Did Smokefree Legislation in England Reduce Exposure to Secondhand Smoke among Nonsmoking Adults? Cotinine Analysis from the Health Survey for England

Background On 1 July 2007, smokefree legislation was implemented in England, which made virtually all enclosed public places and workplaces smokefree. Objectives We examined trends in and predictors of secondhand smoke exposure among nonsmoking adults to determine whether exposure changed after the introduction of smokefree legislation and whether these changes varied by socioeconomic status (SES) and by household smoking status. Methods We analyzed salivary cotinine data from the Health Survey for England that were collected in 7 of 11 annual surveys undertaken between 1998 and 2008. We conducted multivariate regression analyses to examine secondhand smoke exposure as measured by the proportion of nonsmokers with undetectable levels of cotinine and by geometric mean cotinine. Results Secondhand smoke exposure was higher among those exposed at home and among lower-SES groups. Exposure declined markedly from 1998 to 2008 (the proportion of participants with undetectable cotinine was 2.9 times higher in the last 6 months of 2008 compared with the first 6 months of 1998 and geometric mean cotinine declined by 80%). We observed a significant fall in exposure after legislation was introduced—the odds of having undetectable cotinine were 1.5 times higher [95% confidence interval (CI): 1.3, 1.8] and geometric mean cotinine fell by 27% (95% CI: 17%, 36%) after adjusting for the prelegislative trend and potential confounders. Significant reductions were not, however, seen in those living in lower-social class households or homes where smoking occurs inside on most days. Conclusions We found that the impact of England’s smokefree legislation on secondhand smoke exposure was above and beyond the underlying long-term decline in secondhand smoke exposure and demonstrates the positive effect of the legislation. Nevertheless, some population subgroups appear not to have benefitted significantly from the legislation. This finding suggests that these groups should receive more support to reduce their exposure.

Creating the dummy variables for a logistic regression model to compare secondhand smoke exposure between two time periods.
In this section we describe how the dummy variables in the logistic regression model were coded so that the odds of having undetectable cotinine in a six month period with the odds in the previous six month period could be compared. A similar approach was also taken to compare geometric mean cotinine levels between a six month period and a preceding one.

Defining the dummy variable
Let Z equal the date that the nurse visited the respondent and p represent the proportion of non-smoking adults with undetectable cotinine. We define 12 dummy variables as follows: In our analysis, equation (1) also includes an intercept and the predictors listed in Table   1, but without the linear and quadratic terms for time and the binary smokefree legislation predictor. The exponential of a regression coefficient β associated with a dummy variable coded 0 or 1 represents the odds ratio between a six month period and the preceding one. For example, ) exp( 1 β is the odds ratio between July-December 1998 and January-June 1998.
During periods when no cotinine data were collected, we assume that the odds ratios between a six month period and the preceding one remain constant and we can estimate this constant odds ratio by recoding the 1 in the dummy variable associated with this time period with a value equal to the number of missing six month comparisons. For example, no cotinine data were collected for the period 1999 and January to June 2000.
If this data had been available, we would have calculated four odds ratios between July 1998 and July 2000 (i.e. four comparisons of a six month period to the preceding one) and we therefore use the value 4 in 2 X instead of 1. To demonstrate how the constant odds ratio is derived, we define the odds of having undetectable cotinine for each 6 month period from July 1998 to December 2000 as: The odds of having undetectable cotinine in July-December 2001 is: The odds of having undetectable cotinine in January-June 2001 is: Therefore the odds ratio between these two periods can be calculated as: The odds of having undetectable cotinine in July-December 2003 is: Therefore the odds ratio between these two periods is given by:

Comparing two odds ratios
In our paper, we investigated whether the odds ratio between the second and first half of 2007 (i.e. a comparison of the six months post-and pre-legislation) was significantly higher than other six month comparisons between 1998 and 2008. In this section, we illustrate how to test for a significant difference between this and another odds ratio using the ratio of odds between the second and first half of 2001 as the comparison odds ratio.
We assume that the odds ratio between the second and first half of 2001 is ) exp( 4 β , as was described in Example (i). We then set the odds ratio between the second and first half of 2007 to equal ) exp( 4 β multiplied by a factor N. We can test whether N is significantly different from 1, i.e. whether there is a difference in magnitude between the two odds ratios, by fitting the logistic regression model defined in equation (1)  (2) N is equal to ) exp( 10 β in equation (2) and we can therefore test whether the magnitude of the two odds ratios are different by testing whether 10 β is significantly different from zero.  Figure 1. Comparison of secondhand smoke exposure in non-smoking adults in England in a six month period compared with the previous six month period using: a) odds ratio of the proportion with undetectable cotinine a , b) ratio of geometric mean cotinine. Error bars indicate 95% confidence intervals. SFL shows the odds ratio comparing the six months prior to the legislation being implemented with the six months post-legislation. a for example: jan-jun03:jul-dec03 represents the odds of having undetectable cotinine in July to December 2003 compared with January to June 2003. jul-dec03:jan-jun07 represents the ratio odds of having undetectable cotinine in a six month period compared with the previous six month period between July 2003 and June 2007 (i.e. we assume a constant odds ratio for each six month comparison during this time period). a) b)