Systematic review of the link between tobacco and poverty.

CIAPPONI, AGUSTÍN; BARDACH, ARIEL; GLUJOVSKY, D; ARUJ, PATRICIA; MAZZONI, AGUSTINA; COMANDE, D; GIBBONS, LUZ; LINETZKY, B; CASETTA, B

World Health Organization

Año: 2011 p. 227

978-92-4-150054-8

BACKGROUND This study explores the link between tobacco use and poverty, as well as the broader relationship between income, tobacco use, and tobacco-related health consequences, using a meta-analysis of existing research literature. An estimated 5 million deaths are caused by tobacco each year, with this fi gure expected to reach more than 8 million per year by 2030 given current trends in tobacco use. The proportion of this burden borne by people living in low- and middle-income countries at that time is expected to be above 80%. Many of the risks to health and life caused by tobacco consumption develop over a long period of time. However, tobacco use can also infl ict immediate harm on users and their families ? for example, when scarce family resources are spent on tobacco products instead of other essential needs. Even a small diversion of the resources of poor families who live at or below the edge of poverty can have a signifi cant impact on their health and nutrition, and in many countries the percentage of total expenditures allocated for tobacco products was highest for the lowest-income households. The association between socioeconomic position and health risk factors varies over time and between regions of the world. Smoking is acknowledged to be a contributor to differences in mortality and morbidity between socioeconomic groups, especially in key diseases. A signifi cant variation in the prevalence of use of tobacco is based on income level, in addition to other variables including ethnicity, altitude of residence, occupation, and religion. The aim of this study is to assess the association between income level and tobacco consumption, tobacco expenditures, and morbidity and mortality attributed to tobacco. SELECTION CRITERIA Type of exposure: Income level categories (low, middle and high), determined only by validated methods of direct assessment; papers where indirect assessment was performed were not included. As defi nitions varied between authors, strata of income level were interpreted as an income gradient. Type of outcome measures: Prevalence of current smokers; intensity of tobacco consumption; incidence of death attributed to tobacco; disease attributed to tobacco; household expenditure on tobacco. Types of study designs: Observational studies and baseline or control arms of intervention studies published in the last 20 years. Types of participants: General populations of different income levels around the world. SEARCH STRATEGY Multiple electronic databases were searched systematically, including MEDLINE, EMBASE, CENTRAL, SOCINDEX, AFRICAN IDEX MEDICUS and LILACS. We have also reviewed the International Tobacco or Health Conference Paper Index from 2006. The authors personally contacted key referents in tobacco control to obtain unpublished information and referrals to other key researchers. They also consulted the web pages of numerous tobacco control agencies, as well as contacting known Tobacco Control international networks in search of grey literature and contact information for key researchers. METHODS Two independent researchers per citation pre-screened titles and abstracts of all studies retrieved to identify those that could be included. Disagreements were solved by consensus, and fi nal decisions were made by the review team in the case of continued discrepancies. After 20% of the citations were screened, only one researcher per citation performed the screening because there was more than 90% agreement between cases with two researchers. The authors obtained the full text of all articles that were not excluded, and two independent researchers assessed the full text of selected articles to confirm their classifi cation and evaluate whether they met the inclusion criteria or not. Any discrepancies were solved by consensus, with the review team making the fi nal decision. If data from included studies were unclear or insuffi cient, the author(s) were contacted, and if it was not possible to obtain necessary information the article was excluded. Data Collection: An electronic chart, previously tested in a pilot study, was used to collect data. One reviewer extracted data from the included studies, and a second one checked this data. Methodological Quality Assessment A tool for assessing susceptibility to bias in observational studies was developed. With a modified STROBE (Strengthening the Reporting of Observational studies in Epidemiology) checklist for cross-sectional studies, together with key methodological papers, an algorithm was programmed in an Excel spreadsheet to assess the quality of the studies, and another algorithm was used for identifi cation of the study design. Pairs of reviewers independently assessed the risk of bias, and discrepancies were solved by consensus of the full work team. ANALYSIS A random effects meta-analysis was performed using Stata 8.0 to calculate summary odds ratios (OR) based on adjusted OR and confi dence intervals, or equivalent data as coeffi cients (β) and standard errors (SE), presented in the included studies. The random effect model was used, considering important possible sources of heterogeneity. Statistical heterogeneity was evaluated using the I(2) statistic and subgroup and sensitivity analyses were performed to evaluate potential heterogeneity. When there was evidence of substantial statistical heterogeneity, the following preplanned subgroup analysis could be performed: decade of dataset, continent, WHO region, mortality rate stratum, risk of bias, gender and age group. We also performed a sensitivity analysis considering only prospective studies, and excluding studies with methodological fl aws. We presented summary and descriptive statistics when meta-analysis was not possible (that is, the case of unadjusted smoking prevalence and household expenditure on tobacco). RESULTS A total of 9575 references were initially retrieved. After eliminating duplicates and screening the full text of these references, 765 studies were retrieved for detailed evaluation. From the selected articles assessed by full text, 137 that met the inclusion criteria were selected, together with 17 narrative reviews. Out of these 137 studies, 118 were cross-sectional (86.13%), 13 (9.49%) were prospective studies, and four (2.92%) were case-control studies. From the 137 included studies, 94 were subject to a meta-analysis of current smoking by income level and 17 to a meta-analysis of death or diseases attributable to tobacco by income level. Five studies were not included in the metaanalysis because their quality scores were evaluated as a ?high risk of bias,? while the others were excluded because of the absence of adjusted data. A total of 125 papers reported smoking prevalence data. A total of 31,146,096 people were included in the analysis. The median of the mean age in all the studies reporting them was age 41, from those who were 15 or more years old. The median current smoker rate was 27%, ranging from 2.5% to 73.7%. In the analyzed population, low-income people smoke more than higherincome people (OR 1.48, 95%CI 1.38-1.59). This result is seen in each of the evaluated world regions except the Eastern Mediterranean Region (EMRO), and especially in those reports performed beyond the year 1990. Considering only studies that reported results in three income categories, a gradient is shown, with the highest tobacco use prevalence in the lowest income level versus the high income level (OR 1.54, 95%CI 1.39-1.72), and a less marked increase prevalence in the middle income level people versus the high income level (OR 1.25, 95%CI 1.16-1.33). This trend was observed both in female and in male populations. A total of 20 studies (N=987,885) reporting data for tobacco-attributable diseases by income were analyzed. The most common tobacco-attributable diseases evaluated were: low birth weight for gestational age (LBWGA) (5 studies); coronary heart disease (4); cardiovascular death (3); periodontitis/ tooth loss (3); chronic obstructive pulmonary disease (COPD) (2); all causes of death (2); lung cancer (2); and stroke (1). Only one study used as independent variables the current smoking status and the income level category, showing the independent effect of smoking on LBWGA by income category. The other studies only evaluated the independent effect of income category on different outcomes of known relationship with tobacco exposure, adjusted by current smoking status and other variables. No study showed statistically significant higher odds of tobacco-attributable diseases at increasing income strata. Three studies showed no statistically signifi cant differences between strata. All the other studies exhibited statistically signifi cant higher odds ratios at decreasing income strata. The median proportion of tobacco spending related to total expenditures was 10.7%, 3.7%, and 1.8% in low, medium and high income levels respectively. DISCUSSION This review demonstrated an inverse relationship between income level and tobacco use prevalence, particularly in the last two decades. This coincides with the social gradient of cigarette diffusion taken into account by different authors. In the 1990s, the tendency to smoke by the poorest was well established by the literature, and this has been further confi rmed in the latest studies, despite a heterogeneous representation of continents and countries. This trend was consistent in all continents, with Oceania having the highest association. EMRO was the only region not showing this trend, although it was represented by just two countries, Pakistan and Saudi Arabia. There is clear evidence in America as well as the Western Pacifi c Region that smoking prevalence in low-income groups is approximately 50% higher than in higher income ones. This social gradient is confi rmed when middle and wealthy strata are compared, resulting in intermediate prevalence in the middle category of income. The results also show a consistently higher prevalence inthe poorest smokers for both genders, even though some other analyses state that the infl uence of economic strata on women should be less due to later incorporation of tobacco consumption. Those studies evaluating age categories showed a greater impact for this in people under age 44. An increased susceptibility to tobacco related illnesses was also found in low income groups, especially in all cause mortality, lung diseases and low birth weight. This effect was not as evident for cardiovascular disease and coronary disease, but became statistically signifi cant after performing sensitivity analyses. CONCLUSIONS This was an exhaustive and methodological rigorous systematic review examining the real magnitude of the impact of tobacco use by income level, which was previously unknown. This complex issue deals with a great heterogeneity of exposure and outcome variables, as well as populations and settings. Nevertheless, this study presented a solid base from which to support its conclusions of an inverse relationship between income level and tobacco use prevalence, and its related consequences. Greater efforts to reduce tobacco use among poor people are clearly needed. This research may be useful for policy makers as well, to improve strategies in tobacco control and inequity.