Lung cancer mortality in Sub‐Saharan Africa

In a recent paper in the IJC,1 Ferlay et al. report on the results of the latest update of GLOBOCAN, released on 1st June 2010 by the International Agency for Research on Cancer.2 GLOBOCAN 2008 is an important resource for cancer research and health policy. Mortality and incidence estimates of crude rates, age-standardized rates and absolute number of cases by country and sex for the year 2008 are provided. We agree with their conclusion that "Already the majority of the global cancer burden now occurs in developing countries, these proportions will rise in the next decades if rates remain unchanged." Given the demographic transition that is ongoing in developing countries, the absolute numbers will almost certainly rise in the future. In Sub-Saharan Africa (SSA), population-based data on lung cancer incidence or mortality are not available, except South Africa and smaller island states. GLOBOCAN uses local registry data if available for incidence estimation. For 22 African countries, the incidence estimates "represent simply those of neighboring countries." Absolute numbers are derived using United Nations Population data.3 The authors acknowledge these limitations and discuss these appropriately.1, 2, 4 According to GLOBOCAN, the lowest lung cancer mortality worldwide is seen in SSA (with the exception of South Africa). Their estimated number of lung cancer deaths in SSA (Africa without Algeria, Egypt, Libya, Marocco, Tunesia, West-Sahara) has risen from 11503 in 2003 to 13732 in 2008. We think that the reported numbers for lung cancer and possibly for other cancers as well underestimate the true numbers. We have developed and validated an alternative estimation procedure which we think is appropriate in this setting and applied it to lung cancer.5 Briefly, we combine age-, country- and sex-specific smoking prevalence data (available from WHO Global InfoBase) with lung cancer mortality rates in non-smokers (estimated from several international studies) and with relative risk estimates for tobacco smoking (taken from large cohort studies).5-7 This yields age-specific lung cancer mortality rate estimates for country k via, λjk by , where λj0 is the estimated lung cancer mortality rate in black non-smokers in age group j, pjks the proportion of smokers in dose group s, age group j and country k, and RRs the relative risk associated with that smoking dose. We obtained estimated numbers of lung cancer deaths for SSA by combining these rates with UN population figures (year 2005). We applied this method to data from Indonesia, Vietnam and Ethiopia and found very good agreement between our estimates and the GLOBOCAN estimates for the two Asian countries, where the quality of cause of death statistics is much better, and a strong disagreement for the African country,2 where our estimate was 2.5 times larger. In an extended ongoing analysis, we refined our model and applied it to data from Germany, UK and Canada where also a very good agreement between the model estimates and the reported numbers was found.8 Here, we have estimated mortality rates and absolute number of deaths for SSA. Figure 1 shows the rate estimates by GLOBOCAN and by the above method. While our estimates show some, but not very striking differences between countries for male and female lung cancer mortality, the GLOBOCAN estimates show high mortality rates for South Africa and very low rates for all other countries. With the exception of South-Africa, the ratios between the country-specific age-adjusted rate estimates of both methods are above one, with ratios up to 83·0. Our estimates add up to 43,498 deaths in 2005, and to 46,699 in 2008 (using the UN population projection, similar as in GLOBOCAN), about threefold higher than in GLOBOCAN. Sensitivity analyses assuming lower smoking prevalences and lower rates in non-smokers confirmed this magnitude. Assuming a non-smoking population in SSA, we estimate about 27,500 deaths per year, and assuming the same rates in blacks as in whites, we estimate 31,132 deaths per year. To obtain the GLOBOCAN numbers, the population mortality rates in SSA at large would have to be 50% lower than the non-smoker rates. Ferlay et al.1 acknowledge this; however, they argue that the rates are based of urban areas, and that the rates in rural areas are likely to be lower, as found in India, implying an overestimation of the country specific rates. We do not think this is the case. For example, we found in a recent study in Burkina Faso, West Africa, that the life expectancy in a remote rural area is not lower, but slightly higher than the reported country average. Map of estimated age-adjusted lung cancer mortality rates in Sub-Saharan Africa by sex according to GLOBOCAN and our method. We believe that the local registries on which the GLOBOCAN estimates are based are not sufficiently reliable at present and are the main cause of the observed differences. The only African country with reliable mortality data, South Africa, has high rates according to GLOBOCAN. Since smoking prevalence is similar in many other African countries,6 and since smoking is the dominant factor for lung cancer, similar rates should be expected. Longitudinal data on smoking prevalence in Africa are not available; therefore, we cannot rule out that smoking in South Africa started earlier than in other countries, thus yielding to higher rates at present. This, however, cannot explain such a difference. We also consider it unlikely that other risk factors for lung cancer, such as occupational exposures, can explain the differences. Other estimates are presented in the Global Burden of Disease report of the WHO, 2004 update. Here, another estimation method was used, and the reported projection figures for 2008 are 21,843 male and 6,121 female lung cancer deaths for the WHO region Africa, thus being closer to our estimates.9 The demographic transition is ongoing in SSA and the proportion of population above 60 is quickly increasing, and the numbers will therefore increase further. Smoking prevalence has increased over the last decade in Africa.6, 10 As a result, chronic disease prevention becomes more relevant. In summary, we agree with the conclusions of Ferlay et al. Our results, however, suggest that the burden is even higher. More prevention efforts are therefore needed for the future, in particular tobacco control. Heiko Becher*, Volker Winkler*, * Institute of Public Health, University of Heidelberg, Heidelberg, Germany.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.