Personal care product use patterns in association with phthalate and replacement biomarkers across pregnancy

The study investigates whether patterns of personal care product use during pregnancy are associated with urinary biomarkers of phthalates and their replacements. Phthalates, used as plasticizers and solvents and common in fragranced products (e.g., DEP), are linked to adverse pregnancy outcomes such as preterm birth and fetal growth restriction. Women, particularly those of reproductive age, often have higher biomarker concentrations due to greater use of personal care products. While personal care products are a potentially modifiable exposure source compared to diet or medical exposures, prior studies show inconsistent associations between individual products and specific phthalate biomarkers, likely due to differing populations, temporal changes in product formulations, lack of consideration of simultaneous product use, and product availability. The study aims to: (1) identify groups with similar personal care product use patterns in early, mid-, and late pregnancy; (2) assess changes in these patterns over gestation; and (3) relate period-specific group membership to urinary concentrations of phthalate and replacement biomarkers.

Previous research on personal care product use and phthalate biomarkers in pregnant and non-pregnant populations has been inconsistent, often examining products individually without accounting for concurrent use. Studies vary by socioeconomic and geographic context and by queried product types. Some consistent links have been reported (e.g., higher MEP with use of fragranced or cosmetic products like hairspray/gel, cosmetics, lotion, deodorant, perfume, sunscreen), yet overall findings remain heterogeneous. Evidence also suggests lotion use may be associated with DEHP metabolites, and certain cosmetics with DBP and DiBP metabolites. No prior work systematically contrasted fragranced versus fragrance-free product use. Intervention data are limited; one trial among adolescent girls using products labeled as free of certain chemicals showed modest decreases in MEP and MBP. Methodologic challenges include non-persistent biomarkers, multiple exposure sources (notably diet), lack of ingredient transparency (phthalates often embedded in “fragrance”), evolving formulations, and racial/ethnic differences in product use that can confound associations.

Design and population: The Human Placenta and Phthalates Study enrolled 303 pregnant women (<14 weeks’ gestation) from prenatal clinics at Eastern Virginia Medical School (EVMS) and the University of Texas Medical Branch (UTMB) during 2017–2018. Eligibility: age 18–50, singleton pregnancy, and no detected fetal/placental abnormalities. Participants attended up to 8 visits (approximate medians: 13, 15, 17, 21, 25, 29, 33, 37 weeks). IRBs at EVMS and UTMB approved procedures; informed consent obtained. CDC laboratory analysis of de-identified samples was not considered human subjects research. Personal care product assessment: At each visit, participants reported products used in the prior 24 hours, including fragrance status (fragranced, fragrance-free, or unknown) where applicable. Analyses focused on personal care/cosmetics (vitamins and other items excluded). For analysis, visits were condensed into three pregnancy periods: early (visits 1–2; 13–15 weeks), mid (visits 3–5; 17–25 weeks), and late (visits 6–8; 29–37 weeks). For each period, the first completed response within that window was used to represent product use. Products with very low use (<10 users per visit) or those highly colinear with race/ethnicity were excluded; some similar products were combined (e.g., conditioner and leave-in conditioner). Product response coding: For most items, a 3-level variable captured no use, fragranced use, or fragrance-free use. If use was reported but fragrance status was unknown, the item was set to missing (with limited exceptions for combined hand soap). Cosmetics and perfume were coded as any use (yes/no). Exposure biomarker collection and analysis: Spot urine was collected at each visit in polypropylene cups; no wipes were used pre-collection. Specific gravity (SG) was measured (PAL-105 refractometer). Samples were stored at −80°C and shipped to CDC on dry ice; CDC stored samples at ≤−40°C before analysis. Eighteen phthalate and replacement metabolites were quantified using enzymatic deconjugation, automated online SPE, HPLC separation, and isotope-dilution tandem MS. Metabolites included: MEP, MBP, MHBP, MiBP, MHiBP, MBzP, MCPP, MEHP, MEHHP, MEOHP, MECPP, MONP, MCOP, MCNP; replacements: MEHHTP, MECPTP (DEHTP metabolites), MHINCH, MCOCH (DINCH metabolites). Handling of values < LOD: instrument-read values >0 were retained; zeros were imputed as LOD/√2; instrument-read < LOD values were not imputed. Molar sums were calculated for parent compounds: ΣDEHP (MEHP, MEHHP, MEOHP, MECPP), ΣDiNP (MCNP, MCOP), ΣDnBP (MBP, MHBP), ΣDiBP (MiBP, MHiBP), ΣDiNCH (MHINCH, MCOCH), and ΣDEHTP (MEHHTP, MECPTP). Molar sums were converted to ng/mL using specified molecular weights. Urine dilution adjustment: Biomarkers were SG-corrected using covariate-adjusted standardization. Natural log(SG) was modeled as a function of maternal age, gestational age at collection, pre-pregnancy BMI, education, and race/ethnicity to predict SG; observed concentrations were divided by (observed SG / predicted SG). Period averaging: To improve exposure stability, biomarker concentrations were averaged within each pregnancy period by geometric mean across available visits; single measurements were used if only one was available. Covariates: Race/ethnicity (composite), education, insurance (private vs self-pay/uninsured/government-assisted), parity, age, smoking before pregnancy, BMI. Statistical analysis: Latent class analysis (LCA; SAS PROC LCA) identified product-use patterns separately in early, mid, and late pregnancy. Class number was selected based on AIC/BIC, membership proportions, posterior probabilities, and interpretability. Latent transition analysis (LTA; PROC LTA) assessed changes in class membership across periods, producing conditional item response probabilities, class prevalences, and transition probabilities. Associations with biomarkers: Period-specific weighted mean biomarker concentrations by latent class were estimated using linear regression with generalized estimating equations. Inverse probability of treatment weights standardized class-specific estimates to the distribution of maternal age, education, insurance, and race/ethnicity in the overall population for each period. Women missing covariates were excluded from these models (n=8, 7, 6 at early, mid, late respectively). Because product-use responses and biomarker measures could come from different visits within a period, analyses used period-averaged concentrations. Visualization: Heat maps displayed percent differences of class-specific means relative to period-specific overall means for low molecular weight (LMW) phthalates (MEP, ΣDnBP, ΣDiBP), high molecular weight (HMW) phthalates (MBzP, MCPP, ΣDEHP, ΣDiNP, MCNP), and replacements (ΣDiNCH, ΣDEHTP). Sensitivity analyses: Repeated analyses with alternative LCA solutions (e.g., 3-class models), different product subsets, and retaining “fragrance unknown” responses, to assess robustness.

• Four latent classes of product use were identified in early and mid-pregnancy and three similar classes plus a modified class in late pregnancy: (1) low fragranced product use (16–23%); (2) fragranced product and low body wash use (22–26%); (3) fragranced product and low bar soap use (26–51%); and (4) low product use (7–34%) in early/mid, replaced by a mixed use with high cosmetics/perfume class (~10%) in late pregnancy. - Class stability: 62–92% of women remained in the same class between early and mid-pregnancy; transitions were more common between mid and late pregnancy. The “fragranced product and low bar soap use” class was most stable. - Biomarkers across pregnancy were relatively stable; MEP and ΣDEHTP were highest in early pregnancy. - Consistent associations: Compared to period-specific means, women in the “low fragranced product use” class had: • MEP concentrations 7–10% lower across all periods. • Replacement biomarkers lower: ΣDEHTP 15–21% lower across periods; ΣDiNCH generally lower (approximately 9–21% lower, with wide CIs). - No class showed universally lower levels across all biomarkers. - LMW phthalates: Patterns varied by period; often lower ΣDiBP and sometimes ΣDnBP among fragranced classes in early pregnancy; in mid and late pregnancy the direction differed by fragranced class. Of note, ΣDnBP was paradoxically higher (22–38%) in the “low fragranced product use” class in early and late pregnancy. - HMW phthalates: Patterns were inconsistent and estimates imprecise; concentrations tended to be lower among the “fragranced product and low body wash use” group in early and late pregnancy. - Sensitivity analyses with alternative class solutions and handling of “fragrance unknown” produced similar conclusions, including consistently lower MEP and replacement biomarkers among the “low fragranced product use” class.

The study found that classifying pregnant women by patterns of personal care product use—accounting for fragrance status and concurrent use—yields meaningful behavioral groups that are relatively stable across early to mid-pregnancy. However, no single product-use pattern was associated with uniformly lower phthalate or replacement biomarkers, reflecting the ubiquity and diversity of exposure sources (e.g., diet, other consumer products) and the non-persistent nature of these chemicals. A consistent and biologically plausible finding was lower MEP among women with low fragranced product use, reinforcing the known role of DEP in fragranced products and demonstrating that behavior-based groupings can recapitulate established exposure relationships. Additionally, lower replacement biomarkers (ΣDEHTP, ΣDiNCH) among the low fragranced group suggest these replacements may also be used in fragranced or cosmetic formulations, not only as DEHP substitutes in other applications. Inconsistencies for LMW (ΣDnBP, ΣDiBP) and HMW phthalates likely reflect multiple sources beyond personal care products, low baseline concentrations for some biomarkers (leading to imprecision), formulation variability, labeling opacity, and demographic patterns in product use. The absence of a universally protective product-use pattern implies that individual behavioral modifications may have limited impact on overall exposure profiles, and that broader strategies (policy changes, manufacturer reformulations, improved labeling transparency) may be necessary for substantive exposure reduction.

Using repeated measures and latent class methods, the study identified distinct personal care product use patterns during pregnancy but found no single pattern that consistently minimized all phthalate or replacement biomarkers. Women reporting low use of fragranced products had consistently lower MEP and replacement biomarker concentrations, suggesting that minimizing fragranced product use may reduce DEP and some replacement exposures. Given the complexity and ubiquity of phthalate sources and the limitations of current product labeling, interventions solely focused on personal behaviors may yield limited benefits. Future research should: (1) test targeted interventions (e.g., fragrance-avoidance strategies) with rigorous exposure assessment; (2) incorporate detailed product composition/brand data; (3) evaluate dietary and other sources concurrently; (4) explore policy and manufacturer-level actions to reduce or replace phthalates and replacements with safer alternatives; and (5) examine differential impacts across demographic groups to address exposure disparities.

• Exposure misclassification: Product use within periods was represented by the first available response; while biomarker averaging reduces random error, misalignment of timing between product use reports and urine sampling could attenuate associations. - Non-persistence and low concentrations: Short half-lives and low baseline levels (e.g., MCPP, MCNP, ΣDiNCH) led to imprecision; percent differences may appear exaggerated at low concentrations. - Lack of ingredient transparency: Unknown and changing formulations, and poor regulation of “fragrance-free” and “phthalate-free” labeling complicate interpretation. - Confounding by other sources: Diet and other consumer products are major exposure contributors; demographic differences in product use patterns may also confound associations despite weighting adjustments. - Racial/ethnic collinearity: Some products showed strong racial/ethnic usage patterns and were excluded from LCA to avoid stratified classes, potentially omitting relevant exposure behaviors. - Missing affiliations or brand-level data: No brand/formulation details were collected; fragrance status was sometimes unknown and handled via missingness in primary analyses.