Socioeconomic factors influence tobacco consumption worldwide [8]. In the developed world, a strong inverse relationship between socioeconomic status and smoking exists such that the poorest and least educated populations are more likely to smoke [9-12]. While the few studies characterizing tobacco use in the developing world are equivocal, they generally support the findings observed in the developed world [9]. Moreover, it has been suggested that tobacco use has become widely prevalent in developing countries and the public health significance of smoking-related morbidity and mortality will continue to grow [9]. As the tobacco epidemic shifts from developed to developing countries, countries like Peru are in an ideal position for primary intervention programs prior to the realization of full range of smoking-related morbidity and mortality.

Peru has recently transitioned to a rapidly growing upper middle-income country [13]. Chronic disease profiles for many developing countries in transition are substantially and rapidly worsening [14,15]. Peru might well be expected to follow such an epidemiological transition involving a decreasing proportion of infectious diseases and an increasing proportion of chronic diseases. However, data on mortality and detailed studies of smoking are sparse in many developing countries where smoking has already become epidemic [9]. As a country in transition, accurate characterization and quantification of Peru's tobacco burden and use patterns is of public health importance. Previous studies have shown overall smoking prevalence in Peru at 22% to 38% and a recent report in 2005 from the Center for Information and Education for the Prevention of Drug Abuse found prevalence of daily and occasional smoking in urban Peru to be 8% and 20%, respectively [16-21]. However, study methodologies varied greatly and none focused on the low-income, peri-urban shantytowns. These communities that comprise a large portion of the population are less well studied because they are more difficult to access than their urban counterparts. Moreover, these communities are often associated with poor health infrastructure and receive little public-health attention from the government.

While estimates of tobacco smoke prevalence are available for multiple studies in Latin America, few of these studies have been conducted across multiple countries using standardized approaches. Two multi-country studies that used standardized questionnaires found that prevalence for tobacco smoke in Latin America ranging from 24% to 39% in one study [16] and from 22% to 45% in the other [22]. These studies, however, appear to have sampled populations that are either more urbanized [16] or with an overall higher educational level [22] than would be expected from peri-urban communities, which represent a larger share the urban populations in many Latin American cities.

Recent studies have shown current smoking rates of 27% in Lima, which is substantially higher than the rates we observed [16,17]. The disparity between values for smoking prevalence we observed in Pampas de San Juan de Miraflores and those observed in previous studies may partially stem from differences in the manner in which smokers were classified in each study. Information regarding methods of determining smoking status was unclear in several prior studies of smoking in Peru [19,31]. Other studies have been more explicit. The Center for Information and Education for the Prevention of Drug Abuse focused on urban centers. They reported lifetime prevalence, but broke down smokers into those who smoked daily, those who smoked at least once per week, and those who smoked less than once a week but greater than once a month [17,18]. To classify subjects into smoking categories, the PLATINO group implemented the Lung Health Study Questionnaire; and classified individuals who had smoked in the previous 30 days as current smokers. The PLATINO study found < 1% of the individuals identified current smokers reported not to be daily smokers [22]. In contrast, 88% of the individuals we observed in our study to be smokers were not daily smokers. The CARMELA study classified individuals as current smokers if they were current daily or occasional smokers with a lifetime smoking history of at least 100 cigarettes [16]. Standardized methods for collecting and reporting data on smoking status, such as those employed by the Global Adult Tobacco Survey and CARMELA groups will help to provide a clearer picture of the burden of tobacco smoking worldwide. We suggest that using a cutoff of at least one cigarette per day is a more appropriate measure of current smoking status and avoids overestimation that may occur by simply asking participants if they had smoked in the last 30 days or had a history of a lifetime consumption of ≥100 cigarettes. Despite slight differences in our approach to the quantification of smoking, the number of current daily smokers in our study was remarkably low.

The low smoking prevalence we observed in adults ≥ 40 years of age in Pampas San Juan de Miraflores is likely due to several factors that influence smoking behaviors that need to be investigated further. It has been previously suggested that providing health information and prominent warning labels are highly relevant tobacco control policies for low-middle income countries [32]. The vast majority of the study population understood that smoking could cause serious health problems including lung cancer, heart attacks, and cerebrovascular accidents. While education is important in reducing the burden of tobacco use, the economics of tobacco consumption likely serves as the principal determinant of tobacco smoking behaviors. San Juan de Miraflores is an economically depressed community, with a household income of less than 400 USD per month. In populations with low socioeconomic status, individuals are more responsive to changes in prices of cigarettes (Townsend et al., 1994). It has been suggested that price increases of 10% would be the most cost effective intervention for reducing smoking prevalence and smoking-related mortality and that price interventions would be particularly effective in low and middle-income countries [32,33]. Hence, taxation in developing countries would likely lead to decreased smoking prevalence in poor peri-urban communities. Moreover, as income levels and development continues to improve communities like Pampas de San Juan de Miraflores, longitudinal studies tracking tobacco costs versus smoking prevalence in a representative group of the population would likely clarify the relationship between economics and smoking in this community.

The majority of people observed cigarettes being sold on the streets. These are often sold as single cigarettes or packs of five cigarettes. As cigarettes sold on the street are cheaper than large packs and more accessible, this is likely an area for effective policy intervention. By eliminating the sale of cigarettes on the street, the already low tobacco burden in this population could likely be further reduced. Moreover, cigarettes sold on the street are more accessible to young individuals and thus intervention might help reduce the incidence of smoking in this community.

The majority of study participants observed the sale of tobacco products on the street. Strategies aimed at preventing the sale of cigarettes on the street would likely lead to a reduction of smoking prevalence and incidence. While many strategies may be employed to prevent the initiation of smoking, we suggest that education, taxation, efforts to curtail the sale of cigarettes on the streets, and strict enforcement of recent laws prohibiting smoking in eating and drinking establishments, government buildings, and public transportation would be particularly beneficial.

By focusing on primary prevention, governmental and public health organizations can help maintain a low smoking prevalence in poor, peri-urban communities in Lima, Peru. Further epidemiological studies in these populations may yield important information suggesting the reasons behind the low prevalence of smoking. We also suggest the need for repeat surveys over time to assess changing incidence and prevalence with the implementation of new tobacco control policies. These longitudinal studies may well elucidate the factors leading to the remarkably low smoking prevalence in this community and may offer new strategies to be implemented on the international stage.

A limitation of this study is that data were obtained via self-report. In the future use of quantitative measures of tobacco exposure such as exhaled carbon monoxide, urine or saliva cotinine, or hair nicotine would allow us to more definitively characterize tobacco exposure in this population. However, self-reported smoking status is the primary measure of smoking status for research and policy, and it provides both a cheaper and widely accepted indicator. Another limitation of this study is its cross-sectional nature. Longitudinal studies of this population in the future might yield insight into trends in tobacco prevalence and incidence and the outcomes of current and future tobacco control measures. Third, we focused on a group of adults ≥ 40 years of age and did not include adolescents and younger adults who may be at a high of risk of tobacco use. While younger people may be smoking at greater rates and have younger ages of onset, we were interested in first estimating the current level of pack-years which requires a long history of smoking; and, second, we wanted to compare the prevalence of smoking with the two large-scales studies PLATINO and CARMELA. This study was a pilot to better understand tobacco use and dependence in adults, the segment of the population most at risk for tobacco-related lung disease. Our group is conducting a 3000-person study in adults ≥ 35 years of age to determine lung function decline and we sought to find the best epidemiological measure of tobacco use in our region.

Our study has several strengths. First, we obtained adults ≥ 40 years were randomly selected. Therefore, we are confident that our sample is representative of the study population and largely representative of the population in Lima. Second, we used the well-validated Global Adult Tobacco Survey and Fagerstrom Test for Nicotine Dependence, both which provide a standardized set of questions that can be directly comparable with other studies. Third, we surveyed a relatively large number of male and female participants so we are confident that the low prevalence of smoking is not due to a small sample size. Finally, we do not believe that our findings are attributable to low reporting. In previous studies, self-reporting smoking in adults has been shown to be accurate in most studies [34,35].