Lung cancer is the cause of death that most directly reflects the (historic) burden of smoking. Figure 1 shows lung cancer mortality rates by ethnicity and household income, for the 1980s and 1990s, as calculated from the New Zealand Census-Mortality Study (NZCMS) that uses linked census and mortality datasets covering millions of person-years of observation 17. Lung cancer mortality rates among Māori were over four times the non-Māori/non-Pacific rate for women and over three times for men (for 1996–1999). The rates for Pacific people were also relatively high (at over 2 times for men and 1.4 times for women, compared to non-Māori/non-Pacific). Over the same 1981–1999 time period, the inequality in male lung cancer mortality rates by household income persisted despite a decline in deaths in all income groups. However, in women there was a large increase among the low-income group compared to a decrease among the high-income group. Over this time period there was also an overall increase in ethnic inequalities in mortality rates from lung cancer (in both men and women). The authors of this study concluded that these inequalities will probably widen in future decades – unless there is concerted public health action. All these patterns are consistent with differently phased tobacco epidemics 18 by ethnicity and SEP, resulting in changing inequalities in lung cancer over time.

The very large inequalities in lung cancer mortality by ethnicity are probably greater than would result alone from historically (still large) differences in smoking prevalence, pointing to other independent, and likely interacting, risk factors. These may include such factors as: varying passive smoking exposure 1415, environmental pollution exposure 19 and hazardous occupational exposure such as from asbestos 20. Diet may also be relevant to this differential (eg, given evidence around fruit intake lowering lung cancer risk 21) and so might genetics given some New Zealand evidence for variability in nicotine metabolism by ethnicity 22.

Differential survival, due to differential access to care and more advanced stage at presentation will also contribute to ethnic inequalities in lung cancer mortality. Māori are more likely than non-Māori to have lung cancer identified at a later stage and have a lower survival rate after diagnosis 23. Possible factors involved include access to specialised cancer services and the quality of care received 24.

The NZCMS includes active smoking data for the 1981–84 and 1996–99 cohorts, allowing direct estimations of the active smoking-related burden within and between social groups. The measure of smoking is simply "never", "ex-" and "current" smokers, meaning there will be inevitable misclassification biases of smoking that probably lead to modest underestimates of the contribution of active smoking.

Table 1 shows population-attributable risk percents (PAR%) for 45–74 year olds in 1996–99 from NZCMS output. They are the percentage reduction in all-cause mortality that might be expected if, in a counterfactual world, all people who were either current or ex-smokers had actually been "never" smokers. Because of slightly (and necessarily) different methods between the ethnic and educational group analyses (see footnotes to Table 1), they are not fully comparable. Nevertheless, they do robustly point to the following conclusions:

• active smoking is a major contributor to all-cause mortality in all educational and ethnic groups,

• about a quarter of 45–74 year old all-cause mortality in each educational group is due to active smoking. This figure is slightly higher in lower educational groups, and slightly less in higher educational groups,

• about a third of 45–74 year old all-cause mortality among non-Māori/non-Pacific is due to smoking. This figure is slightly higher among males, and slightly less among females,

• a fifth to a quarter of 45–74 year old all-cause mortality among Māori is due to smoking.

These above estimates for Māori are less than expected based on previous Ministry of Health estimates that a third of all Māori deaths (not just 45–74 year olds where a greater proportion of deaths will be due to smoking than at other ages) are due to tobacco 25. There are two key reasons why the more recent Ministry of Health estimates for Māori are likely overestimates. First, other recent work from the NZCMS finds that the relative risk of death associated with tobacco use varies by ethnic group and over time 26. All-cause rate ratios (RRs) for mortality associated with smoking were significantly greater within non-Māori/non-Pacific than within Māori: 2.22 compared to 1.51 respectively for men, and 2.20 compared to 1.45 respectively for women (for 1996–99). One of the likely reasons for this rate ratio heterogeneity is the greater role of competing non-tobacco causes of mortality among Māori and Pacific peoples. But other factors may also be relevant eg, different patterns of what cigarettes are used and how they are smoked. Second, the Ministry of Health estimates have used the standard WHO/Peto methodology whereby lung cancer mortality rates are used to estimate the total mortality impact of smoking. However, as mentioned above, Māori lung cancer mortality rates are higher than would be expected on the basis of tobacco smoking alone, which would lead the WHO/Peto method to overestimate the total tobacco-related mortality burden among Māori.

What of the contribution of smoking to gaps in mortality between ethnic and socioeconomic groups? Poisson regression analyses adjusting for smoking reduced the all-cause mortality RRs for men with nil educational qualifications compared with men with post-school qualifications from 1.34 to 1.29 in 1981–84 and from 1.31 to 1.25 in 1996–99. This equated to 16% and 21% reductions in relative inequalities respectively. The equivalent results for women were 3% and 11% reductions in relative inequalities for these time periods. Such higher mortality rates for men and women with poorer education were due to the impact of smoking on cardiovascular, cancer and respiratory deaths. The patterns identified in this study were considered to reflect the historically differential phasing of the tobacco epidemic by sex and SEP.

The most recent NZCMS study on smoking examined its contribution to ethnic inequalities in mortality 27. It found that the apparent contribution of smoking to mortality differences between Māori and non-Māori/non-Pacific was greatest for women in 1996–99 (8% reduction in standardised rate difference), and had increased from 1981–84 to 1996–99 for both men (from -1% to 5%) and women (from 3% to 8%). That is, the contribution of smoking to ethnic gaps (in percentage terms) is notably less than for socioeconomic gaps. But a fuller understanding of this requires also considering the actual underlying mortality rates.

Figure 2 attempts to pull together the above findings for 1996–99, and addresses the need to consider absolute mortality rates and absolute differences in mortality rates (as well as relative risks and percentage contributions). It shows actual mortality rates by ethnicity and education (partitioned by the proportions estimated to be smoking- and non-smoking related). A floating column representing the gap in mortality rates is included (again partitioned into smoking and non-smoking-related components). The figure should be considered indicative only. There are unavoidable differences in methodology between: the ethnic versus educational analyses as stated above; the determination of PAR% within ethnic and socioeconomic group versus the percentage contributions to gaps between ethnic and socioeconomic groups; and standardisation versus regression methodologies for different components of analysis behind the figure. Nevertheless, there are a number of robust findings:

• mortality rates for Māori are 2–3 times greater than non-Māori/non-Pacific, compared to an approximately 40% higher mortality for people with no qualifications compared to post-school qualifications;

• in absolute terms, the mortality rate attributable to smoking among both Māori and less educated groups is considerably greater than among non-Māori/non-Pacific and post-school educated people, respectively – a different perspective from considering the PAR% estimates in isolation;

• in absolute terms, the gap in mortality rates between ethnic groups attributable to smoking is as great or greater than between educational groups – a different perspective from considering just the percentage contribution to gaps.

As mentioned already, the estimates above are likely to be modest underestimates due to likely non-differential misclassification bias of smoking status. The analyses did not include the impact of exposure to second-hand smoke, which is more common among Māori and lower socioeconomic groups 1415. This would mean that percentage contributions of active and passive smoking combined to mortality are probably greater than given above. Figure 2 also clearly demonstrates that ethnic gaps in mortality not explained by smoking are much greater than socioeconomic gaps in mortality not explained by smoking. This points to other determinants of health (eg, differential access to health services, racism) that must be more important for ethnic inequalities than socioeconomic inequalities in health.

There have been arguments in the New Zealand context for having a Tobacco Authority type agency 38 with a public health mandate to control the marketing of tobacco. This approach has also been proposed by others internationally 3940. Such an agency could allow the nicotine market to be realigned to strongly favour (in terms of price and availability) pharmaceutical-grade nicotine, over smoked forms of tobacco. Such a market could also favour reduced-harmed tobacco products such as nasal or oral snuff, though the idea of health sector endorsement of such a market is controversial in New Zealand. Nevertheless, a switch to snuff could plausibly facilitate reductions in overall harm to the health of users 4142 and facilitate quitting 4344 and therefore health inequalities attributable to tobacco use. Any shift to smokeless forms of nicotine or tobacco would also be likely to reduce the health inequalities associated with different levels of exposure to second-hand smoke. A key aspect for maximising the impact for reducing inequalities would be the extent to which the price differential could be managed given the suggestive evidence that low-income and Māori populations are more price sensitive (see the discussion around taxation below). This shift to alternate forms of nicotine or tobacco could also be accelerated by increasingly tight restrictions and raising the price of smoked tobacco.

In the long-term however, if a large proportion of low-income New Zealanders remained dependent on pharmaceutical nicotine or snuff – then this could still represent a drain on their financial resources (if the price was not kept relatively low). Such issues could be further explored by modelling work and studies on the price elasticities and acceptability of these alternative nicotine products to low-income New Zealanders.

Finally, enhancing tobacco regulation will inevitably be a political decision. Debate and discussion is therefore essential not only among the tobacco control community, but also politicians and the public at large. It is imperative that tobacco control advocates reinforce at all times that ridding New Zealand of tobacco smoking will benefit all sectors of society, and reduce inequalities – a win-win situation.

New Zealand could more intensively pursue all the key components of a comprehensive tobacco control programme (eg, tax policy, smokefree environments, and smoking cessation support – as detailed below). All these could be funded by increasing the relatively low level expenditure on tobacco control, currently less than 3% of tobacco tax revenue 45. Added to these interventions could be litigation against the tobacco industry by government and a more strongly industry-focused approach to tobacco control 46.

There is some international evidence that tobacco taxes are relatively more effective in reducing tobacco consumption among low-income or poorly educated populations 474849. There are also some New Zealand data to support this differential benefit for low-income groups and Māori 5051. Other New Zealand modelling work 52 provides some justification for tobacco taxation, as it indicates that the harm from smoking for low-income New Zealanders greatly exceeds the likely harm from financial hardship that is associated with the tax. Despite this, there is concern regarding the potential for increased economic hardship (with subsequent impacts on health) among low-income groups from increased tobacco taxation in the future. If tobacco taxes were to be increased, it would be necessary and ethical 53 for a greater proportion of the tax revenue to be used for smoking cessation support, especially for Māori, Pacific and low-income New Zealanders. Indeed, this country previously introduced a programme of providing heavily subsidised nicotine replacement therapy in the year of the last tax increase (ie, 2001).

In late 2004 a new smokefree law came into effect in New Zealand and the evidence to date is that it is working well 545556. This law covers all indoor workplaces and hospitality settings which suggests that it should reduce exposure to second-hand smoke among low-income workers and patrons of venues such as bars, clubs, and casinos. In addition, recent government-funded mass media campaigns may be contributing to increasing smokefree homes among low-income families (though post-campaign follow-up data have not yet been published). There is some indirect evidence for benefit from such campaigns internationally on smokefree homes 57.

The prevalence of smoking in cars amongst people from more deprived areas in New Zealand is significantly higher than less deprived areas 58. While some campaigns have incorporated the hazard from smoking in cars in New Zealand, these have been of low intensity. Nevertheless, laws have now been passed in other jurisdictions (eg, Arkansas, Louisiana and Puerto Rico) and this approach could be considered in New Zealand.

There is evidence that mass media campaigns (both generic and those designed by Māori) are effective in stimulating calls to the national Quitline from Māori and other low-income New Zealanders who are the priority audiences for this service 5960. Evaluation work has also shown that culturally appropriate mass media campaigns are regarded as acceptable to a Māori audience 6162. A recently launched media campaign with a Pacific peoples focus 63 has also successfully stimulated increased call rates to the Quitline [Personal communication, Helen Glasgow, Director Quit Group, 28 March 2006].

The national free-phone Quitline service has been successful in reaching a Māori audience 6064. The popular uptake of heavily subsidised smoking cessation services provided via the Quitline 65 also suggests that it is reaching low-income New Zealanders. However, the idea of building long-term relationships with smokers and recruiting them as volunteers to promote smoking cessation services 66 has yet to be tried in this country.

Culturally appropriate smoking cessation services such as the Aukati Kai Paipa services for Māori women have been evaluated and found to be acceptable and effective 67. New Zealand has also had some success with running quit and win contests 68 which may differentially appeal to low-income smokers (although this aspect has not been studied). A randomised controlled trial of bupropion for Māori smokers has reported successful smoking cessation outcomes in this population 69.

Other countries are also trying to reduce health inequalities associated with tobacco. A review of 16 studies that aimed to reduce smoking in low-income groups found that half of these had demonstrated effectiveness 70. Out of another nine studies that were not actually targeted at low-income groups, in five of these the intervention was at least as effective in low as in high-income groups. Nevertheless, in four of the studies, including one New Zealand study 71, the intervention was less effective for those in low-income groups. In particular, there is evidence from the United Kingdom that smoking cessation services are reaching disadvantaged communities 72, that the services to such communities are qualitatively better 73, and that these services are reducing inequalities in smoking prevalence rates 74.