Temporal trend of tuberculosis in Brazil Tendência temporal da tuberculose no Brasil Evolución temporal de la tuberculosis en Brasil

The objective was to analyze the temporal trend of tuberculosis incidence rates in Brazilian states and regions to identify patterns and inequalities. We carried out an ecological study of incidence rates per 100,000 inhabitants aged between 20 and 59 years with tuberculosis in Brazil and in their respective states from 2001 to 2017 according to annual percentage change and joinpoint regression. A significant decrease of -15.1% in the annual variation (95%CI: -27.2; -1.0) was observed in Roraima between the years 2003-2007. In the Central Region, Goiás registered an annual average variation of -2.5% per year (95%CI: -3.6; -1.3). Five out of nine Northeastern states had decreasing annual variations throughout the studied series. Espírito Santo and Rio de Janeiro were the states of the Southeast that presented decreasing between 2001-2017. In the Southern Region, the state of Paraná was the only one with the same annual variation of -3.7% (95%CI: -4.1; -3.2). In Brazil, the average annual percentage change was -1.8% (95%CI: -2.4; -1.1). The results of this study showed a decline in the incidence of tuberculosis in Brazil. However, regional and intra-regional differences were observed. Determining the trend pattern of tuberculosis incidence may assist in the planning and implementation of national tuberculosis control policies. Tuberculosis; Incidence; Epidemiologic Measurements; Regression Analysis; Uses of Epidemiology ARTIGO ARTICLE This article is published in Open Access under the Creative Commons Attribution license, which allows use, distribution, and reproduction in any medium, without restrictions, as long as the original work is correctly cited.


Introduction
Global confrontation of tuberculosis (TB) requires progress in prevention and care of patients in countries with high incidence, including Brazil. The percentage reduction in TB incidence rates was 1.5% between 2014 and 2015 in the world, but an average annual reduction of 4% to 5% is expected by 2020 to achieve the goals of the reduction strategy stipulated by the World Health Organization (WHO) 1 .
In the Americas, Brazil has a significant role in the fight against tuberculosis, although it presents characteristics common to the region, such as high levels of urbanization, social inequalities, and ethnic and cultural diversity. Other characteristics may help in understanding the Brazilian response capacity, among them: having the second largest Gross Domestic Product (GDP) among the countries of the Americas; ranking 15th in Human Development Index (HDI) among American countries, 5th in the list of countries with the highest income per capita in South America, and 6th in life expectancy, among Latin American countries 2 .
Since 2003, TB has been considered a priority disease in the political agenda of the Ministry of Health, as Brazil is one of the countries with the highest number of cases in the world. However, there are still barriers in the population's access to diagnosis and treatment, which are performed universally and free of charge by the Brazilian Unified National Health System (SUS). About 69,000 new cases of tuberculosis and 4,500 deaths were reported in 2016 2 .
In 2016, 66,796 new cases were diagnosed and registered in Brazil, of which 12,809 (19.2%) were cases of retreatment. Between 2007 and 2016, the incidence coefficient of the disease showed an average annual variation of -1.7%, changing from 37.9/100,000 in 2007 to 32.4/100,000 inhabitants in 2016 3 .
Operational indicators for the monitoring of TB control proposed by the National Tuberculosis Control Plan are important tools for assessing the trend of TB and its impact, reflecting the performance of health services and the quality of care provided for people with the disease 3,4 .
According to the national plan for elimination of TB as a public health hazard 2 , scenarios were defined to be worked in the following years with the objective of investigating their relationship with the reduction of TB incidence in Brazil. These scenarios present particularities that facilitate the planning of states and municipalities to combat the disease and assist in the identification of priorities.
The National Tuberculosis Control Program has implemented several actions to prevent and detect TB in the last decades. Studying the pattern of TB incidence in various parts of the country can assist in assessing the impact of the Program.
The objective of this study is to analyze the temporal trend of TB incidence rates in Brazilian states and regions to identify patterns and inequalities.

Method
This is an ecological study of the temporal trend of TB incidence rates per 100,000 inhabitants. The rates were constructed based on the total number of confirmed cases reported by the SINAN (Brazilian Information System for Notificable Disease) of people aged between 20 and 59 years in Brazil and in the respective states from 2001 to 2017.
Incidence rates were calculated by dividing the number of TB cases in the age group by the estimated population of the same range in each year of study, multiplied by 100,000. Population estimates were retrieved from a study sponsored by the Inter-Agency Health Information Network (RIPSA) and by the Brazilian Ministry of Health/Health Surveillance Secretariat/General Information Coordination and Epidemiological Analysis of the mentioned years as denominators per 100,000 inhabitants.
The annual percentage change (APC) was calculated using segmented regression (SR), allowing the turning points that indicate the variations of rates with greater significance in the increase of the line to be identified. The number of points required for the adjustment of each segment was automatically selected by the default configuration of the software (without logarithmic transformation of the response variable). The measure that explains several variations is the average annual percentage Cad. Saúde Pública 2020; 36(6):e00081319 change (AAPC). In situations where only one APC comprises the entire period studied, the AAPC corresponds to the APC.
The assumption of homoscedasticity, autocorrelation, and normality of distribution of residuals in the most suitable models for adjustments in each region were tested in order to verify the compatibility of the proposed analysis. For APC and AAPC, 95% confidence intervals (95%CI) were calculated. Data were presented in the form of tables and graphs of segmented regression (joinpoint regression), annual percentage changes with the respective 95%CI. The Excel 2016 (https://products.office.com/), Joinpoint Regression Software version 4.5.0.1 (https://surveillance.cancer.gov/joinpoint/) provided by the US National Cancer Institute (NCI) 5 , and the Stata Software version 11 (https://www.stata.com) were used. The results of the predictions are presented for each Brazilian Region.
The project was approved by the Ethics Research Committee with Human Beings -School of Medical Sciences, University of Campinas (FCM/Unicamp) under process number 2,903,348.

Results
A total of 69,661 TB cases were reported in 2017 in Brazil, with the lowest incidence in the historical series (55.2/100,000 inhabitants). Among the incidence rates in the Brazilian Regions, the Northern Region had the highest (69.2 per 100,000 inhabitants) and the Central Region had the lowest (32.1 per 100,000 inhabitants). We observed a decrease in incidence in all regions in the period, although high incidences of disease were recorded in the first years of the time series ( Figure 1).       Table 1).
The Central Region was characterized by decreasing APC throughout the study period in the Federal District and Mato Grosso do Sul. Mato Grosso had a decrease between 2013 and 2017 and Goiás registered an annual average variation of -2.5% per year (95%CI: -3.6;-1.3) ( Table 1).
Five out of nine Northeastern states had decreasing annual variations throughout the studied series, namely, Alagoas, Bahia, Ceará, Paraíba and Sergipe. For the other states, the AAPC was negative; they were Maranhão, Pernambuco, Piauí and Rio Grande do Norte (Table 1) Table 1).
As for the South, the State of Paraná was the only one with the same APC for the whole time series, while Rio Grande do Sul recorded a decrease in APC between 2011 and 2017, and Santa Catarina a decrease of -5.6% per year between 2013 and 2017. In Brazil, the AAPC was -1.8% per year (95%CI: -2.4; -1.1) ( Table 1).
The evaluation of the historical series showed that the trend is uniform throughout the period in 12 Brazilian states (Acre, Alagoas, Bahia, Ceará, Espírito Santo, Mato Grosso do Sul, Pará, Paraíba, Paraná, Rio de Janeiro, Sergipe and Tocantins) and the Federal District. The adequacy of the adjustment of the TB incidence coefficients in the Brazilian states according to the most adequate segmented regression was the one that considered two, three, four or five turning points in the period depending on the state, that is, four segments in the states of Rio Grande do Sul, Santa Catarina and Pernambuco, three in Mato Grosso, Rio Grande do Norte and Roraima, two segments in Amapá, Amazonas, Goiás, Maranhão, Minas Gerais, Piauí, Rondônia and São Paulo, and a straight segment was enough to represent the trends of the other states, all with a constant decreasing trend (Figure 2).
The states with two-line segments, Amapá, Amazonas, Maranhão, Rondônia and São Paulo, were characterized by a decreasing initial and increasing final observation in APC. The states of Minas Gerais and Piauí had increasing APC at the beginning and decreasing at the end, while Goiás had a decreasing APC at two points between 2001 and 2017. Brazil (three segments) presented increasing and decreasing variations in the incidence of TB (Figure 2).

Discussion
In this study, high rates of TB incidence were observed at the beginning of the series in all Brazilian states.
In 2009, there were changes in the national policies of active search, follow-up and treatment of TB in Brazil, which resulted in the reduction of new cases 6 .
The WHO intensified the goals to reduce the TB incidence rate by 90% (< 10/100,000 per inhabitant) by 2035, in relation to 2015. In 2014, WHO approved the Global Plan to End Tuberculosis. Its intermediate stages are expected to occur in 2020 (20%, < 85/100,000), 2025 (50%, < 55/100,000 inhabitants) and 2030 (80%, < 20/100,000 inhabitants) and it requires greater commitment on the part of governments, and focusing on providing better services to vulnerable populations 7 . Brazil has the lowest rates in the list of 30 countries with high TB burden 1 .
The SUS has the proper structure to reduce the disease, counting on universal coverage, access to diagnosis and treatment. However, the prioritization of investments in intersectoral articulation must be reinforced, as well as the participation of the civil society and public policies that combat the occurrence of the disease 8 .
The Brazilian National Tuberculosis Control Program recognizes the importance of the horizontalization of TB care, aiming at the integration of TB control, especially with primary health care    (PHC), which includes the Community Health Agents Program and the Family Health Program (FHP), expanding access to diagnosis and treatment 4 . Prevention is still the fundamental strategy for the elimination of TB in cases of first contact, reinfection or latent reactivation of the disease 8,9 .
The region with the highest incidence rate of TB was the North, above the national average, with high endemism and marked impact on morbidity and mortality affecting mainly the indigenous population. Intra-regional heterogeneity has been identified in the spatial distribution of the disease, influenced by socioeconomic determinants and greater concentration of cases in the larger cities of the Northern Region 10 .
A recent study that evaluated TB control actions in Brazil in PHC showed that 47.56% of the basic health units (UBS in portuguese) in the North Region did not collect material for laboratory examination as one of the protocols for identifying the disease. The absence of protocol items for TB control shows the still precariousness in all Brazilian Regions, and not only in the Northern Region 11 .
The concepts of TB transmissibility that are based on the biomedical model of educational and curative interventions do not fit the reality of Northern populations (mainly indigenous and riverbased) since educational and therapeutic actions should consider the unique cultural issues of these populations 12 . The sensitivity of the health system to capture new cases is directly associated with the coverage of primary health care and epidemiological surveillance, particularly in case investigation and early diagnosis, contributing to the reduction of incidence 13 .
Poor access to services and lower diagnostic and therapeutic options in remote areas make it difficult to control the disease and increase the impact on morbidity and mortality 14 . Specific work processes with qualified teams and efficient action tools are needed to ensure access to the diagnosis in vulnerable people. These actions also depend on health education to populations that have specific demands or greater vulnerability 15,16 .
The high rates may suggest flaws in health services, such as decentralization of care, insufficient supervised treatment, early diagnosis, effective protocols for dealing with the confirmed cases, and adequate training of professionals with strengthened teamwork 10 . However, there was a downward trend in incidence rates in the North Region in the study period, which may mean improved access and quality of services.
The state with the highest incidence of TB is Amazonas, followed by Rio de Janeiro, both with rates above 100/100,000 inhabitants. The cases in the former are concentrated in the capital and large cities. As this state has a great territorial extension, cities in the countryside have less access to reference services 3,17 .
The Northern Region presents specific needs in view of the large territory, and difficult access to diagnosis, treatment and clinical follow-up. In addition, the vast international border, especially the northern arc of the state, requires integrated strategies of state and national governments and of neighboring nations for integrated epidemiological surveillance actions to address this public health problem 12 .
The region with the lowest incidence is the Central, which showed an increase in the years 2012 and 2013, and then experienced a decline. These findings are like other studies in the country 18,19 .
TB in the Northeast Region is characterized by its heterogeneous spatial distribution. The highest rates are found in the coastal cities, as they are more populous and have poor neighborhoods 20 . Social inequalities between Brazilian Regions and states show the difficulty in accessing basic health services, causing delayed diagnosis of various diseases, including TB 21 .
Among the nine states in Northeast Region, Pernambuco recorded the highest incidence rates. The difficult access of patients to the capital Recife was 62.9%, the lowest percentage registered in a research that compared this factor in three Northeastern capitals (Salvador, Bahia State, with 84.6% and João Pessoa, Paraíba State, with 90.9%) 18 . Care for TB cases in these three capitals was predominantly provided in public health units (86% in Salvador, 98.8% in Recife, and 100% in João Pessoa), demonstrating the public financing and accountability of patient care under the coordination of the Tuberculosis Detection and Treatment Program at the state and national levels 21 .
The results obtained in this study indicate the reduction of confirmed cases of TB in the Southeastern Region, but not in a homogeneous way. São Paulo stands out as the only state with a sharply increasing trend since 2007 while the rates have decreased in Rio de Janeiro and Espírito Santo. The Southeastern Region is characterized by consolidated actions to combat and provide care for TB Cad. Saúde Pública 2020; 36(6):e00081319 patients through the primary health care network, emergency care, Tuberculosis Control Programs and Specialized Services. The need to strengthen the Health Care Network in the different care settings and in the organization of work in an integrated way for the effective diagnosis of TB and reduction withdrawal, especially among HIV co-infected patients 22 , should be emphasized. Thus, these actions in São Paulo have been insufficient to achieve the goals of the National Plan for Elimination of Tuberculosis and for reduction of the disease with a central public health strategy 23 .
Although the Southern Region presents incidence rates below the national average, during 2006 to 2011, these remained stable, and decreased after 2012. This regional trend was strongly influenced by the states of Rio Grande do Sul and Santa Catarina. A study conducted in 2016 identified that the South Region had the lowest percentage of clinical follow-up (46.4%) of TB cases in Brazil. This follow-up includes the operationalization of strategies such as directly observed treatment (DOT) by UBS, search for missing persons to follow-up of DOT, surveillance of intra-household contacts and lack of knowledge about priorities of care for TB patients 24 . Rio Grande do Sul, which is the state with the highest incidence rate, has implemented short-to medium-term projects to improve TB control. The measures include intensifying training in priority municipalities, increasing the number of supervision visits to patients with TB, consolidating partnerships with other areas of care (HIV/AIDS, viral hepatitis, PHC, LGBT population) with awareness of civil society, among others 25 .
The results of this study showed that, although five Brazilian states (Amapá, Amazonas, Maranhão, Rondônia and São Paulo) showed an increase in new cases 26 and consequently in annual percentage, the incidence had a decreasing trend in the time series studied in most states.
TB remains a major global health problem, affecting around 10 million people each year and one of the top 10 causes of death in all continents. In the last five years, it has been the leading cause of death by a single infectious agent, outweighing HIV. It is noteworthy that, with early diagnosis and correct treatment, most infected people can be cured 1 .
The overall burden of the disease remains high in the world, but significant progress shows the impact of specific TB control policies. The rate of case detection has increased from 46% to 64% and the prevalence in the world had dropped by 47% by 2015. Since the year 2000, the global incidence of TB has been declining, albeit slowly, along with new reported cases each year; however, rates remain high in countries of Southeast Asia and sub-Saharan Africa 27,28 .
In some regions of East Asia such as the metropolitan areas of Hong Kong, China, Japan, South Korea and Singapore, the incidence of TB is stagnating or increasing. This phenomenon can be explained by cases that arise due to reactivation of the disease in a previously infected human population 29 or by immigration from high incidence countries, an event that contributes to the transmission 28 .
Countries such as Brazil, Indonesia, South Africa, Thailand and Vietnam have good indicators of TB control, ensuring mechanisms of notification of detected cases, as well as linking the supply of first-line drugs (as in Brazil) to rapid notification through a specific information system 1 . There has been tremendous progress over the last ten years when it comes to the development of new diagnostic and therapeutic methods and tools that support WHO policies. However, there is yet no prediction of a rapid and accurate test to diagnose patients outside the health care setting, which could result in clinical advances and reduced transmission 30 .
Guidelines on disease, treatment, prevention and follow-up of patients are pointed as the strengths of PHC, but autonomy during treatment, related to DOT, is the main negative point of follow-up of TB 31 , besides the weakness in the involvement of professionals in control actions, verticalization of control actions within PHC, turnover of professionals and weaknesses in professional training 32 .
Segmented regression indicated significant turning points in the incidence rate trend of TB in the Brazilian states by means of the annual percentage change. The calculation of the crude incidence by region identified the behavior of the disease and the possible turning points in the time series identified by segmented regression and the refined estimates in each segment of the line.
The joinpoint regression is an efficient and flexible statistical method able to determine the turning points and the trend pattern of TB incidence rates per state in different periods. The behavior of the response variable (TB incidence) can be analyzed either in the different periods of the explanatory variable (time), or separately, linking the trends to cultural, socioeconomic and political specificities of each state.
Cad. Saúde Pública 2020; 36(6):e00081319 It should be noted that an effective TB control program, including primary prevention, early detection and treatment, when well structured, tends to increase the number of new cases as a result of access to diagnosis and notification. This contributes to treatment and to reducing disease mortality, with emphasis on places with high social vulnerability. The use of secondary data can be pointed out as a limitation of this study, because of the underreporting and poor quality of information in several regions of the country.
However, the possibility of feedback for the official TB databases may partially lessen the insufficiency of the data. Changes in the TB reporting system since the implementation of new TB care policies do not allow the sharing of variables among information systems, making it difficult to analyze more widely national data. One limitation of the segmented regression is that the turning points identified by the segmented regression do not always coincide with the real points and variations, what happens because the regressions are connected continuously over time. Despite these limitations, all cases of TB reported in Brazil in the selected period were analyzed.

Conclusion
The results of this study showed a decline in the incidence rates of TB in Brazil. However, regional and intra-regional differences were observed. This heterogeneity can be explored in the epidemiological and quality contexts and access to health services in each state. The relevance of the use of the joinpoint regression in temporal trend analyses of epidemiological indicators must be highlighted as it allows the monitoring of variations with greater statistical precision. Determining the trend pattern of TB incidence may assist in the planning and implementation of national TB control policies. The support of Public Health Policies focused on TB in the definition of preventive and control actions, especially in PHC, is broadened regarding the dynamics of transmission in risk subgroups present in the Brazilian regional contexts.

Contributors
All the authors took part in the conception of the study, data analysis and interpretation, writing and critically reviewing the intellectual contents of the manuscript. They have approved the final version and take on responsibility for all aspects of this paper, including the guarantee of its accuracy.