Water hardness and atopic dermatitis in the first year of life in the New Hampshire Birth Cohort Study

To the Editor,

Atopic dermatitis (‘eczema’), is a chronic inflammatory skin disease, characterized by itchy, inflamed and dry skin. Eczema has a multifactorial aetiology with genetic factors such as filaggrin gene mutations (FLG), and environmental factors such as antibiotic exposure, breastfeeding, smoking, pet ownership and water hardness.¹ Studies have reported a higher prevalence of childhood eczema in hard water areas, especially in children carrying the FLG skin barrier gene loss-of-function mutation. A UK cross-sectional study in the 1990s reported that primary school-age children living in hard water areas had increased risk of eczema in comparison to those in softer water areas, (OR: 95% CI) (1.28: 1.04 to 1.58).² This association was confirmed amongst school children in Japan (1.12: 1.06 to 1.18)³ and Spain (2.29: 1.19 to 4.42).⁴ We subsequently confirmed this relationship in a cross-sectional study in 1300 UK infants, adjusting for confounders (1.87: 1.25 to 2.80).⁵ A Danish birth cohort study further reported that eczema prevalence 0-18 months increased with water hardness (1.05: 1.03 to 1.07). In contrast, the INMA Spanish birth cohort found no significant difference in eczema prevalence between harder and softer water areas in infants at 14 months of age (0.79: 0.45 to 1.39).⁶

A UK randomized trial of water softeners in children with moderate to severe eczema failed to show any improvement in eczema severity in the water softener group.⁷ However a recent randomized feasibility trial found that the introduction of a water softener from birth led to lower prevalence of eczema at age 6 months, although the trial was not powered to be confirmatory.⁸ These studies were conducted in predominantly hard water areas and little research has explored the association between domestic water hardness, assessed by calcium carbonate (CaCO₃) level, and eczema in soft water areas. We sought to fill this gap in the evidence base using data from the New Hampshire Birth Cohort Study (NHBCS), an ongoing cohort whose participants live in a rural region in the United States with naturally soft water.⁹ We also sought to assess the evidence for a gene-environment interaction in a soft water area following the UK hard water area study where eczema risk was higher in children exposed to the hardest water who had filaggrin (FLG) loss-of-function variants.¹

The current analyses include 962 mother-infant pairs recruited between January 2009 and November 2020 who provided reports of their child’s health at four, eight and twelve months from birth for: any eczema, eczema lasting more than two days, eczema requiring a doctor visit, eczema requiring prescription medication. We focus here on overall eczema prevalence 0-12 months. CaCO₃ content of the domestic water supply was obtained from analysis of a tap water sample obtained at recruitment. Factors related to eczema were explored using logistic regression with adjustment for confounders: infant sex, gestational age, birthweight, parity, family history of atopy, pet exposure and use of a device to purify, soften or treat the home water supply. Pet exposure was recorded when the child was 18 months and so sensitivity analyses were performed without this variable. In the replication ‘gene-environment’ analyses, family history of eczema was analysed used as a potential modifier of the effect of water hardness on eczema (FLG data were not available). CaCO₃ level was cut at the current dataset’s CaCO₃ median (23.5mg/L) following the method of the UK study.¹ The characteristics of those with and without eczema data were compared to check the representativeness of the sample, and multiple imputation was used to account for missing eczema responses as a sensitivity analysis. Analyses were performed in R version 4.04 and Stata v17. Additional information about study methods and findings are available in the following repository: https://osf.io/ukzm5/files/osfstorage/6381dba9a98e5f04981034b8.

For mothers: mean age was 31 years, mean BMI 26.1, 44% were nulliparous and 56% were educated beyond high school. For the children: mean birthweight was 3429g, mean gestation 39wks, 51% were male. The population characteristics of those with and without eczema data not appreciably different.

Parent-reported eczema at 4, 8, 12 months was 7.9%, 14%, 14% with an overall prevalence 0-12 months, 28% (95% CI: 25 to 31%). Overall prevalence of eczema lasting more than two days and eczema needing a doctor visit were 26% (23 to 29%), 13% (11 to 15%) respectively.

Male sex, birthweight, gestational age, maternal eczema, maternal hay fever, paternal eczema, and family history of atopy were all positively correlated with eczema. Seventy-one percent of families reported pets or animals at home but there was no statistical association between type of pet and eczema (table 1). The distribution of CaCO₃ was positively skew with median 23.5 mg/L and interquartile range 10 to 46 mg/L. There was no clear trend in eczema prevalence by CaCO₃ quartile, and while prevalence overall was higher in those with CaCO₃ above the median, this was not statistically significant. These trends did not change after adjustment for confounding factors.

Table 1.

Prevalence of parent-reported eczema 0-12 months by participant characteristics

	N=962		Difference (95% CI)	P-value1
	r/n	% with eczema
Infant characteristics
Male sex	153/496	31%	5.7% (0.0% to 11.3%)	0.050
Female sex	117/465	25%
Birthweight, g
<2500	7/38	18%		0.007
>=2500	222/792	28%
Gestation age (weeks)				0.013
<37	11/76	15%
>= 37	260/886	29%
Maternal characteristics
Enrolment age(years)				0.685
18-24	15/63	24%
25-34	184/658	28%
35-45	72/232	31%
Pre-pregnancy BMI (kg/m2 )				0.407
<18.5	9/25	36%
18.5 to <25	124/482	26%
25+	134/441	30%
Parity				0.525
0	108/411	26%
1	102/347	29%
2+	61/204	30%
Ethnicity/race
White	261/927	28%	0.1% (−15.8% to 15.9%)	>0.999
Other	9/32	28%
Highest educational level				0.479
High school or less	20/88	23%
Junior college/college	152/524	29%
Postgraduate	87/307	28%
Smoked during pregnancy				0.473
Yes	11/49	22%	−5.9% (−17.9 to 6.2%)
No	252/892	28%
Delivery mode				0.286
Vaginal	195/664	29%	3.6% (−9.7% to 2.5%)
C-section	76/295	26%
Breastfeeding postpartum				0.601
Yes	217/758	29%	2.8% (−11.3% to 5.7%)
No	31/120	26%
Maternal eczema:				0.006
Yes	43/108	40%	13.1% (3.4% to 22.8%)
No	228/854	27%
Maternal hay fever:				0.041
Yes	45/124	36%	9.3% (0.34% to 18.3%)
No	226/838	27%
Maternal asthma:				0.109
Yes	56/167	34%	6.5% (−1.3% to 14.3%)
No	215/795	27%
Paternal eczema:				0.032
Yes	23/55	42%	14.4% (1.0% to 27.7%)
No	226/824	27%
Paternal hay fever:				<0.999
Yes	45/160	28%	0.3% (−7.4% to 7.9%)
No	206/725	28%
Paternal asthma:				0.431
Yes	19/80	24%	4.8% (−14.6% to 4.9%)
No	252/882	29%
Parental history of atopy:				0.003
Yes	127/380	33%	11.8% (2.8% to 14.6%)
No	144/582	25%
Pets at home:				0.204
Cats only	21/73	28%
Dogs only	58/192	30%
Mixed pets	27/134	20%
None	125/439	29%
Device to purify/soften/treat home water supply				0.139
Yes	42/464	31%	3.6 % (−1.2% to 10.6%)
No	112/432	26%
CaCO3 quartiles				0.744
<10.5	71/252	28%
10.5 to 23.8	58/228	25%
23.9 to 46.3	70/237	30%
>46.3	72/245	29%

¹For all continuous variables, eg birthweight, P value is calculated from continuous form

In the gene-environment replication analysis with CaCO₃ dichotomized at the median value, the interaction between CaCO₃ and family history of eczema was not statistically significant, although the OR was highest in the group of children living in areas with a high level of CaCO₃ and who had a family history of eczema. Estimated odds ratios in this group was: i) any eczema (OR=1.88; 95% CI 1.09 to 3.24) with a non-significant trend (P=0.10) and ii) eczema lasting more than two days (OR=1.77; 95% CI 1.02 to 3.08), test for trend (P=0.056) (table 2). The OR and 95% CIs for these analyses were very similar following multiple imputation of missing data with the equivalent OR of 1.78 (1.09 to 2.92) for any eczema and 1.76 (1.05 to 2.94) for eczema lasting more than two days.

Table 2.

Prevalence of parent-reported eczema 0-12 months by calcium carbonate content of domestic water and family history of eczema

N=830	OR1	95% CI	Interaction P value	P for trend2
Any eczema			0.953	0.10
Low CaCO3, no family history eczema	1.0 (ref)
High CaCO3, no family history eczema	1.25	0.87 to 1.79
Low CaCO3, family history eczema	1.46	0.76 to 2.8
High CaCO3, family history eczema	1.88	1.09 to 3.24
Eczema > 2 days			0.93	0.056
Low CaCO3, no family history eczema	1.0 (ref)
High CaCO3, no family history eczema	1.09	(0.75 to 1.59)
Low CaCO3, family history eczema	1.56	0.81 to 2.99
High CaCO3, family history eczema	1.77	1.02 to 3.08
Eczema with Dr visit			0.47	0.89
Low CaCO3, no family history eczema	1.0 (ref)
High CaCO3, no family history eczema	0.86	(0.53 to 1.39)
Low CaCO3, family history eczema	1.48	0.66 to 3.35
High CaCO3, family history eczema	0.83	0.37 to 1.88

¹model fitted is eczema (outcome, yes/no), birthweight, gestational age, baby sex, parity (0/1/2), family history of eczema (yes/no), calcium carbonate (below/above median), Device to purify/soften/treat home water supply (yes/no) and pets (none/cats only/dogs only/other pets)

²trend test was post-hoc

³interaction P value is 0.49 fitting CaCO₃ as a continuous variable

This study found little evidence for a positive correlation overall between eczema and the level of water hardness among young children in this rural birth cohort in the United States. Using the US Government cut-point for hard water, CaCO₃ >60mg/L, the overwhelming majority of mother/child pairs in this sample, 84%, had naturally soft water at home. The period prevalence, 0-12 months, for parent-reported eczema was 28%, in keeping with other birth cohort studies and so the lack of an association between eczema and water hardness per se seems likely to reflect its multifactorial genetic and environmental causes. The observed associations between eczema and male sex, lower mean birthweight, lower gestational age, and parental history of atopy were consistent with other studies, and we saw an elevated risk of eczema in infants living in the hardest of the soft water areas whose parents had a history of eczema which while not statistically significant, did mirror findings in a hard water area.¹

This study has strengths and limitations. Strengths are first that domestic water hardness was measured directly from a sample from the family’s domestic water supply, increasing the precision of the estimated exposure in comparison to studies where water hardness was estimated from postal/zip codes and municipality. Second, the eczema data came from responses to three prospective parental questionnaires in infancy, and these have previously been shown to correlate well with medical notes. Third, the statistical analyses were thorough - we adjusted for neonatal and family factors, and then used multiple imputation to test the sensitivity of the complete dataset analyses to missing data. Fourth, we sought to determine whether there are high-risk subgroups by modelling the interaction between water hardness and family history of eczema. We acknowledge this study’s limitations: we did not have either of the Eczema Area and Severity Index (EASI) or the SCORing Atopic Dermatitis (SCORAD) assessments which are more objective than parental reports of eczema. Neither did we have bathing frequency, types of body soaps or emollient applications in infancy so could not include these as covariates in our models. Finally we used parental history of eczema as a proxy for FLG mutation as genetic data were not available and acknowledge that the correlation between these two is not perfect. Our sample was too small to do a subgroup analysis for preterm and/or low birthweight new-borns and we plan to do these analyses in the future when we have greater numbers of participants, as the NHBCS is still recruiting.

In conclusion, analyses of a birth cohort from a rural US state with predominantly soft water did not find an overall positive relationship between eczema and the level of water hardness. Our findings do raise the possibility of effect modification by family history of eczema with a higher risk of eczema among children with a parental history of eczema exposed to higher CaCO₃ domestic water levels. Future work is needed across the United States with larger numbers of children, wider ranges of water hardness levels and full covariate data to enable subgroups of children who may be particularly vulnerable to water hardness to be identified.

Key Messages:

• Water hardness is not correlated with infant eczema overall in a naturally soft water area.

• Parental history of eczema may be an effect modifier in presence of high CaCO₃ levels.

• Future research needs to identify children who are particularly vulnerable to effects of water hardness