Epidemiology of Chronic Obstructive Pulmonary Disease in Brazil: a systematic review and meta-analysis.

Chronic Obstructive Pulmonary Disease (COPD) is among the leading causes of morbidity and mortality in developing countries; however, few systematic reviews are available in the literature. This review examines the prevalence of COPD in the Brazilian population. For this purpose, a systematic review and meta-analysis was conducted of epidemiological observational studies indexed in the databases PubMed, Cochrane, Ovid, Scopus, ScienceDirect, SciELO, Lilacs, and Google Scholar published up to May 2018. The prevalence was estimated using a random effects model. Of the 1,182 articles identified, 12 were eligible. The prevalence of COPD in Brazil was 17% (95%CI: 13-22; I2 = 94%) and the region with the highest prevalence of COPD was the Center-western region (25%), followed by the Southeastern region (23%). The Southern region had the lowest prevalence among the studies (12%). We found that Brazil has a high occurrence of COPD, higher than the estimates of Latin American and world population Hence, additional studies are necessary to support intervention strategies, as well as formulation of specific public health policies to control and prevent COPD.

introduction Chronic Obstructive Pulmonary Disease (COPD) is a public health problem worldwide and is among the leading causes of morbidity and mortality, especially in developing countries 1,2 . In 2017, it was considered the fourth leading cause of death in the world and is estimated to become the third leading cause of death in 2020 1 . This severe condition is characterized by the persistence of respiratory symptoms, due to airflow limitation, such as dyspnea, and cough with or without expectoration 1,2 .
Epidemiological studies have shown that COPD is a multifactorial disease and is associated with exposure to environmental pollution, unfavorable socioeconomic conditions, and genetic abnormalities. Among the environmental pollutants, the exposure to smoke from tobacco and burning biomass 1,3 is associated with COPD.
In 2015, more than 37,000 people died in Brazil due to COPD, according to the Mortality Information System of the Unified Public Health System 4 . Data from a 9-year follow-up study showed that the rate of new cases diagnosed in the city of São Paulo (1.4% to 4%) was similar to the rates found in Europe, Asia, and North America 5 . This demonstrates the need to study this subject to contribute to the development of disease prevention programs.
The disease is underdiagnosed and undertreated 2 . In addition, the literature on this disease in Brazil is sparse, which makes it difficult to estimate the impact of COPD on quality of life. This is the first systematic review with meta-analysis to evaluate the prevalence of COPD in Brazil.

Methodology
This systematic review study was developed according to answer the question: "what is the prevalence of COPD in the Brazilian population?".
To answer the question, a review protocol was developed and registered at the International Prospective Register of Systematic Reviews (PROSPERO).

Eligibility criteria
Observational (cross-sectional, cohort, and case-control) studies conducted in Brazil and evaluating the prevalence of COPD in adults were considered eligible.
Evaluated studies must meet the following criteria: FEV1/FVC ratio < 0.70 obtained through spirometry; mean ages; and city where the study was conducted. Only studies carried out from the year 2000 onwards were considered in meta-analysis. No limits on language or publication status were applied. Studies that evaluated adults with specific health conditions (myocardial ischemia, pulmonary hypertension) were excluded.

selection of studies and extraction of data
Following the eligibility criteria, the authors MMC and MP-S independently selected the articles in two steps, evaluating the title and abstract and, later, reading the full text. Disagreements were resolved by consensus.
To extract the data, a spreadsheet was prepared to record information about: the authors; publication year; city; region of the country; age group; sample size; prevalence of COPD; mean ages; standard deviation of age; percentage of prevalence, and equipment used.
For the articles that did not contain all the information, the author was contacted, at least twice, for the necessary clarifications. However, in three studies [23][24][25] it was not possible to obtain all because the authors did not respond.

Evaluation of the methodological quality of included studies
To assess the risk of bias, a tool developed by Hoy et al. 26 was used, which adopts the criteria 1) probabilistic or census sampling; 2) adequate sampling source (official census, school census, among others); 3) sample size previously calculated; 4) adequate post-bronchodilator spirometry FEV1/ FVC; 5) adequate response rate (> 70.0%) and description of refusals; 6) presentation of confidence intervals and analysis of subgroups of interest; and 7) study subjects well described and similar to the research question. Each item received a score of 1 (yes) or 0 (no). Scoring on all items can generate an overall quality score ranging from 0 to 10. According to the scores, risk of bias is classified as low (> 8 points), moderate (6)(7)(8), or high (≤ 5).

Data analysis
The primary outcome was the prevalence of COPD, with a 95% confidence interval (95%CI).
The prevalence of the disease was estimated for the total population and defined subgroups of interest: women; sample size, region of the country, and risk of bias.
The DerSimonian and Laird method was used to estimate the variability parameter between the studies. The heterogeneity was assessed by the Cochran test and Q test, and its magnitude was ascertained by I-square (I 2 ) 27 . Data from the studies included in the meta-analysis were transformed through the logit function to satisfy the normality assumption of the meta-analytic model of random effects. The confidence intervals for the results of the individual studies were calculated using the Coppler-Pearson method.
Meta-regressions were used to identify the causes of heterogeneity using the Knapp and Hartung test to analyze the following variables: quality score, sample size, proportion of females and males, publication year, and region of the study. The publication bias was not evaluated since it is not adequate in the meta-analyzes of prevalence.
In all analyzes the p value < 0.05 was considered statistically significant. Statistical analyzes were performed using the STATA 12 program (Stata Corp., College Station, TX, USA).

Characteristics of the analyzed studies
The search strategies are presented in Figure  1. We identified 1,182 articles in the eight databases. After the removing the duplicate records, 47 remained for title and abstract analyses; 20 were selected for qualitative synthesis 2,12,22-24,28-42 and 12 were included in the meta-analysis 2,12,23-25,29,34-39 . The reasons for excluding the articles were the absence of the association measure (n = 2), absence of COPD prevalence in Brazil (n = 25).
The main characteristics of the selected studies are presented in Table 1. Most of the articles selected (67%) were published after 2010, with samples varying from 67 to 1,059 participants; 50% of the studies have a sample number above 300. The total number of participants was 4,812. For geographic regions of the studies, there was a higher concentration in the southern region (42%) of Brazil (Table 1).
A predominance of cross-sectional design (n = 12) was observed, as well as a predominance of articles with methodological quality classified as strong (n = 7; 58%). Articles from cross-sectional studies obtained a higher frequency of strong scores in risk of bias assessment (Chart 2).

Meta-analysis results
The results of the meta-analysis of COPD prevalence are shown in Table 1. The prevalence of COPD in Brazil was 17% (95%CI =13-22) (Figure 2). There was signifcant heterogeneity across the included studies (I 2 = 94%).
The meta-regression for the variables: risk of bias, sample size, proportion of female and male subjects, publication year, and region of the study analyzed did not explain the possible causes of the heterogeneity found in the analysis of the studies (p > 0.05).

Discussion
This study was the first meta-analysis to estimate the prevalence of COPD in Brazil, based on observational studies that used as a diagnostic criterion from spirometry for COPD.
The occurrence of COPD in this study was 17% (95%CI = 13-22), which is higher than the estimated 11.4% for the world population (95%CI = 8.4-15.0) 43 . In 2010, the World Health Organization estimated that the region with the highest prevalence of COPD was in the Americas (15.2%, 95%CI = 14.9-15.5) and the lowest in Southeast Asia (9.7%). Thus, the prevalence in Brazil exceeds the average of the world regions 44,45 .
Most studies report a prevalence of COPD ranging from 8 to 10% among adults 33 , with cigarette smoking the most important risk factor. However, occupational and environmental exposures are important for disease progression and onset, especially in developing countries. The latest statistics from COPD identify a gradual decline in some parts of the world in relation to morbidity and mortality. However, in Brazil, we found that the prevalence among adults and the elderly is significant (19%), which implies large economic and social expenditures, such as hospitalizations [46][47][48] . In the US, each exacerbation costs $88 to $7,757 19 . The estimated mortality in hospitalizations for COPD in the Caribbean ranged from 6.7% to 29.5% 44 . Therefore, COPD is a severe condition with substantial impact on the Brazilian public health system. In view of this, measures to prevent environmental exposures, such as biomass burning and programs to encourage smoking cessation, are essential to combat the current situation.
A possible explanation for the high prevalence of COPD in the Brazilian population may be related to the extensive use of wood stoves for domestic needs, because the use of solid fu-  A study published in 2014 analyzed COPD mortality in Brazil from 1980 to 2014. It found an increase from 1998-2004, followed by a decrease, with the highest rates concentrated in the Southeastern and Southern regions 48 . Another study that also evaluated COPD mortality between 1996-2008 showed higher rates among men (2.7%, 95%CI = 3.6 to -1.8) than in women (-2.0%, 95%CI = 2.9 to -1.0) 47 . These studies ver-ified that this disease has irregular affections rates in the population, which entail substantial costs to the Health System.
A study carried out in São Paulo showed that, among the 206,764 hospitalizations due to respiratory diseases between 1995 and 2000, COPD was the third main cause (5.8% of the total), with the higher rate in men (6.09% versus 5.53%) and the mean age was 57.7 years 48 .
The Central-Western region had the highest prevalence of COPD. Explanation for this record needs to be established, as few studies associate possible environmental exposures or population factors to the occurrence of COPD. However, the prevalence found in the Central-western and Southeast regions were similar. A possible factor associated with COPD may be exposure to biomass burning. Especially in the Southern and Southeastern regions of Brazil, the production of firewood and charcoal increased between 2016 and 2017 49 . The Northern and Northeastern regions were not represented in this study because there was no study published the fit the scope of this review. Thus, studies on the prevalence of COPD in these states needs to be carried out, because this region has major producers of risk factors for the disease.
In this meta-analysis, the prevalence of COPD in men was 16% (95%CI = 6-26), which is 1.02 times higher than in women (13%, 95%CI = 5-20). In the United States, the prevalence in women (5.4%) was higher than in men (2.8%). In other countries, such as China and Finland, COPD in both men and women is lower than the results of the present study 10,50 . Our result is similar to that found in the Americas, where the prevalence of the disease among men was 17.6% (95%CI = 14.6-20.6) and women 11.8 (95%CI = 9.7-13.9) 34 , as well as in the Caribbean and Latin America in 2010 36 , in which the prevalence in men was 1.76 times higher than women (10.4%, 95%CI = 7.6-13.6 vs. 18.2%, 95%CI = 14.1-22.6). The prevalence of COPD among men is estimated to be 14.3% (95%CI = 13.3-15.3) and women, 7.6% (95%CI = 7.0-8.2), which is lower than our results.
The results of this meta-analysis evidenced that COPD predominantly affects people over 55 years old. The lowest prevalence was recorded in the population of Novo Hamburgo, a city in the Southern Brazil, in contrast to Rio de Janeiro, in Southeastern Brazil, where the highest prevalence of the disease was described 36 .  38 8 The largest sample size in the studies occurred mostly after 2010. On the other hand, the highest prevalence recorded were concentrated in the period between 2004 and 2007, because they had a smaller sample size, as well as selection bias or inadequate sampling technique.
The great volubility in the quality of the studies attributed to the high probability of non-response bias, as the type of sampling was often not random. Therefore, many of the selected studies were considered to have moderate risk of bias. Thus, the appropriate sampling process as well as the methodological quality of the study should be considered in future studies to minimize the risk of bias.
Cross-sectional studies were predominant, as was the diagnostic method through spirometry, which was applied individually in most of the studies. In addition, all studies applied consolidated protocols in the literature to evaluate COPD. Thus, the spirometric result, in which the ratio of forced expiratory volume (FEV1) divided by forced vital capacity (CFV) is less than 0.7, according to the latest guideline of the Global Initiative for Chronic Obstructive Lung Disease 1 , which was the primary diagnostic criterion for COPD applied in the selection of studies. However, one study used medical diagnostic criteria and adherence to the recommended treatment for the disease 42 .
This meta-analysis has limitations. Possibly because there are few studies conducted in Brazil, and none in the Northern and Northeastern Regions. In addition, due to different age categories, it was not possible to estimate the prevalence of COPD by age group in Brazil. We also had difficulties in obtaining data from selected studies, due to the lack of response from authors of the studies included in the meta-analysis. Nevertheless, methodological rigor and well-established criteria make this study consistent.
Finally, the results found in this study indicate that COPD is a frequently occurring problem in Brazil, higher than the estimates of Latin American and world population. Nevertheless, new studies covering not only the criteria already mentioned, but also the Northern and Northeastern regions are necessary to extensively investigate the impact of COPD in Brazil and help to develop interventions.