD [11,13]. Although the reason for this inconsistency is not clear, it is likely because of differences in study design. Our study population, obtained from a randomized trial, was different from the two cohort studies that did not detect an association with IL-6. Important differences from these studies included our study population of ART-naive pregnant WHIV who initiated ART during the second trimester of pregnancy, and samples were all obtained in the second trimester prior to and exactly 4weeks after ART initiation. These differences, along with our analytical approaches (e.g. a categorical analysis usingpersistently high levels as avariable was statistically significant whereas continuous variables had nonsignificant associations) could potentially explain the discrepancy in findings. Consistent with other recent studies in people with HIV on ART, we did not observe a statistically significant association of inflammatory markers other than IL-6 (acute phase proteins, other cytokines and gut inflamma- tion markers) and PTD [8,10,12]. This is in contrast to two earlier studies from Spain and India, which showed that WHIV with higher levels of intestinal barrier dysfunction, microbial translocation and monocyte activation had higher rates of PTD [8,9]. However, these were small studies; and women in the study in India were not on ART.Recent larger studies have not seen this Inflammation, preterm delivery and HIV treatment Shivakoti et al. 1117 Fig. 1. Forest plot for treatment effect on change in immune marker concentrations. All estimates use ZDVþsdNVPþTRV tail as reference treatment group. Models additionally adjusted for baseline inflammation, age, BMI, gestational age at entry, country, education level, CD4þ and CD8þ cell counts, hemoglobin, and HIV-1 RNA. Units for all cytokines are log10-fg/ml, for monocyte activation biomarkers are log10-ng/ml, and for I-FABP are log10-pg/ml. 3TC-ZDV/LPV-RTV, zidovudine, lamivudine, and lopinavir-ritonavir (zidovudine-based ART); FTC-TDF/LPV-RTV, tenofovir, emtricitabine, and lopinavir-ritonavir (tenofovir-based ART); ZDVþsdNVPþTRV tail, tenofovir and emtricitabine (zidovudine alone). association [8,10,12], in line with our findings in this study. As the women in these more recent studies were on ART, one possible explanation could be that the partial resolution of gut dysfunction with ART might be sufficient to reduce risks related to PTD. Alternatively, as described with IL-6 above, the discrepancy between gut markers and PTD could be related to differences in other factors such as study design, timing of sample collection during pregnancy or ART/HIV factors. It should also be noted that several of these immune markers did have an association with increased PTD in our study. Notably, we observed a greater than two-fold higher odds of PTD for participants with high TNF-alpha levels at both week 0 and week 4. It is possible that there is a true association between other immune markers and PTD, but our study had insufficient power to detect it. We also observed higher inflammation after treatment initiation for the tenofovir-based and zidovudine-based ART arms. In nonpregnant ART-naive adults, a reduction in soluble markers IL-6 and sCD14 was observed within 4weeks of tenofovir/raltegravir/emtri- citabine ART initiation, with further decreases over the first year [19]. Over longer periods (e.g. 24 and 48weeks) post-ART initiation, reduction of multiple soluble inflammatory markers have also been generally observed compared with baseline in nonpregnant adults, although some studies did not show a significant difference for specific soluble markers or at specific time points (e.g. 24 vs. 48weeks) [19–21]. Data on changes in inflammation is limited in ART-naive pregnant women initiating ART [22], and it is unclear whether the differences between pregnant women and nonpregnant adults is due to pregnancy-related factors or HIV-related factors. Pregnancy is characterized by immune changes, and particularly the second trimester of pregnancy is considered to be an ‘anti-inflammatory 1118 AIDS 2024, Vol 38 No 8 state’ [23]. It is possible that the interaction of pregnancy and more specifically second trimester-related immune profile with HIVor ART initiation results in increases in inflammation, at least in the short-term, in contrast to nonpregnant adults who have reductions. There is data to suggest that levels of some markers, including IL-6 and I- FABP, increase over time during pregnancy in WHIV [22]. Alternatively, or in addition, the discrepant results between the populations could also be due to factors such as the specific ART regimen, timing of assessment (i.e. within 4weeks of ART initiation) and study population characteristics. Related to ARTregimen, it is possible that use of lopinavir/ritonavir-based regimen might partly explain the increased inflammation (e.g. potentially through its lipid-raising effects [24]), as a study in nonpregnant adults with HIV showed an increase in soluble markers including TNFa after initiating lopina- vir/ritonavir-based treatment [25]. Further studies in pregnant women initiating ART could help us better understand these results. We also note that I-FABP levels were higher in the TDF-based ART regimen compared with zidovudine-based ART, a finding that has previously been reported in an HIV prevention trial [26], although future studies are needed to confirm this and understand the potential mechanisms. Our findings of persistently elevated IL-6 being associated with PTD and also being higher in ART-based regimens raised the possibility that IL-6 could be a potential mediator in the observed relationship in the parent PROMISE trial between ART-based regiments and increased PTD as compared with zidovudine alone. However, formal mediation analyses suggested that inflammatory markers (including IL-6) had only minimal mediating effects. This suggests that increased inflamma- tion does not explain the increased PTDwith ART-based regimens. We do want to point out that these findings do not preclude IL-6 or other inflammatory markers having a mediating role for other adverse perinatal outcomes (e. g. low birth weight) or for preterm birth at other time points in pregnancy. Our study has some limitations, including potential limited generalizability to other periods of pregnancy (e. g. third trimester), a limited set of inflammatory markers assessed only in plasma samples (i.e. no data on biomarker production by specific immune cells), potential unmea- sured confounders, and wide confidence intervals in select mediation analyses, and no data on the effect of treatment on inflammation at times beyond 4weeks post- ART initiation. Despite these limitations, our study also has a number of strengths including nesting within a multicountry randomized controlled trial where HIV treatment regimen was randomized, standardization of timing of sample collection in second trimester, longitudinal data, a large sample size of PTD, and assessment of inflammatorymarkers relevant in pregnancy and HIV. In conclusion, we observed that persistently elevated IL- 6, but not other markers, during the second trimester of pregnancy was associated with PTD in a population of WHIV initiating ART. Although initiation of ART regimens resulted in increases in inflammation, our results indicate that this increased inflammation is not a mediator between the observed relationship of ART regimens and increased PTD. This suggests the importance of factors other than inflammation in terms of risk of PTD, and should be addressed in future studies. Acknowledgements The authors thank the study participants for their time and contributions. We also thank the PROMISE study team and the sites, the IMPAACT leadership and the support from pharmaceuticals and NIH. Data availability agreement: the data is not publicly available because of the restrictions in the study’s informed consent documents and in the International Maternal Pediatric Adolescent AIDS Clinical Trials (IMPAACT) Network’s approved human subject protec- tion plan. However, data are available to all interested researchers upon request to the IMPAACT Statistical and DataManagement Center’s data access committee (e-mail address: sdac.data@fstrf.org or sdac.data@sdac.harvard. edu) with the agreement of the IMPAACT Network. Funding: overall support for the International Maternal Pediatric Adolescent AIDS Clinical Trials Network (IMPAACT) was provided by the National Institute of Allergy and Infectious Diseases (NIAID) with co-funding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) and the National Institute of Mental Health (NIMH), all components of the National Institutes of Health (NIH), under Award Numbers UM1AI068632-15 (IMPAACT LOC), UM1AI068616-15 (IMPAACT SDMC) and UM1AI106716-15 (IMPAACT LC), and by NICHD contract number HHSN275201800001I. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. The study products were provided free of charge by Abbott, Gilead Sciences, Boehringer Ingelheim, and Glaxo- SmithKline. Author’s contributions: R.S. conceived of the research question, obtained funding, led this study’s implementa- tion, helped guide the analysis and wrote the primary version of the manuscript. M.J.G., R.K., and S.B. helped with the study design and conducted the data analysis for this project. They also contributed to the interpretation of study findings and manuscript writing. M.L. and D.M. led the laboratory analysis and contributed to the study design, implementation, interpretation, and manuscript mailto:sdac.data@fstrf.org mailto:sdac.data@sdac.harvard.edu mailto:sdac.data@sdac.harvard.edu Inflammation, preterm delivery and HIV treatment Shivakoti et al. 1119 review. S.D., D.M., A.V., L.C., M.O., and T.E.T. were key leading investigators of the parent PROMISE study and contributed to manuscript review and interpreta- tion. A.G., J.S.C., and M.G.F. helped obtain funding, contributed to the study’s design, implementation, interpretation, and manuscript review, and were also key leading investigators of the parent PROMISE study. All authors read and approved the final version of the manuscript. All authors meet criteria for authorship as recommended by the International Committee of Medical Journal Editors (ICMJE) and were fully responsible for all aspects of manuscript development. Conflicts of interest The authors have no relevant conflicts of interest to declare. References 1. Perin J, Mulick A, YeungD, Villavicencio F, Lopez G, Strong KL, et al.Global, regional, and national causes of under-5mortality in 2000-19: an updated systematic analysis with implications for the Sustainable Development Goals. Lancet Child Adolesc Health 2022; 6:106–115. 2. Wedi CO, Kirtley S, Hopewell S, Corrigan R, Kennedy SH, Hemelaar J. Perinatal outcomes associated with maternal HIV infection: a systematic review and meta-analysis. Lancet HIV 2016; 3:e33–e48. 3. Eke AC, Mirochnick M, Lockman S. Antiretroviral therapy and adverse pregnancy outcomes in people living with HIV. New Engl J Med 2023; 388:344–356. 4. Zack RM,Golan J, Aboud S,MsamangaG, SpiegelmanD, Fawzi W. Risk factors for preterm birth among HIV-infected Tanza- nian women: a prospective study. Obstet Gynecol Int 2014; 2014:261689. 5. Slyker JA, Patterson J, Ambler G, Richardson BA, Maleche- Obimbo E, Bosire R, et al. Correlates and outcomes of preterm birth, low birth weight, and small for gestational age in HIV- exposed uninfected infants. BMC Pregnancy Childbirth 2014; 14:7. 6. Watts DH, Williams PL, Kacanek D, Griner R, Rich K, Hazra R, et al. Combination antiretroviral use and preterm birth. J Infect Dis 2013; 207:612–621. 7. Ikumi NM, Matjila M. Preterm birth in women with HIV: the role of the placenta. Front Glob Womens Health 2022; 3:820759. 8. Lopez M, Figueras F, Coll O, Gonce A, Hernandez S, Lonca M, et al. Inflammatory markers related to microbial translocation among HIV-infected pregnant women: a risk factor of preterm delivery. J Infect Dis 2015; 213:343–350. 9. Shivakoti R, Gupte N, Kumar NP, Kulkarni V, Balasubramanian U, Bhosale R, et al. Intestinal barrier dysfunction and microbial translocation in human immunodeficiency virus-infected preg- nant women are associated with preterm birth. Clin Infect Dis 2018; 67:1103–1109. 10. ShafiqMMJ, Naik S, Alexander M, Yadana S, Araujo-Pereira M, Kulkarni V, et al. Association of maternal inflammation during pregnancy with birth outcomes and infant growth among women with and without HIV in India. JAMA Netw Open 2021; 4:e2140584. 11. Powell AM, Persaud D, Anderson JR, Kacanek D, Huo Y, Psoter K, et al., International Maternal Pediatric Adolescent AIDS Clinical Trial (IMPAACT) 1025 protocol team. Markers of in- testinal immune activation and inflammation are not asso- ciated with preterm birth among women with low level HIV viremia. Am J Reprod Immunol 2023; 89:e13680. 12. Kirby MA, Lauer JM, Muhihi A, Ulenga N, Aboud S, Liu E, et al. Biomarkers of environmental enteric dysfunction and adverse birth outcomes: an observational study among pregnant wo- men living with HIV in Tanzania. EBioMedicine 2022; 84:104257. 13. Rittenhouse KJ, MwapeH, Nelson JAE, Mwale J, Chipili G, Price JT, et al. Maternal HIV, antiretroviral timing, and spontaneous preterm birth in an urban Zambian cohort: the role of local and systemic inflammation. AIDS 2021; 35:555–565. 14. Fowler MG, Qin M, Fiscus SA, Currier JS, Flynn PM, Chipato T, et al., IMPAACT 1077BF/1077FF PROMISE Study Team. Ben- efits and risks of antiretroviral therapy for perinatal HIV pre- vention. New Engl J Med 2016; 375:1726–1737. 15. Sebikari D, Farhad M, Fenton T, Owor M, Stringer JSA, Qin M, et al. Risk factors for adverse birth outcomes in the PROMISE 1077BF/1077FF Trial. J Acquir Immune Defic Syndr 2019; 81:521–532. 16. Ballard JL, Khoury JC, Wedig K, Wang L, Eilers-Walsman BL, Lipp R. New Ballard Score expanded to include extremely premature infants. J Pediatr 1991; 119:417–423. 17. Borges AH, O’Connor JL, Phillips AN, Neaton JD, Grund B, Neuhaus J, et al., INSIGHT SMART Study and ESPRIT Groups. Interleukin 6 Is a stronger predictor of clinical events than high-sensitivity C-reactive protein or D-dimer during HIV infection. J Infect Dis 2016; 214:408–416. 18. Prairie E, Cote F, Tsakpinoglou M, Mina M, Quiniou C, Leimert K, et al. The determinant role of IL-6 in the establishment of inflammation leading to spontaneous preterm birth. Cytokine Growth Factor Rev 2021; 59:118–130. 19. Funderburg NT, Andrade A, Chan ES, Rosenkranz SL, Lu D, Clagett B, et al. Dynamics of immune reconstitution and activation markers in HIVR treatment-naive patients treated with raltegravir, tenofovir disoproxil fumarate and emtricita- bine. PLoS One 2013; 8:e83514. 20. Kelesidis T, Tran TT, Stein JH, Brown TT, Moser C, Ribaudo HJ, et al. Changes in inflammation and immune activation with atazanavir-, raltegravir-, darunavir-based initial antiviral ther- apy: ACTG 5260 s. Clin Infect Dis 2015; 61:651–660. 21. Mave V, Erlandson KM, Gupte N, Balagopal A, Asmuth DM, Campbell TB, et al., ACTG PEARLS andNWCS 319 Study Team. Inflammation and change in body weight with antiretroviral therapy initiation in a multinational cohort of HIV-infected adults. J Infect Dis 2016; 214:65–72. 22. Schnittman SR, Byakwaga H, Boum Y, Kabakyenga J, Matthews LT, Burdo TH, et al. Changes in immune activation during pregnancy and the postpartum period in treated HIV infection. Open Forum Infect Dis 2021; 8:ofab245. 23. Mor G, Cardenas I. The immune system in pregnancy: a unique complexity. Am J Reprod Immunol 2010; 63:425–433. 24. Martinez E, Domingo P, Galindo MJ, Milinkovic A, Arroyo JA, Baldovi F, et al. Risk of metabolic abnormalities in patients infected with HIV receiving antiretroviral therapy that con- tains lopinavir-ritonavir. Clin Infect Dis 2004; 38:1017–1023. 25. Squillace N, Trabattoni D, Muscatello A, Sabbatini F, Maloberti A, Giannattasio C, et al. Evaluation of adhesion molecules and immune parameters in HIV-infected patients treated with an atazanavir/ritonavir- compared with a lopinavir/ritonavir- based regimen. J Antimicrob Chemother 2018; 73:2162–2170. 26. Huang Y, Zhang L, Karuna S, Andrew P, Juraska M, Weiner JA, et al. Adults on preexposure prophylaxis (tenofovir-emtricita- bine) have faster clearance of anti-HIV monoclonal antibody VRC01. Nat Commun 2023; 14:7813. Temporal changes in the relative frequency of K103N variants in plasma�RNA Temporal changes in the relative frequency of K103N variants in plasma�RNA Temporal changes in the relative frequency of K103N variants in plasma�RNA Temporal changes in the relative frequency of K103N variants in plasma�RNA Temporal changes in the relative frequency of K103N variants in plasma�RNA Temporal changes in the relative frequency of K103N variants in plasma�RNA Temporal changes in the relative frequency of K103N variants in plasma�RNA Temporal changes in the relative frequency of K103N variants in plasma�RNA Temporal changes in the relative frequency of K103N variants in plasma�RNA Temporal changes in the relative frequency of K103N variants in plasma�RNA Temporal changes in the relative frequency of K103N variants in plasma�RNA Temporal changes in the relative frequency of K103N variants in plasma�RNA Temporal changes in the relative frequency of K103N variants in plasma�RNA Temporal changes in the relative frequency of K103N variants in plasma�RNA Temporal changes in the relative frequency of K103N variants in plasma�RNA