CHETS Wales aimed to replicate the Scottish CHETS study . Unlike the Scottish study, which demonstrated a 39% reduction in geometric mean cotinine concentrations post-legislation, this study demonstrated a non-significant 12% decline. When data were divided into tertiles, a significant movement towards the lower end of the distribution was observed, with the percentage of children with cotinine concentrations less than 0.1 ng/ml increasing from 45% to 51%. Hence, whilst an unintended outcome rather than an original aim of legislation, a lowering of exposure occurred. However, the lack of change at the top of the distribution indicates that effects were limited to children whose cotinine concentrations were relatively low pre-legislation.
Consistent with other research [7, 8, 10, 15], the greatest self-reported prevalence of SHS exposure occurred within the home. Prevalence of exposure in the home, a car or someone else's home (private places) did not change significantly. By contrast, a reduction in SHS exposure in cafés or restaurants, buses and trains, and indoor leisure facilities (public places) post-legislation was observed. However, less than 3% of children reported being exposed to SHS in public places pre-legislation. Whilst statistically significant, the public health significance of these changes is likely to be limited compared to potential effects on adult SHS exposure.
Significant reduction in the frequency of perceived SHS exposure was observed post-legislation, with the percentage of children reporting being in a place where people were smoking 'about every day' falling by 4%, and a similar increase in the percentage of children reporting 'never' being in a place where people were smoking. Whilst the fact that young people still saw people smoking in public places may have implications for enforcement, it is possible that children did not distinguish between indoor and outdoor spaces in their responses, potentially overestimating exposure.
As in Scotland, there was no evidence of displacement of smoking from public places into the home. Exposure was greatest at both time points amongst children with two parent figures who smoked in the home, followed by those with a mother figure who smoked in the home, those with a father figure who smoked in the home, and lowest amongst those without parent figures who smoked in the home. However, linear analyses demonstrated no significant changes in geometric mean cotinine concentrations for any subgroup post-legislation. Categorical analyses demonstrated significant increases in the likelihood of children with no parent figures who smoked in the home providing a 'low' cotinine concentration post-legislation. However, as in linear analyses, no significant changes were observed for children with one or more parent figures who smoked in the home. These findings support population-level analyses which indicated greater beneficial impacts of smoke-free legislation post-legislation amongst children with relatively low baseline SHS exposure. A number of studies have assessed the health effects of SHS exposure amongst children. However, the quantification of SHS exposure using salivary cotinine measures is scarce [see, [36, 37], for examples] and there is limited published evidence regarding the health effects of very low levels of SHS exposure. Whilst plasma cotinine concentrations of ≥ 0.17 ng/ml, have been linked to health outcomes such as endothelial dysfunction , the lower threshold at which SHS exposure begins to pose health risks is largely unknown. Given the small nature of declines and the fact that they occurred only at the bottom of the distribution, it is at present unclear what implications if any these observed changes may have for health outcomes in these children. However, the acknowledged susceptibility of children to the detrimental effects of SHS, particularly long-term exposure, supports the need to reduce home-based exposure amongst children who have parent figures who smoke in the home. This remains a challenge.
There was a small but significant decline in the percentage of children reporting having parent figures who smoked in the home, consistent with research which shows a tendency for smoke-free public places to stimulate reduction in smoking within the home . It is possible that this decline is a result of increased success of cessation efforts following legislation, as reported elsewhere  or raised awareness of the dangers of passive smoking impacting on smoking within the home amongst smokers with children. The discrepancy between the non-significant change in reporting of exposure in the home on the previous day, and significant change in frequency of parental smoking in the home may be due to persons other than parent figures smoking in the home. In addition, analyses did not distinguish between parents who smoked in the home 'every day and those who smoked in the home 'sometimes', and it is possible that parents who stopped smoking in the home were those who had previously smoked relatively infrequently in the home.
Children whose parent figures did not smoke in the home were significantly more likely to report 'never' being in a place where people were smoking post-legislation. These children were also more likely to report not knowing how often they were in a place where somebody was smoking post-legislation. It is possible that reduced awareness of the frequency of SHS exposure may reflect a downward trend in exposure. Amongst all children with one or more parent figures who smoked in the home, there was a significant increase in the percentage of children reporting not knowing how often they were in a place where people smoked, suggesting that the lack of change in subgroups with one or both parent figures who smoked in the home may have been a result of diminished statistical power in these smaller groups.
It is possible that differences between the findings of the current study and those of the Scottish CHETS study  are due to floor effects associated with the lower pre-legislation salivary cotinine observed in Wales (median concentration = 0.10 ng/ml) compared to Scotland (median concentration = 0.30 ng/ml). The reasons for this difference are unknown but may be due to a general lower exposure in the population, sample coverage, timing of data collection or pre-legislation changes in Wales due to exposure to the extensive UK media coverage of the legislation in Scotland. Imputation of 47% of cotinine concentrations placed substantial limits on linear analyses compared to the Scottish study, in which only a quarter of cases were imputed.
Limitations of study
In addition to the aforementioned statistical limitations arising from the reliance upon imputation for 47% of children, which were overcome by supplementing linear analyses with multinomial logistic regression analyses, a clear limitation is the difficulty in firmly attributing change to legislation, due to the absence of a counterfactual. However, the maintenance of such a group would have been impracticable. A longitudinal study, following up children over time may have facilitated examination of change but would have made it impossible to distinguish between changes occurring due to increases in children's age, or due to legislation. Whilst the rationale for targeting this age group was to eliminate exposure to other SHS from, for example, peer smoking, generalisability is limited by focusing upon one year group.
Whilst there was relatively low uptake of the study by schools (63%), reducing the intended sample of 80 schools to 75, both the number of schools and students included exceeded the minimum required to achieve the intended power of 80%. Furthermore, the stratified random sampling procedure ensured a representative sample of schools were included. The suggestion that non-participation in CHETS may have been due to collection of saliva samples was not observed in CHETS Wales; only 5 (of 44) schools reported this as a concern. The main reason (n = 10) for non-participation was involvement in other health initiatives and research.
Finally, it is probable that self-reports underestimated actual exposure compared to cotinine measurements since cotinine measures reflect exposure over the previous few days, whilst self-report exposure reflected exposure on the previous day. Furthermore, whilst FAS is recognised as an appropriate self-report method for identifying different levels of affluence , we acknowledge that with evolving lifestyles, use of FAS to measure SES may involve a degree of misattribution.