Results

We do not know how total mortality varies with changing levels of exposure to SHS. Therefore calculations of attributable risk must begin with the question of which fatal diseases are influenced by SHS. Table1 lists conditions which are potentially life threatening and have been associated with SHS. The first group includes conditions that have been closely studied in many populations using a variety of study methods. The second group includes two conditions that have been less extensively researched, but large well conducted studies in New Zealand populations report associations between SHS and fatal outcomes.13 14 The evidence of an association with sudden infant death syndrome (SIDS) is stronger than that for stroke, but in both instances the New Zealand studies are consistent with reports from other countries.15-17 Non-causal explanations are, in our view, unlikely, but estimates of the strength of the effect of SHS must be treated with caution. Third, we have listed conditions that are well known as causes of morbidity caused by exposure to SHS, but there is scant evidence of an effect on deaths. In the final group are conditions that have been linked with SHS, but the literature in these instances is limited.

The first two groups in table 1 were included in the calculation of deaths caused by SHS. We provided for each condition our best estimate of the increased risk of death caused by SHS (the source of these figures is shown in the table) and an estimate of the average time from exposure to death, based on what is known about the effects of active smoking.

In table 2 we report what is known about historical exposures to SHS in New Zealand, and these data are combined with the relative risk estimates to calculate numbers of deaths (table 3). The first figure in the final column refers to current deaths related to exposures to SHS in the past. We have also calculated the deaths that will occur in the future, associated with present exposures to SHS. This latter figure, which represents potentially avoidable deaths, assumes that the lag time between exposure and mortality has elapsed for all conditions. In other words, the number we have calculated refers to the deaths avoided, from about 2015 onwards, if exposure to SHS ceased immediately.

Discussion

Attributable risk estimates should be regarded as a guide to policy decisions rather than a precise prediction of the number of lives that will be saved by public health interventions. For one thing, the number of attributed deaths depends on the order in which the component causes of different diseases are removed.18 We have assumed that control of SHS is the first intervention in each instance although in practice this may not be the case. For example, parents may change the position in which babies are put to sleep sooner than change their smoking patterns. If this happens the total number of sudden infant deaths will fall, as will the number attributable to SHS, although these latter deaths will make up a greater fraction of the total (that is, the PAR% will rise).

Estimates of attributable mortality contain other uncertainties, some of which are quantifiable. In table 4 we have listed what we see as the most important assumptions underlying our estimate of deaths caused by SHS. The values in the first column include key inputs to the attributable and avoidable mortality equations. In the next column we have laid out plausible, alternative conditions, and have then recalculated the numbers of deaths caused by SHS.

We have treated ex-smokers as susceptible to the effects of SHS on heart disease and stroke. There is no a priori reason why this group should be immune, and there are some data in support. The US Nurses Health Study reported an 82% increase in risk of heart attack (95% CI 0.76 to 4.34) among ex-smokers occasionally exposed to SHS and a relative risk of 2.38 (95% CI 0.98 to 5.79) among those regularly exposed to SHS.19 Howard and colleagues20measured the thickness of the wall of the carotid artery using ultrasound, and found among never smokers that the increase in vessel wall thickness over three years was 22% greater in those reporting exposure to SHS: among past smokers there was a similar effect (an 18% greater increase associated with SHS exposure, after adjustment for potential confounding factors). We know of no empirical studies and no theoretical reasons indicating that time since quitting is likely to affect the relation between passive smoking and risk of heart disease, although the absolute risk of heart disease will tend to be higher in recent ex-smokers. For this reason we have treated all ex-smokers as susceptible to the effects of SHS, but acknowledge the uncertainty that surrounds this assumption because of paucity of data.

What age groups are at risk of disease caused by SHS? Kawachi and colleagues7 argued that most of the effects of SHS on heart disease are likely to be acute and transitory and therefore work exposures would have little impact on disease risk past the age of retirement. However, there is evidence of effects of SHS on platelet function, vascular endothelium, and myocardial exercise tolerance, at levels commonly found in the workplace.21 This means that, as with active smoking, there are likely to be both thrombogenic and atherogenic components of the increased risk caused by SHS22. The relative importance of the different routes of action is not known, but it is implausible that exposures before age 65 should have no effect on risk of disease over age 65. However, we have included an upper age limit of 75 for the workplace exposures, on the grounds that the time from exposure to heart death is unlikely to be greater than 10 years. In the home, the estimate of a 30% increase in risk of disease is based on studies which had no age bounds (in fact, the average age in this pooled analysis was 65 years23). There are no data on stroke risk from SHS by age, but we have excluded from our analysis deaths at age 75 and over to match the age restrictions in the study by Bonita and colleagues.13

Depending on how “exposure” is defined, in the recent New Zealand surveys the proportion of all those exposed to SHS who are exposed at both home and work is between 10–30%.24 This is consistent with recent data from elsewhere. For example, in the NHANES III study in the USA 20% of those exposed at work reported that they were also exposed at home.25 Should those who are exposed at home and work be treated as susceptible to the effects of SHS in both settings? We think so, since the relative risk estimates in the epidemiological literature were calculated on the basis of exposure at home or at work (none, as far as we know, were calculated on the basis of exposure in one setting but not the other). Should the effects at home and at work be regarded as independent of one another? There are no data to directly answer this question, so we conclude the most appropriate course of action is to add the numbers of cases attributed to SHS at home and at work. A priori, there are probably stronger grounds for suspecting a negative interaction than a positive one: for coronary heart disease at least the dose–response relation suggests that the pathways leading to disease may be rapidly saturated following exposure to SHS. Therefore, as part of the sensitivity analysis we repeated the calculation of avoidable deaths under the extreme assumption that those exposed at home would not be affected by exposures at work. The effect is small (a reduction of 5% in the attributed deaths).

The rate of deaths caused by SIDS has been high in New Zealand by international standards, although there has been a notable decline (among non-Maori largely) in recent years.26 It appears that this decline is associated with changes in infant sleeping position, and possibly also a reduction in smoking by parents. We have used the results of the most comprehensive study in New Zealand14 to estimate the prevalence of SIDS and the impact of SHS (defined in this case by whether or not the mother smoked after birth). A case–cohort study was conducted including all infants born in New Zealand between October 1991 and September 1993. There were 232 SIDS cases in the cohort and these were compared with 1200 randomly selected control subjects. Maternal smoking was found to be the strongest risk factor, and its effect was modified by bed sharing. Using the exposure prevalence and relative risk figures from the paper by Mitchell and colleagues,14 and including only the effect of SHS independent of bed sharing, we estimate that there are approximately 50 SIDS cases a year attributable to SHS. This does not take account of maternal smoking during pregnancy, but epidemiologic and mechanistic data suggest that SHS does act independently of transplacental exposure to smoke products from the mother's (active) smoking.16

Population attributable mortality is an average and it is important to note that the effect of tobacco on health is unevenly shared across the population. For instance, exposure to SHS is known to vary by socioeconomic status.27 Ethnicity is also a marker of susceptibility: in New Zealand, disease rates are generally higher among the indigenous Maori who make up approximately 15% of the national population, and smoking rates are also higher among Maori. A Maori non-smoking adult or child is likely to be surrounded by twice as many smokers per household, on average.28

The list of assumptions in table 4 is not comprehensive, and we have taken the simplest possible approach by varying only one parameter at a time. But we suggest the results do indicate a plausible range of future deaths caused by current exposures to SHS. The total number was most sensitive to varying the relative risk estimates between the upper and lower bounds of the 95% confidence interval. The spread of attributed deaths (174–490) obtained in this way is not a statistical confidence interval, but it might be viewed as a policy confidence interval. It includes almost all the other estimates shown in table 4(the effect of including active smokers is the only exception). We repeat that this is a snapshot picture of the impact of SHS. We have not attempted to model future trends although it is almost certain that smoking patterns and disease rates will change. However, as an indication of the magnitude of the gains that might be achieved by effective interventions, we have calculated the reduction in mortality caused by SHS that would occur if prevalence of exposure to SHS was halved in the home and reduced by 90% at work. We have not included in this estimate effects on the prevalence of active smoking—social policies that restrict smoking at work and in other public places have important secondary effects on smoking rates.29

There has been a pronounced fall in New Zealand in the number of cigarettes consumed per smoker since 1987, which may have led to a reduction in the intensity of exposure to SHS. (Estimates of mean cigarettes per smoker, based on tax paid consumption, have fallen from 23 per day in 1987 to 14 per day in 1999.24) Workplaces are also likely to be less smoky, on average, as a result of the changes that have occurred since the Smoke-free Environments Act. As a consequence, extrapolation of relative risk estimates derived from an earlier, heavier smoking period may overestimate the impact of SHS, but in our view there is insufficient information available to attempt to quantify the possible overestimate.

In conclusion, we estimate that past exposures to SHS are responsible for about 347 deaths per year in New Zealand. Some of these deaths result from exposures that occurred 10–20 years ago, and since that time prevalence of smoking has fallen, and so have exposures to SHS at work and in public places. This means that the number of “avoidable” deaths in the future is slightly less: we estimate that there would be approximately 325 fewer deaths in New Zealand each year if present exposures to SHS ceased altogether. This number is about 7% of the deaths in New Zealand estimated to be caused by direct smoking. There are many uncertainties associated with this calculation, but an indication of the magnitude of the problem may be obtained by comparisons with other causes of death. Our calculations suggest that the number of deaths caused by SHS lies between the number of people who die from melanoma of the skin (about 200 per year in New Zealand) and the number killed in road crashes (509 deaths in 1997).30