Mothers’ Memories Mislead: Biomarkers Reveal Hidden Infant Tobacco Smoke Exposure

When scientists try to understand how tobacco smoke harms developing children, they usually begin with a question posed to a mother: did your baby breathe other people’s smoke? A new commentary in Pediatric Research argues that this seemingly simple question may be one of the least reliable instruments in all of developmental science. Writing about a study by Micalizzi and colleagues that validated maternal reports against biochemical evidence, Lynn T. Singer of Case Western Reserve University describes a field confronting an uncomfortable truth: the memories, admissions, and silences of parents can systematically distort the very data on which public health warnings and clinical guidelines are built. The stakes are considerable, because environmental tobacco smoke, or ETS, has been consistently linked to respiratory illness in children, and a growing body of evidence points to additional damage to behavioral and intellectual development. Yet despite decades of research, scientists still cannot establish what constitutes a safe level of exposure, largely because the exposure itself is so difficult to measure accurately.

The Micalizzi study stands out because of an unusual longitudinal resource spanning seventeen years and drawing on two linked investigations. The first, known as Baby’s Breath, was a pregnancy intervention designed to reduce maternal smoking and infants’ exposure to secondhand smoke during the perinatal period. The second, Pathways, was a randomized controlled trial that enrolled 738 women and later followed 109 mother-child pairs from the perinatal study into adolescence. During the perinatal phase, mothers reported their own smoking and their infants’ exposure to tobacco smoke when the children were six months old, and these reports were checked against bioverified cotinine levels. Cotinine, a metabolite of nicotine, is widely regarded as the biomarker of choice for measuring tobacco smoke exposure because it can be detected in saliva, blood, and other biological samples and provides an objective chemical record of recent nicotine intake. Years later, when the children had reached adolescence, the same mothers were asked to recall, retrospectively, how much smoke their infants had been exposed to.

The results reveal a striking asymmetry in the reliability of human memory and self-report. When mothers reported their own smoking at six months postpartum, their answers concorded strongly with their own salivary cotinine levels, suggesting that adults can accurately describe their personal habits when those habits have already been openly acknowledged. But the picture changed dramatically for the infants. Maternal reports of infant smoke exposure did not align with the infants’ own cotinine levels, and the retrospective accounts given during adolescence were likewise inconsistent with the bioverified measurements taken years earlier. Most tellingly, the researchers identified a pattern of systematic bias that depended on whether mothers had acknowledged or denied exposure. When mothers admitted that their infants had been around smoke, their reports were largely reliable. When mothers reported no exposure at all, discrepancies with infant cotinine levels emerged that pointed toward underreporting or misclassification. In other words, denial, not memory alone, appears to be the primary corrupting force in the record.

This pattern has profound implications for developmental neurotoxicology, a field that depends heavily on maternal report to reconstruct exposures that occurred before birth or in early infancy. Stigma and social undesirability exert powerful pressure on what parents are willing to say, and the passage of time degrades memory in ways that are neither random nor benign. The commentary emphasizes that these forces are not unique to tobacco. Alcohol, marijuana, cocaine, methamphetamines, and even prescribed drugs such as opiates all carry social weight that can shape reporting, and all have been implicated in media accounts and research as potential threats to children’s development. Wherever disclosure carries a cost, whether legal, social, or emotional, the integrity of self-reported exposure data comes into question, and researchers who rely on such data without biochemical verification risk building their conclusions on systematically biased foundations.

Converging evidence from other studies reinforces the concern. In a large cohort of polydrug-exposed infants, researchers used multiple measures of exposure, including infant and maternal urine, infant meconium, and a Timeline followback interview. Concordance between interview data and biomarkers was high for cocaine exposure, but the two methods disagreed about the amount of exposure: mothers self-reported lighter use while meconium measurements indicated heavier use. Notably, the mothers in that study were all volunteers who had already been identified by their drug use, a circumstance that, as in the Micalizzi study, likely increased the reliability of their reports. The lesson is sobering: even under favorable conditions, with motivated and already-identified participants, quantitative discrepancies persist between what mothers say and what biology reveals.

The discrepancies become even more dramatic when bioverification occurs outside the supportive context of an intervention study. In a cohort drawn from the Health Outcomes and Measures of the Environment Study, a large prospective birth cohort, researchers found significant gaps between maternal self-report of smoking during pregnancy and serum cotinine levels measured in the same women. Only 12 to 26 percent of mothers reported active smoking or combined active and passive exposure, yet cotinine was detectable in 68.6 percent of maternal serum samples collected at sixteen weeks of gestation, and in an even higher percentage of child serum samples. Maternal serum levels were also the best predictors of childhood wheezing, a common respiratory outcome. Had the researchers relied on self-report alone, they would have substantially underestimated the health risks posed by tobacco smoke exposure, quietly erasing a large fraction of the exposed population from their analyses.

Timing of assessment adds another layer of complexity. In a large prospective study of prenatal alcohol exposure, mothers reported their pregnancy drinking, drug use, and smoking both during pregnancy and again four years after delivery. Retrospective reports of drinking were higher than antenatal reports, yet the antenatal accounts more accurately predicted infant neurobehavioral outcomes, suggesting that memory does not simply fade but actively reconstructs the past. For smoking, antenatal and retrospective reports were highly correlated and both predicted some cognitive outcomes, even without biomarkers. The authors of that study stressed the importance of measuring substance use during pregnancy itself, because most women decrease their use once they realize they are pregnant, and many biomarkers or interviews conducted at birth capture only recent use. A single measurement at delivery can therefore miss the critical early weeks of gestation when exposure may have been highest.

Some exposures resist biomarker measurement altogether. Prenatal alcohol exposure remains uniquely difficult to detect because no valid or reliable biomarker currently exists, although promising work has measured fatty acid ethyl esters in infant meconium and linked their presence and quantity to maternal reports and child outcomes. Other candidate biomarkers include gamma-glutamyltransferase, carbohydrate-deficient transferrin, ethyl glucuronide, ethyl sulfate, and mean corpuscular volume, each with limitations in sensitivity and detection window. For most other substances of concern, reliable biomarkers do exist, but they are scattered across an array of biological matrices: saliva, buccal cells, urine, finger and toenails, deciduous teeth, meconium, stool, hair, blood, breast milk, and cord blood. The ongoing Healthy Brain and Child Development study chose to collect multiple maternal and child biomarkers at multiple points in pregnancy and postpartum, precisely because each biomarker differs in breadth of detection window, substance sensitivity, patient burden, cultural acceptability, cost, and feasibility.

The modern exposure landscape is also far more complicated than it was when earlier cohorts were assembled. Polysubstance use among pregnant women is increasing, and prescription medications, both as prescribed and illicitly obtained, including stimulants, antidepressants, benzodiazepines, gabapentin, and non-benzodiazepine hypnotics, are increasingly used during pregnancy. The legalization of cannabis for medical and recreational purposes has led to greater prenatal exposure, and one large health maintenance organization study found that smoking during pregnancy clustered with alcohol and caffeine use. Because multiple substances can produce additive or synergistic adverse effects, or even moderating beneficial effects, study designs must account for these combinations. Beyond drugs, the neurotoxicant exposome, the constellation of chemical pollutants including lead, cadmium, arsenic, phthalates, polychlorinated biphenyls, and polybrominated diphenyl ethers, is now recognized as an important determinant of health and development, with several of these contaminants more prevalent in low-income homes such as those in the present study.

The clinical and methodological implications of the Micalizzi findings are clear. Neither biomarkers nor self-reports alone are sufficient; both are critical to improving the accuracy of understanding the occurrence of, and the thresholds for, prenatal and postnatal impacts of substances such as environmental tobacco smoke. The commentary alerts researchers to a specific threat to validity: bias and misclassification are most likely when exposure is denied, meaning that the absence of reported exposure should never be treated as the absence of exposure. The paper also offers strategies to reduce uncertainty in cases of denial, including the use of novel biomarkers, sensitivity analyses, and combining self-report with third-party interviews to develop more accurate inferences about exposure thresholds and outcomes. For a science that seeks to protect the developing brain, the message is simple and demanding: trust what mothers say, but always verify it against the chemistry of the body.

Subject of Research: Validity of maternal self-report versus biomarker-based assessment of prenatal and infant substance exposure

Article Title: Trust but verify…the importance of multimethod and biomarker assessments of prenatal substance exposures

Article References: Trust but verify…the importance of multimethod and biomarker assessments of prenatal substance exposures. (n.d.). https://doi.org/10.1038/s41390-026-05544-w

Image Credits: AI Generated

DOI: 10.1038/s41390-026-05544-w

Keywords: environmental tobacco smoke, cotinine, biomarkers, maternal self-report, prenatal exposure, retrospective recall, developmental neurotoxicology, secondhand smoke, polysubstance use, pediatric health, exposure misclassification, public health