Global burden and trend of acute lymphoblastic leukemia from 1990 to 2017

Acute lymphoblastic leukemia (ALL) is a common malignant hematologic disease that is characterized by large numbers of dedifferentiated lymphoid cells. Statistical data of ALL's incidence and mortality are fundamental for policymakers to allocate resources optimally. In this study, we reported the incidence, death, and disability-adjusted life year (DALY) of ALL in the globe from 1990 to 2017. Our analysis showed that the incidence case of ALL increased by 30.81%, while the age-standardized incidence rate (ASIR) maintained stable. Subgroup analysis by social-demographic index (SDI) showed that ALL's ASIR was significantly decreased in high SDI countries, but were moderately increased in high-middle SDI countries. The change trends of age-standardized death rate and DALY rate were similar to ASIR trends. Subgroup analysis by age groups showed that children and the elderly were more likely to suffer ALL. Risk factor analysis demonstrated that smoking was the most significant contributor to ALL's death and DALY in the globe. Besides, the high body-mass index is playing an increasingly important role in ALL-caused mortality. Multiple methods to counteract potential risk factors should be adopted, such as controlling body-mass index in all regions and avoiding occupational exposure to carcinogens in low SDI countries.

AGING survival rate of children with ALL is over 90% [11]. However, the clinical outcomes of ALL vary in different age groups and countries. Older adult patients tend to have a poor prognosis, and basic treatment of ALL might not be available in low-or middle-income countries [2]. A comprehensive analysis of ALL's global burden and incidence trend helps assess the health cost and make corresponding policies.
Global burden disease study 2017 (GBD 2017) dataset provides the burden of 354 diseases in 195 countries and regions [12,13]. The incidence rate, death rate, and disability-adjusted life year (DALY) rate could adequately reflect the disease burden and potential public health cost. In this study, we collected statistic data of ALL's incidence rate, death rate, as well as DALY rate and analyzed their change trends in different regions and countries. Socio-demographic Index (SDI) is developed by GBD researchers, which is a comprehensive indicator of social development degree. For some tumor types, the incidence and death rates significantly change along with social development degree [14,15]. Therefore, we explored the relationship between the change trends of ALL' burden and SDI.

The incidence trends of ALL
In the globe, the incidence cases of ALL increased from 49.07*10 3 in 1990 to 64.19*10 3 in 2017 (Table 1). However, the ASIR (per 100,000 individuals) kept stable (from 0.89 in 1990 to 0.85 in 2017). Our analysis showed that males were more likely to have ALL (incidence cases of ALL in male/female = 1.54:1 in 1990 and 1.41 in 2017). Subgroup analysis by SDI showed that the ASIR of ALL gradually decreased in high SDI countries during the 28 years (EAPC = -1.89, 95% CI -1.94 ~ -1.83) ( Figure 1A-1C). Contrarily, the ASIR of ALL was relatively stable in middle SDI countries (EAPC = 0.82, 95% CI 0.73 ~ 0.90). Subgroup analysis by geographical zone demonstrated that Australasia had the fastest reduction in ASIR (per 100,000 individuals) (EAPC = -2.33, 95% CI -2.53 ~ -2.13). On the contrary, Andean Latin America had the most pronounced growth in ASIR (EAPC = 1.51, 95% CI 1.37 ~ 1.64).

AGING
We constructed a Pearson's correlation model to assess the relationship between social development degree and ALL's incidence or mortality trends. Generally, the EAPC values of ASIR (Pearson correlation coefficient, abbreviated to r = -0.42, P < 0.0001), ASDR (r = -0.41, P < 0.0001), as well as age-standardized DALY (r = -0.44, P < 0.0001) were negatively correlated to SDI ( Figure 3A-3C). It was notable that this decline was most significant when the SDI value was above 0.6. To further investigate the role of social development degree in ALL's incidence or mortality trends, we drew the scatter diagrams to present dynamic changes of SDI and ASRs (ASIR, ASDR, and age-standardized DALY) of 21 regions in the globe during past 28 years ( Figure  4A-4C). We found that for most regions with middle or low SDI (expect Andean Latin America and Central Latin America), the ASRs were relatively stable. However, for most regions with high SDI, the ASRs were dramatically decreased, accompanied by gradually elevated SDI.

Age distribution of ALL
Firstly, we presented the incidence rate of ALL in all age groups ( Figure 5A, 5D). We found that the incidence rate of ALL had two peaks: children (under five years old) and the elderly. Almost in all age groups, males had a higher incidence rate than females. Besides, the elderly had the highest death rate ( Figure 5B, 5E), while children had the highest DALY rate ( Figure 5C, 5F). Compared to 1990, the incidence rate and death rate of ALL in children were reduced in 2017.

ALL attributable risk factors
For the world and regions with different SDI values, smoking was the most significant contributor to ALLcaused death and DALY ( Figure 6A-6L). However, the role of tobacco in ALL-caused death and DALY was gradually declined (Supplementary Figures 1, 2). High body-mass index was the second most significant risk factor for ALL-caused death and DALY. Notably, the contributing ratio of the high body-mass index was increased in all regions. Occupational exposure to carcinogens, including benzene and formaldehyde, was the second most significant risk factor for ALL-caused death and DALY. We found that the contributing ratio of occupational exposure to carcinogens was markedly higher in the low SDI region than the high SDI region (Supplementary Figures 1, 2).

DISCUSSION
Our study reported the status and trend of ALL's incidence and mortality in the globe from 1990 to 2017.
Considering the change of population size and distribution structure, the incidence rate and death rate of ALL were relatively stable during past 28 years. Despite this, we still noticed that the disparities of the region, gender, and age existed in ALL's incidence and mortality. Besides, the investigation of ALL's AGING According to previous statistical data, the peak of ALL's incidence rate occurred at 3 ~ 5 years of age in the United States [16]. We found that ALL's incidence rate indeed had a single peak (under five years old) in 1990 in the world. However, due to the decreased incidence rate in children, ALL's incidence rate had two peaks in 2017: one peak in children and the other peak in the elderly. Besides, we observed that the death rate of children (under five years old) was significantly decreased from 0.96 per 100,000 individuals (1990) to 0.78 per 100,000 individuals (2017). Correspondingly, ALL's incidence rate in the elderly (above 70 years old) was slightly increased (from 1.54 to 1.77 per 100,000 individuals in 2017).
Benefiting from the development of modern pediatric regimens, the ten-year survival rate of children was elevated from 11.1% (1962 ~ 1966) to 91.1% (2000 ~ 2007) [17]. At the same time, the clinical outcome of the elderly was less encouraging: for ALL patients over 55 years of age, the tolerability of intensive pediatric-derived therapy was weak, and the compliance to planned chemotherapy was relatively low [18]. The data from the SEER database showed the overall 5-year survival rate of ALL (60 ~ 69 years old) was still below 20% in the United States [19].
The ASDR and age-standardized DALY rates of ALL were gradually declined in the high SDI region but kept stable in other areas. This decline was partially attributed to the improvement in the management of ALL during the past decades. In some countries, especially developed countries, the 5-year survival rate of childhood ALL was considerably high. Nevertheless, for some low-income or middle-income countries, basic interventions for ALL were not consistently available, and the 5-year survival rate of ALL was relatively low [20]. Besides, we noticed the ASIR of ALL in high SDI region was also gradually decreased. We supposed that the decline of mortality rate in high SDI region might relate to reduced predisposing factors of ALL such as pesticide exposure and ionizing radiation.
Moreover, the ASIR, ASDR, and age-standardized DALY rates in low or middle-low regions stayed at a low level. This situation is related to several factors. A truly low incidence rate, limited cancer registration, or restricted diagnosis level could contribute to a similar statistical result. The under-diagnosis and underregistration for cancer patients might be interferences when interpreting ALL's burden in some countries [21].

AGING
We found that smoking, high body-mass index, occupational exposure to benzene, and formaldehyde were risk factors contributing to ALL-caused death and DALY. Until 2017, tobacco had been the most critical risk factor of ALL's mortality. According to the data from UK Biobank, smoking was closely related to myeloid clonal hematopoiesis and ASXL1 mutation. It has been reported that prenatal and early-life tobacco exposure could lead to the generation of ALL-related genomic deletions [22]. Additionally, the second-hand tobacco exposure of children was related to the increased risk of ALL with RAS mutation [23]. Apart from smoking, we observed that high body-mass index was the second most significant risk factor contributing to ALL's mortality. A previously multicenter cohort study showed that overweight or obesity at diagnosis was a marker AGING predicting early mortality for ALL patients [24]. High body-mass index was closely related to increased risk of traumatic lumbar punctures and radiographic osteonecrosis during diagnosis and treatment, which might lead to the poor outcomes of ALL patients [25,26]. The high body-mass index affected ALL patients' survival by upregulating insulin-like growth factor-1 signaling, increasing the levels of leptin, circulating glucose, certain amino acids, free fatty acids, and promoting chronic inflammation [27]. Notably, the survivors of ALL are at high risk of obesity and obesity-related metabolic disorders such as type 2 diabetes [15,28].
Furthermore, in this study, occupational exposure to carcinogens such as benzene and formaldehyde, also contributed to ALL-caused death or DALY. A retrospective study showed that benzene exposure during childhood was associated with increased risk of ALL and acute myeloid leukemia [29]. Even parental or maternal occupational exposure to benzene could elevate the risk of ALL in their offspring [30][31][32]. Moreover, it was found that formaldehyde exposure could disrupt hematopoietic function and induce leukemia-related chromosome alterations [33]. Notably, the contribution ratio of occupational exposure to carcinogens was significantly higher in low and lowmiddle SDI regions than the high SDI region. There was much room for improvement in reducing exposure to carcinogens for low or low-middle SDI regions.

CONCLUSIONS
Generally, the cases of ALL were continuously increased in the past 28 years in the globe. Excluding the change in population size and structure, the global incidence rate was relatively stable. Notably, the ASRs of ALL in the high SDI region were gradually decreased from 1990 to 2017, which might be attributed to the development of ALL treatment strategy and the declined exposure to predisposing factors. For most middle/low SDI regions except Andean Latin America and Central Latin America, the ASRs of ALL kept stable. Smoking, high body-mass index, and occupational exposure to benzene and formaldehyde were the main risk factors contributing to ALL-caused mortality. Measures should be taken to reduce ALL-caused loss, including reducing exposure to tobacco and carcinogens, especially for low or middle SDI countries considering the relatively high contribution ratio of carcinogen exposure in these countries.

Data acquisition
Statistical data, including ALL's incidence rate, death rate, and DALY rate, were downloaded by the Global Health Data Exchange tool (http://ghdx.healthdata. org/gbd-results-tool). Moreover, we collected some disease-associated parameters, such as SDI and age or sex distribution data. Besides, data on ALL-caused death and DALY attributable risk factors such as occupational exposure to carcinogens, smoking, and high body-mass index were extracted.

Statistical analysis
The burden of ALL was estimated by the number of incidence cases, death cases, and DALYs from 1990 to 2017. To offset the changes in total population quantity and age distribution, the age-standardized incidence and death rate (ASIR and ASDR), as well as agestandardized DALY rate, were also used in this study. The change trends of incidence rate, death rate, DALY rate were calculated by estimated annual percentage change (EAPC) values. In the present study, EAPC was calculated based on three age-standardized rates (incidence, death, and DALY) and regression model. The following algorithm was adopted to get EAPC: y = α + βx, in which y referred to log10 (ASR) and x meant year. Then, EAPC was calculated as EAPC = 100* (10^β-1). The values of EAPC reflect the changing trend of corresponding ASRs: when EAPC and its 95% CI are above 0, it indicates an increasing trend; when EAPC and its 95% CI are below 0, it shows a degenerating trend. Besides, to assess the influence of social development on ALL's incidence and mortality, we plotted a scatter diagram and trend lines to present the correlation between EAPC and SDI. Lastly, GBD 2017 provides data on cancer death and DALY attributable risk factors, including environmental or occupational factors, behavior factors, and metabolic factors. The three types of factors which could be further subdivided to 84 risk factors. In this study, we found four risk factors contributing to ALL's mortality: occupational exposure to benzene, occupational exposure to formaldehyde, smoking, and high bodymass index.

Data visualization
All calculations were based on R software (version 3.6.0). We drew world maps to reflect the incidence status and change trend of ALL in the globe. All data visualization works were conducted by R software with packages 'ggplot2' and 'map'.

Supplementary Figures
Supplementary Figure 1

Supplementary Tables
Supplementary Table 1