Introduction

Triclocarban and triclosan (together, TCs) are broad-spectrum microbicides active against bacteria, fungi and some viruses [1–4]. Until recently, TCs were found in an ever-changing range of consumer products, including most notably bar soaps (triclocarban) and liquid soaps and toothpaste (triclosan). Introduced in the late 1950s, TCs rapidly became a near-ubiquitous human exposure: in 2011–2013, almost 37% and 72% of US human urine samples contained triclocarban and triclosan, respectively [5, 6]. Although presumably incorporated in wash products to prevent infection, there is no scientific evidence of TC efficacy in community settings. Whereas some studies indicate that the use of triclosan-containing toothpaste decreases plaque and gingivitis [7], use of TC-containing soaps in place of plain soaps has not been shown to decrease infectious illnesses in either households or pediatric populations [8–10].

In the environment, TCs are common contaminants. TCs have been found in 56–100% of surface water samples [11, 12] where adverse effects on aquatic microbiota have been of concern [13]. In experimental animal models, TCs have been reported to have endocrine effects [1]. Without evidence of benefit and with public concerns about environmental and physiological effects, TCs came under increasing scrutiny by regulatory agencies. They were banned from wash products in the European Union in 2010 and, in September 2016, the FDA similarly ruled that TCs can no longer be marketed in antimicrobial wash products. Other uses such as toothpaste, household plastics, industrial cleaning and hospital supplies were not included in this FDA ruling [14].

In 2011 we initiated a prospective birth cohort study, STanford’s Outcomes Research in Kids (STORK), to assess the effect of infectious diseases on infant growth and immune system development [15]. Within this cohort, we nested a randomized intervention of TC-containing wash products to determine whether TCs decreased infection during the babies’ first year.

Materials and methods

Results

From July, 2011 to February, 2015, 158 mothers were enrolled in the STORK cohort, of whom 154 (97%) participated in the TC intervention (78 (51%) randomized to the TC group and 76 (49%) to the non-TC group; Fig 1). The reason for non-participation in the intervention was the desire to continue use of their current wash products (4/4, 100%). At enrollment into the intervention, approximately one third of mothers were expecting a first child; their average age was 30 years, 64% were Hispanic, and 59% were foreign-born (Table 1). Households contained a median of four residents, of whom approximately two were children aged less than 18 years; households were clean but overcrowded [19] with a median of 1.2 people per room. Maternal exposure to cleaning products and chemicals outside of the home was quite common (22%) and most mothers bathed at least once daily (88%). No differences in any of these characteristics were observed between intervention groups.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Participant disposition by randomization group, showing modified intent-to-treat (mITT) and modified per-protocol (mPP) cohorts.

https://doi.org/10.1371/journal.pone.0199298.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Maternal and household characteristics at enrollment, by intervention group (ITT cohort^a).

https://doi.org/10.1371/journal.pone.0199298.t001

Follow-up of mothers during pregnancy was similar for both intervention groups (median proportion of weekly surveys completed from enrollment to delivery: 89% for both TC and non-TC groups, p = 0.75). Loss to follow-up or withdrawal from the study prior to the baby’s birth (n = 21) was more common overall in the TC than in the non-TC group (18 (23%) vs. 3 (4%), p<0.001; Fig 1). Nine mothers could not be contacted and were lost to follow-up; for the remaining 12 mothers, reasons for withdrawal included: a move away from the area (n = 4), time commitment perceived as too great (n = 3), unwillingness to provide blood from the baby (n = 2), a miscarriage (n = 1) and no specific reason (n = 2); no provided reason for withdrawal appeared related to the intervention.

Totals of 60 and 73 babies providing at least one weekly survey were born into the TC and non-TC groups respectively, constituting the analysis sample (mITT; Supporting Information–S1 Table). Both follow-up time and proportion of households withdrawn from the study were similar between groups (Table 2). Babies in the two groups were equally likely to be breastfed and for the same amount of time. Use of day care also was not different between intervention groups. Among babies, infectious disease symptoms were more common in winter months and with increasing age; babies younger than 80 days (~2.6 months) of age rarely experienced infectious diseases symptoms (Supporting Information–S1 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Follow-up, infection and antibiotic use, by intervention group (mITT cohort).

https://doi.org/10.1371/journal.pone.0199298.t002

Primary outcome

Logistic regression modeling did not identify a significant difference in the odds of infectious illness symptoms over time in babies by intervention group in the mITT cohort (likelihood ratio test: mITT analysis, p = 0.88, Fig 2). An analysis excluding the first 90 days after birth (when few illnesses were reported) showed similar findings (p = 0.68). The mPP analysis comparing babies of mothers in the first and fourth quartiles of urinary triclosan also did not show any difference in the odds of illness over time (p = 0.0.83; Supporting Information—S2 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Proportion of babies reporting infectious disease symptoms by week over the first year (mITT cohort).

https://doi.org/10.1371/journal.pone.0199298.g002

Discussion

In this study, randomization of households to “real world” use of wash products with TCs was not related to the occurrence of infectious illness in babies during their first year of life. These findings remained consistent even after taking into account both seasonality and the enrollment period, and after comparing only those babies whose mothers had the highest levels of urinary triclosan to those with mothers with the lowest. Our majority immigrant multiethnic families provided both weekly assessments of illness in their children and at household visits every four months; with this highly granular follow-up, this intervention comparing antimicrobial soaps to regular soaps in the community setting showed that TC-containing soaps did not prevent illness, consistent with prior literature [20].

Several caveats, however, must be noted. In this community setting of primarily underserved families, the rate of reported infections was low in both pregnant mothers and babies. When illnesses did occur, they were typically URIs, with few reported instances of fever, vomiting or diarrhea. Most URIs, however, are viral in etiology [21] and, although TCs are reported to have activity against some viral pathogens, the breadth of viral activity is not well defined. The expectation is that TCs would limit bacterial disease—a fact made evident by the “anti-bacterial” labeling on most TC products. It is notable, then, that children in the TC group received fewer prescriptions for courses of antibiotics than those in the non-TC group according to medical record review. This halving of documented antibiotic prescriptions in the TC group may signal some antimicrobial benefit of TCs with routine household use. Whether this putative benefit derives from killing of pathogens on surfaces or immunomodulatory effects in the host cannot be determined from our data. In both in vitro and animal models, TCs have been linked to a variety of immunologic consequences, both boosting and suppressive in nature [22–24]. However, the findings that TCs did not alter viral symptoms and that triclosan levels in infants were low suggest that alterations in host immunology are unlikely to have led to differences in antibiotic prescriptions.

Although the halving of antibiotic prescriptions in the TC group was intriguing, we could not confirm this finding by review of mothers’ responses to weekly surveys regarding antibiotic use. Half of babies were reported by their mother to have received antibiotics prior to their first birthday, a finding comparable to other studies [25]. Mothers, however, were often unclear as to what medicines were antibiotics: when a random subset of 60 episodes was investigated to confirm their reported antibiotic use, 37% of “antibiotics” were found to be non-antibiotic medications, primarily analgesics (acetaminophen (55%), ibuprofen (5%)) (unpublished data). Because of this inaccuracy, we relied on medical records to determine antibiotic prescription receipt. It should be noted, however, that this too could be inaccurate. Although we requested records from the babies’ primary health care facilities, many families may have visited other locations from which we did not receive records (e.g., urgent care and emergency rooms). There is little reason to suspect, however, that use of these outside facilities was related to randomization by study group, making underreporting of antibiotics from medical records an unlikely confounder.

TC exposure was common even in the non-TC group, as 86% of mothers tested had detectable urinary triclosan levels. While this study successfully randomized households to TC-containing wash products, we did not place restrictions on the use of non-study-provided wash products, and levels of triclosan overall were comparable to population-based studies [26, 27]. Moreover, TC-containing wash products in the home clearly are not the only source of personal TC exposure; indeed, one fifth of our mothers reported exposure to cleaning products or chemicals—possibly including TCs—at their workplaces. Such misclassification might mask a true association between TC exposure from wash products and illness. Additionally, despite the high TC levels in mothers, levels in babies in this study were extremely low. We believe it unlikely that TCs would directly decrease infection rates by altering the babies’ microbial flora. Although this possibility does exist [28], in adults, at much higher levels of exposure, TC-containing wash products have been found to have no effect on the microbiome of the gut or the mouth [29]. The use of TCs in household cleaning products, however, may have a substantial effect on environmental microbes, by decreasing their overall abundance and possibly restricting their variation [30]. Any impact of TCs on antibiotic use, then, would result from preventing transmission from contaminated surfaces that are cleaned with these wash products.

Several limitations are inherent in this study. First, the intervention was not blinded, with household members able to identify their exposure group by reading product labels. To verify exposure, we used spot measures of urinary triclosan that are known to correlate highly with 24-hour urine samples [31]. Nevertheless, although we found that TC-randomized mothers had higher urinary triclosan levels, they were not statistically significantly higher compared to the levels in non-TC-randomized mothers. Second, no exposure dose information was available. While TC is absorbed through the skin [32], triclosan-containing toothpaste is likely the main source of urinary triclosan in mothers; infants, however, are unlikely to be directly exposed and toothpaste use likely will not alter infection risk in children. The amount of exposure to triclosan in the household environment through wash products and dish soap use is not quantifiable. Third, not all infectious illnesses in the child were diagnosed by a HCP. Maternal report of illness, however,—particularly airway-related disease—has been shown to be highly valid [33]. Had we chosen to consider only diagnoses observed by a HCP, mild illnesses that were not brought to medical attention would have been missed. Mothers in each intervention group were equally likely to complete their weekly survey, minimizing the likelihood of differential recall bias. Finally, our analysis of antibiotic prescriptions was incomplete, as not all medical records were available for review.

In summary, using a randomized intervention nested within a well-characterized birth cohort, we found no evidence that “real-life” use of TC-containing wash products changed the occurrence of infectious disease in young children and their mothers. Considerable exposure to TCs from other sources and the preponderance of viral infection, however, may mitigate any effects of household exposure. The finding that children in TC-exposed households had half the antibiotic prescriptions of children in non-TC-exposed houses could indicate the potential of TCs having a large impact on antibiotic use in the US. Regulatory removal of TCs from household wash products, however, will make further investigation of TC effects on child health impracticable.

Supporting information

Acknowledgments

The authors thank T. D. Haggerty and A. Mathur for assistance in data collection, T. Ma and M. Shan for assistance in data management; and K. Krasinska and the Vincent Coates Foundation Mass Spectrometry Laboratory, Stanford University Mass Spectrometry Laboratory (http://mass-spec.stanford.edu) for expertise in urinary triclosan measurements and analysis.