The burden and challenges of tuberculosis in China: findings from the Global Burden of Disease Study 2015

To achieve the End Tuberculosis (TB) Strategy, it is important to understand the characteristics of TB in China, which may provide the government with important clues for controlling TB by 2030. Data from the Global Burden of Disease Study 2015 (GBD 2015) and Institute for Health Metrics and Evaluation (IHME) were reviewed and analysed. The age-standardized death rate decreased by 83.79% [95% uncertainty interval (UI) 73.06–87.10] from 1990 to 2015. The age-standardized prevalence of TB in males decreased steadily by 33.88% (95% UI 29.35–37.67) but nearly increased by 6.24% (95% UI -2.02–15.07) in females from 1990 to 2015. Disability-adjusted life years (DALYs) were higher in males than in females, and the highest TB burden was found in the elderly (70+ years of age). Over the period 1990–2015, the attributable age-standardized DALY rates for smoking decreased by 12.98% (95% UI 2.40–24.27), but increased for alcohol use and high fasting plasma glucose (HFPG). Greater attention should be paid to females especially in the under 5 years of age group, and more latent reasons explaining TB DALYs should be explored in future studies.

Through these efforts, China achieved the Millennium Development Goal of a 50% reduction in TB prevalence and mortality in 2015. However, the incidences of HIV with TB, chronic disease with TB and multi-drug resistant TB (MDR-TB) have all increased in recent years [10][11][12][13] , which may hamper TB control strategies. One study indicated that China could not achieve the global target of the End TB Strategy by 2035 if it were to maintain the current DOTS strategy by using an individual-based computational model of TB transmission, with a declining incidence and mortality of 42% and 41%, respectively 14 .
Therefore, this study aims to provide important clues for the government to achieve its End TB strategy targets, by comprehensively overviewing TB trends and changes in mortality, prevalence and DALYs according to GBD 2015 and the Institute for Health Metrics and Evaluation (IHME) website 4 .

Results
A total of 2.49 million TB deaths were reported from 1990 to 2015 in China, and 67.07% were males (1.67 million). The TB prevalence and DALYs were estimated every 5 years from 1990 to 2015, and a total of 10.38 million TB cases (6.29 million for males) and 20.68 million DALYs (13.50 million for males) were reported from 1990 to 2015 in China. In this study, the means and 95% uncertainty intervals (UIs) and median of the rate changes were determined for the period between 1990 and 2015.
TB Mortality. In 2015, there were an estimated 48,922 TB deaths in China, of which 35,936 (73.46%) were among men and 12,985 (26.54%) among women. The total number of deaths from TB significantly declined by 69.34% (95% UI: 51.38-75.31) from 1990 (159,592) to 2015 (48,922) for both sexes. The age-standardized death rates decreased rapidly by 83.79% (95% UI 73.06-87.10) from 1990 to 2015. The age-standardized death rates steadily decreased annually, from 20.06 (95% UI 18.47-40.85) to 3.40 (95% UI 2.86-5.37) per 100,000 persons from 1990 to 2015 for both sexes (Fig. 1). The age-standardized death rates were always higher in males than in females, but the difference between males and females decreased from 13.33 to 3.33 per 100,000 persons from 1990 to 2015. The age-specific death rates increased continuously with age, and the highest death rate was observed in the elderly (70+ years of age) from 1990 to 2015 for both sexes (Supplementary Table S1).  12). Similar to the death rates and prevalence, the TB burden was much higher for males than for females at all time points (Fig. 2). Furthermore, TB burdens in the elderly (70+ years of age) was much higher than those in the other groups from 1990 to 2015 (Table 2).

Prevalence of TB.
Risk factors for DALYs. In 2015, 33.82% (95% UI 28.14-39.83) of the all-age and 31.99% (95% UI 26.55-37.73) of the age-standardized DALY rates could be attributed to all risk factors of smoking, alcohol use, high fasting plasma glucose (HFPG) and their overlaps, whereas more than 66% could not be explicitly attributed Age-standardized death rates of TB decreased from 1990 to 2015 for both sexes in China. The agestandardized death rates decreased annually, and the male age-standardized death rates were always higher than were the female rates from 1990 to 2015.
The population attributable fraction (PAF) changes varied substantially in age-and sex-specific groups from 1990 to 2015 (Supplementary Table S2). The proportions of PAFs attributed to smoking decreased in all age groups, whereas a higher reduction was observed in females, with the highest reduction of 60.35% in the 50-69-year age group. Moreover, reductions in males increased as age increased. Unfortunately, the proportions of PAFs attributed to alcohol use and HFPG increased in all age groups; the highest increasing proportion in the 15-49-year age group due to alcohol use and HFPG surpassed 28% and 24%, respectively.

Discussion
In this study, we mainly found marked reductions of nearly 70% in both the all-age or age-standardized mortality and DALYs from 1990 to 2015 in China for both sexes, but TB mortality and DALYs have faced a slow decline in recent years, which must be considered. The prevalence changed slightly for the all-age rate but decreased significantly for the age-standardized rate. Notable gender differences were observed: the prevalence steadily decreased in males from 1990 to 2015 but fluctuatingly changed in females with nearly increased by 6.24% in age-standardized prevalence of TB. Additionally, almost 30% of the age-standardized DALY rates were attributable to smoking, alcohol use and HFPG, consistently from 1990 to 2015 in China.
The highest mortality and DALYs have been steadily shifting to the elderly (70+ years of age) in both this study and global surveys 15 . The drift of TB to older populations has several causes. First, longevity is increasing in China. Life expectancy increased from 68.55 years in 1990 to 76.34 years in 2015 6 , and with longer life spans, more people survive to acquire TB infection. Second, the birth rate is decreasing. Reports from the National Bureau of Statistics of China indicate that birth rates have fallen from 21.06‰ in 1990 to 12.07‰ in 2015. Additionally, the entire population is aging. Thus, the proportion of the population older than 65 years was 5.60% in 1990 but increased to 10.50% in 2015 6 . In addition, health problems due to chronic disease or risk factors such as high total cholesterol, high systolic blood pressure, and hypoimmunity would predispose the elderly to TB 11,16,17 .
From the results of GBD 2015 and IHME, the subnational TB burden was not available. According to the GBD 2013, we found TB to be one of the top ten causes of years of life lost (YLLs) in Xinjiang and Tibet ranking 9 th and 10 th , respectively, which ranked 4 th in 1990 18 . When making a comparison with top two TB burden countries, the age-standardized DALY rate per 100,000 persons in 2015 in China was 119.60 (95% UI 100.75-161. 19 4 . On the basis of these findings, it appears that China has made great progress in cultural status and health education compared with other developing countries, but DALYs has faced a slow decline in recent years, and further efforts are needed in the future. The third national TB prevalence survey, which was implemented in 2010 and screened individuals aged older than 15 years, showed a reduced prevalence of TB for both sexes 19 . The GBD 2015 results indicated the same trend of TB prevalence in patients older than 15 years from 1990 to 2010 in both males and females in China. A reduced incidence, a high TB cure rate and a shortened TB duration may have resulted in the reduction of TB prevalence. Notably, the prevalence of TB increased since 2010 in females and in the populations aged under 5 and 5-14 years for both sexes 4 . Children are vaccinated with Bacille Calmette-Guerin (BCG), which prevents the dissemination of TB in young children and reduces the probability of infection TB by 50% 20 . Unfortunately, due to non-specific symptoms and the absence of an ideal diagnostic method, the prevalence of TB in children is often underestimated 21 .
The National Health and Family Planning Commission of the People's Republic of China launched a nationwide scale-up policy on "school TB prevention and control work" in July 2010, which was reinforced by screening, reporting and prevention of new TB cases in kindergarten and all kinds of school students, including those in primary school, junior high school, senior high school and college 22 . In addition, a recent national youth tobacco survey in China indicated that 19.90% of students in junior high school used tobacco and that the rates of smoking in adolescent females under 15 years of age had increased. Furthermore, 72.90% of students had been exposed to second-hand smoking 23 . Active TB screening, early identification and a high rate of smoking may be the most important reasons for the increasing prevalence among children.
The prevalence of TB in females fluctuating changed and began to increase again in 2015 according to data from GBD 2015, which may be attributed to the following mechanisms. One reason was MDR-TB, which is a major threat to the control of TB 13 . In China, a national survey of drug-resistant TB indicated that the rate of MDR-TB among retreated and newly diagnosed TB cases was 25.60% and 5.70%, respectively. Additionally, a female predominance in MDR-TB risk among retreated or newly diagnosed TB patients was observed, with adjusted odds ratio of 2.20 (95% CI 1.40-3.50) and 1.50 (95% CI 1.00-2.10), respectively 24 . MDR-TB treatment is expensive, and patients must pay deductibles in China, which pose a heavy financial burden, particularly to those from poor households. MDR-TB may increase the percentage of patients who default from treatment, thus extending the overall treatment duration and increasing new TB transmission. Another reason may be a larger female population shift from rural to urban than previously, which can expose them to infection and lead to TB disease 6 . Apart from the above reasons and greater longevity, further national surveys to determine the latent reasons for the increasing prevalence in females are needed.
In addition to the limitations reported in GBD studies 7,25,26 , there are potential limitations in this study. First, it is difficult to capture the characteristics of TB sequela because symptoms such as cough, fever, short of breath and weight loss may be mild and thus ignored, which greatly affects YLD calculations 27 . Second, modelling the estimated risk factors of DALYs may not accurately capture how all risks interact. Apart from smoking, alcohol use and well-established HFPG risk factors, specific factors such as ambient air pollution, malnutrition and diabetes are known to increase both the risk of TB transmission and the development of active disease. Studies reported positive associations between particulate matter (PM 2.5, PM 10 ), carbon monoxide (CO), nitrogen dioxide (NO 2 ) and the development of TB in the general population 28,29 . The ambient air pollution PM 2.5 has been a considerable public concern in China in recent years and is ranked as the 4 th greatest risk factor for disease burden, which should be systematically qualified using the available data 30 .
In summary, although we made great progress on TB control, the current epidemic situation of TB still remains serious in females especially in the under 5 years of age group. To meet global targets, more policies such as active case finding and preventive therapy in the latently infected should be implemented to control prevalence. The risk factors of smoking, alcohol use and HFPG were insufficient to explain TB DALYs and more latent reasons should be explored in future studies.

Materials and Methods
Data sources. National disease surveillance points, mortality registration and reporting system, national maternal and child health surveillance system and sample survey on population changes were main sources for China 3 . The TB cases were identified based on International Classification of Diseases and Injuries, versions 10 (ICD-10), discharge diagnosis code (codes A10-A14, A15-A19.9, B90-B90.9, K67.3, K93.0, M49.0, and P37.0). Risk factors, which are associated with an increased risk of DALYs, are potentially modifiable causes of disease or injury 26 . There were three main risk factors included in this study: tobacco smoking, excluding second-hand smoke; alcohol use; and HFPG. Nationally representative survey data, alcohol sales, epidemiological studies and examination surveys for HFPG are the sources of data for estimating TB risk factors. Because of the interdependency of HIV/AIDS and TB, the metrics of TB in this study were for people living without HIV. The datasets generated during analysed during the current study are available in the IHME, [http://ghdx.healthdata.org/ gbd-results-tool].
Statistical analysis. In order to comprehensively overview the TB trends and changes in mortality, prevalence and DALYs, we used Spatiotemporal Gaussian process regression to synthesize the sources and simultaneously correct for biases, and a causes-of-death ensemble modelling (CODEm) strategy to produce a complete dataset of exposure distribution and mortality 5,31 . Meanwhile, the non-fatal TB prevalence was estimated by using the Disease Model-Meta Regression (DisMod-MR 2.1) 25 . Thus, dynamical models, which can do both mathematical analysis and data fitting to reveal the transmission rules of infectious 32,33 , are not necessary in this paper. DALYs which represent how many years of healthy life are lost due to death and non-fatal illness or impairment are relatively simple and widely used to estimate the burden of disease. DALYs are equal to the sum of YLLs and YLDs 7 . YLLs assess the distance between observed mortality and a life expectancy and are computed by multiplying the number of TB deaths at each age χ by a standard life expectancy at age χ, whereas YLDs assess the health loss and are computed as the prevalence of TB sequelae multiplied by the disability weight. The sequelae of TB included a persistent cough and fever, shortness of breath, feelings weak, and substantial weight loss 12 . The standard life expectancy from the normative life table at the age of death was 86.59 years at the age of 0 year and 23.79 years at the age of 65 years, which was computed based on the lowest recorded death rates across countries 3 . The updated world population age standard, which was developed based on the GBD 2013 data sourced from the  The comparative risk assessment method is used to calculate the proportion of the TB burden that is attributable to risk factors 26 . Due to incomplete or missing data for risk factors for both specific populations and years, models such as Bayesian hierarchical regression, mixed effect regression and CODEm were used to generate a complete set of the current exposure distributions. Two formulas are presented to assess the PAF. The first is for polytomous tobacco smoking, and the second is for continuous alcohol use and HFPG 26  Where PAF asy is the population attributable fraction for tobacco smoking for age group α, sex s and year y. Where RR x ( ) asy is the relative risk (RR) at exposure level χ with the lowest level of observed exposure m to the highest n. Over the period 1990-2015, the attributable age-standardized DALY rates decreased for smoking, but increased for alcohol use and HFPG risk factors for both sexes.