Amyotrophic lateral sclerosis mortality rates in Latin America and the Caribbean: a meta-analysis

Abstract Background: Recent studies have described a low occurrence of Amyotrophic Lateral Sclerosis (ALS) in Latin America. Significant differences in ALS risk have been reported among ethnic populations in the region. We conducted a meta-analysis using population-based data to describe ALS mortality rates in Latin America. We explored sources of heterogeneity among key covariates. Methods: National mortality registries from Latin American countries were searched to identify ALS deaths according to the International Classification of Diseases (ICD-9: code 335.2 and ICD-10: code G12.2). Crude and standardized mortality rates were calculated. A random-effect meta-analysis was conducted to estimate pooled mortality rates. Subgroup analysis was performed as a means of investigating heterogeneity. Results: Overall, 28,548 ALS deaths and 819 million person-years of follow-up (PYFU) from ten Latin American countries were considered. Standardized mortality varied among countries. The highest mortality rates were observed in Uruguay and Costa Rica at 1.3 and 1.2 per 100,000 PYFU, respectively. The pooled crude mortality rate was 0.38 (95%CI: 0.28–0.53) and the pooled standardized mortality was 0.62 (95%CI: 0.49–0.77) per 100,000 PYFU. Heterogeneity was high (I2: 99.9%, Cochran’s Q p < 0.001). Subgroup analysis showed a higher mortality rate among countries with a higher proportion of Caucasian populations and higher income levels. Conclusion: There is a lower ALS occurrence in Latin America compared to Europe and North America. This meta-analysis supports the hypothesis of a higher ALS risk among the Caucasian population. Further studies are needed to investigate the role of ancestral origins in ALS, taking socioeconomic status into consideration.


Introduction
Amyotrophic Lateral Sclerosis (ALS) is a rare neurodegenerative disease with a poor prognosis due to respiratory insufficiency, with a median survival time from diagnosis of around 15-20 months (1). Recent studies have suggested heterogeneity of ALS epidemiology among geographical areas and populations (2)(3)(4)(5). Lower incidence and mortality rates of ALS have been described in Latin America compared to Europe and North America (6).
A lower ALS occurrence has been observed in Hispanic populations compared to Non-Hispanic populations in studies performed in the United States (5,7). In Latin America, studies from Cuba and Ecuador have shown significant differences among ethnic groups (3,4). Furthermore, the results described in Latin American countries were consistent with those observed in Hispanic populations in the United States.
This could imply that ancestral origin could play an important role in ALS occurrence, as higher risks of developing ALS have been observed between ethnic groups (3)(4)(5). Variations of ALS occurrence could also be explained, by methodological heterogeneity and differences of socioeconomic status.
There is a major lack of information in some regions of the world, including Latin America. Population-based registries have played a key role in understanding descriptive ALS epidemiology (8). In countries where ALS registries are not available, mortality data has been a valuable source of information as it can be expected to identify all ALS cases due to the fatal outcome of the disease (9). Hence, mortality data could be considered as a proxy of ALS incidence if high-quality methodology is followed (10).
Latin America is an interesting region for ALS studies because of the diversity among ethnic groups. To improve our understanding of the role of ancestral origins in ALS occurrence, there is a need for high-quality studies using standardized methodology. In this context, we conducted a meta-analysis using general population data to describe ALS mortality rates in Latin America. Sources of heterogeneity were also explored.

Methods
We followed the guidelines for Meta-analyses and Systematic Reviews of Observational Studies (MOOSE) (11). We followed specific epidemiological criteria for ALS mortality studies (10). The Moose checklist is shown in Supplementary etable 1.

Source of information
We performed an online search of registries of annual mortality causes from all Latin American countries. Available information for each country was collected from 1990 to 2019. Mortality registries from the National Institutes of Statistics of eight countries were available online: Argentina (12), Chile (13), Colombia (14), Costa Rica (15), Ecuador (16), Guatemala (17), Mexico (18), and Uruguay (19).
We identified mortality studies using data from national registries in Brazil and Cuba, through a previous systematic review of ALS in Latin America (6). The authors were contacted to obtain the mortality data of their countries (Brazil (20) and Cuba (4)).

Case ascertainment
All death records in the mortality registries were searched. National registries collect causes of death from standard death certificates. Health professionals fill in these certificates. Registries in Latin American countries follow guidelines for mortality coding according to the recommendations provided by the World Health Organization (21).
The International Classification of Diseases (ICD) was used to identify ALS deaths: ICD-9 code 335.5 from 1990 to 1996 and ICD-10 code G12.2 from 1997 to 2019. The underlying cause of death was used to identify ALS cases. Cases younger than 15 years of age were excluded to reduce the risk of false positives.

Data collection
Demographic information was collected from each mortality registry including sex and age group at the time of death. The population at risk in each country was obtained from the Annual Demographic Yearbook published by the United Nations Statistics Division (22). The mid-year population per year was considered to calculate the mortality rates.
Subcontinent classification was performed according to the United Nations Statistics Division (23). Country classification by income was obtained from the World Bank's classification by income level. As we include large periods among countries, we took into consideration the middle year of the period for each country as the reference for the income classification (24). The proportion of ethnic groups was obtained from a nonprofit private corporation "Latino Bar ometro," which carries out an annual public opinion survey using a standardized methodology among Latin American countries (25). Self-reported ethnicity is categorized into Caucasian, Admixed, Black, Asians, Indigenous, Mulatto, other race, or no answer. Information for ethnic groups was available from 2007 to 2018. To obtain the proportion of each ethnic group from each country, we collected the crude data for each year and then we calculated the proportion for the overall period. The Latino Bar ometro database and methodology are published elsewhere (25).

Statistical analyses
Qualitative variables were described as frequencies and percentages. The frequency counts of ALS deaths were categorized in 5-year age increments. Crude mortality rates per 100,000 person-years of follow-up (PYFU) were calculated along with 95% confidence intervals (95%CI) based on the Poisson distribution. Direct age and sex standardization was performed using the US 2010 population.
A meta-analysis was performed using a random-effect model and Forest plots were generated. Pooled mortality rates were calculated. The Cochran's Q-test (p < 0.1) and I 2 statistics were used to assess statistical heterogeneity.Heterogeneity was defined as high if greater than 75% (I 2 statistics). We performed stratified analyses as a means of investigating heterogeneity with respect to a number of key covariates: (i) Proportion of Caucasian population, (ii) Income level and (iii) Geographic localization (Subcontinent). We categorized geographic localization into two subcontinent groups: (a) Central America and The Caribbean, and (b) South America. The proportion of Caucasian population was categorized (<25%, 25-50%, 50-75%, >75%). Mortality data from nine countries were used for the proportion of Caucasian population subgroup analysis, as this information was not available for Cuba in Latino Bar ometro.
Analyses were conducted using R statistical software version 3.6.1 (The R project for statistical computing) and Stata version 11.1 (Stata Corporation, College Station, TX).

Results
Overall, 28,548 ALS deaths were identified considering a population of 819 million PYFU covering 10 Latin American countries. Fifty-five percent of ALS cases were males (15,717 cases) and 45% were female (12,831 cases). The total male/female sex ratio was 1.22. The characteristics of ALS mortality data are described in Table 1.

Crude and standardized ALS mortality among Latin American countries
Crude ALS mortality ranged from 0.07 to 1.24 per 100,000 PYFU. Mortality remained variable after standardization (Figure 1

Age-related profile and sex ratio of ALS mortality in Latin America
The peak age of crude mortality was between 75 and 79 years followed by a decrease after 80 years. A similar pattern was observed for standardized mortality ( Figure 2). The crude mortality sex ratio was 1.30, while the standardized mortality sex ratio was 1.27. The crude and standardized mortalities by sex in each country are shown in Table 2.

Discussion
The present meta-analysis reports ALS crude and standardized mortality rates using population-based data from Latin America. Our findings support the evidence of a lower occurrence of ALS in this geographic region compared to Europe and North America. A higher mortality rate was observed among countries describing a higher proportion of Caucasian population and higher income level.

Age pattern and sex ratio of ALS mortality in Latin America
ALS mortality in Latin America exhibits a specific age-related pattern, characterized by low mortality rates in younger ages (before 40 years of age), and a sudden decrease in older ages (70 years of age), similar to reports from Europe, North America and East Asia (26). Recent studies have suggested a younger age at onset in Latin America compared to Europe and New Zealand (6). Age peaks for ALS incidence in European populations ranged between 71.6 and 77.4 years and around 75 years in East Asia (26). In Latin America, the highest peak of ALS standardized mortality was in the 60-69 year age group. This differences could be related to life expectancy, under ascertainment of cases and difficulties in access to healthcare (27).
A male predominance was observed in Latin America, similar to reports in Europe, North America and New Zealand (28). A recent dose-response meta-analysis of population-based studies reported a pooled male to female ratio of 1.28, a crude incidence sex ratio of 1.33 and a sex ratio of standardized incidence of 1.35, which is consistent with our findings (29).

ALS variability in Latin American countries
Recent evidence supports a low ALS occurrence in Latin America. A recent systematic review, found that methodological issues, make it difficult to draw firm conclusions of ALS epidemiology in this region (6). In this meta-analysis, we confirm a low occurrence of ALS in Latin America. This approach provides reliable epidemiological data to gain important insights into ALS heterogeneity among geographical areas.
ALS mortality rates were heterogeneous among Latin American countries. Geographic regions with a similar ancestral origin population (Europe, North America and New Zealand) have shown homogenous incidence rates, while significant differences were found between Europe and Asia (2). A recent study in South Africa described a significant difference in ALS occurrence among populations, with a lower incidence rate among the population of African ancestry compared to those of European ancestry (30). Epidemiological studies performed in the United States showed that incidence rates were lower in Hispanic populations compared to Non-Hispanic White populations (5,7). Differences among ethnic population groups  (20). Mortality rates observed in previous studies in Latin America were consistent with those observed in Hispanic populations in the United States (5,7). We performed a linear regression to explore the correlation between ALS mortality rates and the proportion of Caucasian population. The analysis showed that there was a strong positive linear correlation (Pearson's correlation ¼ 0.83, p ¼ 0.005). Crude mortality increases by 0.012 for each 10% increase in the Caucasian population. R 2 value was 0.70 (Supplementary efigure 2).
Researchers have proposed that ALS risk variability could be associated with ancestral origin, as studies have shown a higher ALS occurrence among European populations which could share common "at risk" alleles, increasing ALS susceptibility (4). In contrast, admixed populations could show a lower risk for ALS because of the combinations of this alleles (4). Latin American populations reflect a continuous admixture of Native American, European and African ancestries shaped by the interaction of migrants and Native American indigenous peoples, due to historical events such as colonization and the slave trade (31,32). Ancestral origin is heterogeneous in Latin American and the Caribbean countries. A genomewide pattern investigation of population structures showed that individuals from Mexico and Ecuador have the highest levels of Native American ancestry (31). Colombia and Brazil have shown widespread genetic patterns, which vary between geographical regions within the country. In Colombia, for instance, the highest levels of African ancestry have been shown to be in the coastal regions, with a higher level of European ancestry in central areas (33). In Brazil, a higher level of European ancestry has been evidence in the south, conversely a higher level of African ancestry has been described in the East (33). A higher proportion of European ancestry has been exhibited in Chile, Argentina and Uruguay (32,34,35). Ethnic identification has been used as a proxy of ancestral origin. However, this is prone to several limitations including the lack of a standardized definition. For instance, ethnicity is self-reported based on cultural and traditional aspects in Ecuador (36), while ethnic classification is performed using skin color in Cuba (37). To address this issue, we used ethnic data collected using a homogenous methodological framework among countries (Latino Barometro). Further studies using objective assessments of ancestral origin are needed in ALS research. Socioeconomic factors have been considered an important determinant to take into consideration for ALS variations. The Global Burden of Disease Study of motor neuron diseases reported a higher age-standardized prevalence and incidence in countries with a higher socio-demographic index (SDI) compared to low and middle SDIs. Agestandardized prevalence, however, was lower in certain geographic regions with high SDIs (38). A population-based study suggests that ALS incidence could be associated with socioeconomic status (SES) and race. After using adjustment models for age, sex and race they showed that the participants in the highest-income quartile had a higher relative risk for ALS compared to the lowest-income quartile. The relative risk of having ALS was significantly lower among Blacks and Asians than among Whites (39). On the contrary, another study in the United States, after using adjusted models to control socioeconomic   factors, type of health insurance and birthplace, a lower risk for ALS was observed in Non-Hispanics Blacks and Hispanics versus Non-Hispanic White populations (40). A lower mortality in lower-middle income countries could be explained by difficulties in health care access, social inequalities, inadequate case ascertainment and scarce medical resources (27). According to the World Health Organization, the median number of neurologists per 100,000 population for lower-middle-income countries was 0.13 and 1.09 for upper-middle-incomes (41). The Pan American Health Organization showed that the ratio of neurologists per 100,000 habitants was 0.92 in Central America, 0.48 in the Caribbean and 1.59 in South America (42).
Regardless of these factors, lifestyle and environmental factors cannot be dismissed as possible sources of epidemiological heterogeneity for ALS. Some authors have propose that ALS could be the   result of a complex interaction between environmental risks and genetic predisposition (43).

Mortality data
Population-based registries are the gold standard methodology for epidemiological studies in ALS. These registries were mostly developed in European countries. The key point for ALS registries is the capacity to identify a reference population in a welldefine geographic area, ensuring case identification among all possible sources, which is possible because of the structured national health systems and specific funding (44). Latin American countries' health systems face different challenges including a double burden of diseases that makes rare diseases have less or inexistent government funding. In countries where no ALS registries exist and incidence studies are challenging, the use of death certificates can be a key tool in estimating ALS occurrence since they allow a population-based approach and provide standard data collection instruments among countries.
The use of death certificates as a source for descriptive epidemiology has raised different   concerns among researchers as mortality studies present limitations that could lead to an underestimation of cases. Various studies have shown, however, that death certificates have good sensitivity for ALS case identification (45). Mortality studies conducted in Hong Kong and England exhibited mortality rates consistent with incidence rates for similar time intervals (46,47). Mortality data can be used as a proxy of ALS incidence; because of the fatal outcome of the disease; we could assume all ALS cases will be identified (10).

Limitations and strengths
Our study had certain intrinsic strengths. First, there was a population-based approach and a homogenous methodology following high standard criteria for reporting mortality studies. Second, we considered the sex and age structure of the population per year in each country to estimate mortality rates. Third, a long period was investigated to include a sufficient number of events. Lastly, we explored ALS heterogeneity considering key covariates through subgroup analysis. Some limitations need to be acknowledged. Self-reported ethnicity was considered as a proxy of ancestral origin which could lead to bias, as this is not an objective assessment of the ancestral origin of individuals and is normally based on traditional cultural aspects and physical appearance (48). Nevertheless, the results of previous studies of the genomic patterns in Latin America are consistent with the reports of the self-reported Caucasian proportion from the "Latino Bar ometro". Another limitation was that some registries only provided the underlying basic cause of death and not the secondary causes, to assure a homogenous methodology we considered only the basic cause of death, which could lead to an underestimation of cases.

Conclusion
This meta-analysis using population-based data confirmed a lower ALS occurrence in Latin America, supporting the hypothesis of a higher risk in developing ALS for populations of Caucasian origin. Furthermore, subgroup analysis showed ALS heterogeneity, with higher mortality among countries with upper-middle income levels. Further studies are needed to investigate the role of ancestral origins in ALS, taking into consideration socioeconomic factors, gene association and environmental interactions.