Cohort Profile: The Cohorts Consortium of Latin America and the Caribbean (CC-LAC)

Why was the cohort set up? 
Latin America and the Caribbean (LAC) are characterized by much diversity in terms of socio-economic status, ecology, environment, access to health care,1,2 as well as the frequency of risk factors for and prevalence or incidence of non-communicable diseases;3–7 importantly, these differences are observed both between and within countries in LAC.8,9 LAC countries share a large burden of non-communicable (e.g. diabetes and hypertension) and cardiovascular (e.g. ischaemic heart disease) diseases, with these conditions standing as the leading causes of morbidity, disability and mortality in most of LAC.10–12 These epidemiological estimates—e.g. morbidity—cannot inform about risk factors or risk prediction, which are relevant to identify prevention avenues. Cohort studies, on the other hand, could provide this evidence. Pooled analysis, using data from multiple cohort studies, have additional strengths such as increased statistical power and decreased statistical uncertainty.13 LAC cohort studies have been under-represented,14 or not included at all,15–17 in international efforts aimed at pooling data from multiple cohort studies. We therefore set out to pool data from LAC cohorts to address research questions that individual cohort studies would not be able to answer. 
 
Drawing from previous successful regional enterprises (e.g. Asia Pacific Cohort Studies Collaboration),18,19 we established the Cohorts Consortium of Latin America and the Caribbean (CC-LAC). The main aim of the CC-LAC is to start a collaborative cohort data pooling in LAC to examine the association between cardio-metabolic risk factors (e.g. blood pressure, glucose and lipids) and non-fatal and fatal cardiovascular outcomes (e.g. stroke or myocardial infarction). In so doing, we aim to provide regional risk estimates to inform disease burden metrics, as well as other ambitious projects including a cardiovascular risk score to strengthen cardiovascular prevention in LAC. 
 
Initial funding has been provided by a fellowship from the Wellcome Trust Centre for Global Health Research at Imperial College London (Strategic Award, Wellcome Trust–Imperial College Centre for Global Health Research, 100693/Z/12/Z). Additional funding is being provided by an International Training Fellowship from the Wellcome Trust (214185/Z/18/Z). At the time of writing, the daily operations and pooled database are hosted at Imperial College London, though a mid-term goal is to transfer this expertise and operations to LAC. The collaboration relies fundamentally on a strong regional network of health researchers and practitioners

Latin America and the Caribbean (LAC) are characterized by much diversity in terms of socio-economic status, ecology, environment, access to health care, 1,2 as well as the frequency of risk factors for and prevalence or incidence of non-communicable diseases; [3][4][5][6][7] importantly, these differences are observed both between and within countries in LAC. 8,9 LAC countries share a large burden of noncommunicable (e.g. diabetes and hypertension) and cardiovascular (e.g. ischaemic heart disease) diseases, with these conditions standing as the leading causes of morbidity, disability and mortality in most of LAC. [10][11][12] These epidemiological estimates-e.g. morbidity-cannot inform about risk factors or risk prediction, which are relevant to identify prevention avenues. Cohort studies, on the other hand, could provide this evidence. Pooled analysis, using data from multiple cohort studies, have additional strengths such as increased statistical power and decreased statistical uncertainty. 13 LAC cohort studies have been under-represented, 14 or not included at all, [15][16][17] in international efforts aimed at pooling data from multiple cohort studies. We therefore set out to pool data from LAC cohorts to address research questions that individual cohort studies would not be able to answer.
Drawing from previous successful regional enterprises (e.g. Asia Pacific Cohort Studies Collaboration), 18,19 we established the Cohorts Consortium of Latin America and the Caribbean (CC-LAC). The main aim of the CC-LAC is to start a collaborative cohort data pooling in LAC to examine the association between cardio-metabolic risk factors (e.g. blood pressure, glucose and lipids) and nonfatal and fatal cardiovascular outcomes (e.g. stroke or myocardial infarction). In so doing, we aim to provide regional risk estimates to inform disease burden metrics, as well as other ambitious projects including a cardiovascular risk score to strengthen cardiovascular prevention in LAC.
Initial funding has been provided by a fellowship from the Wellcome Trust Centre for Global Health Research at Imperial College London (Strategic Award, Wellcome Trust-Imperial College Centre for Global Health Research, 100693/Z/12/Z). Additional funding is being provided by an International Training Fellowship from the Wellcome Trust (214185/Z/18/Z). At the time of writing, the daily operations and pooled database are hosted at Imperial College London, though a mid-term goal is to transfer this expertise and operations to LAC. The collaboration relies fundamentally on a strong regional network of health researchers and practitioners.

Who is in the cohort?
We have harmonized and pooled approximately population-based cohort data on cardio-metabolic risk factors and outcomes, i.e. participants were not recruited based on disease (e.g. cohort of stroke survivors) or risk factor (e.g. cohort of smokers) history. Data have been collated by the CC-LAC, a LAC network of health researchers and practitioners. The database was collated using multiple data identification sources. First, we accessed publicly available cohort data through each V C The Author(s) 2020. Published by Oxford University Press on behalf of the International Epidemiological Association. study's website or data repository. Second, we conducted a systematic search in Medline, Embase and SciELO, a LAC-based search engine to identify cohort studies with peer-reviewed publications in regional journals. The search query included country terms (countries in LAC), cohort studies (e.g. cohort stud*), and cardiovascular outcomes (e.g. stroke); the search query is available in Supplementary data p. 02, available as Supplementary data at IJE online. We invited all eligible cohorts to join the CC-LAC requesting access to anonymized individuallevel data. Third, enquiries among LAC researchers also helped identify or approach principal investigators of the identified cohorts. LAC members of the Non-Communicable Disease Risk Factor Collaboration (NCD-RisC) 20 also helped identify additional data sources. Sources that led to the identification of the collaborating cohort studies are shown in Supplementary data pp. 03-06, available as Supplementary data at IJE online. The project procedures were approved by the Institutional Review Board at Universidad Peruana Cayetano Heredia (UPCH), Lima, Peru.
Data sources were included in the CC-LAC database if they had at least a baseline and one follow-up round, included LAC populations living in LAC, and participants were not recruited exclusively based on the presence of a risk factor or disease. No restrictions were set regarding a minimum sample size at baseline or follow-up duration, sex or age profiles. Anonymized individual record data were received, harmonized and pooled by the CC-LAC. During the harmonization process, we excluded participants who by the time of the baseline assessment had had a cardiovascular event (e.g. stroke or myocardial infarction). Implausible values in selected risk factors were excluded as follows: [5][6][7]21 body mass index (BMI <10 kg/m 2 or >80 kg/m 2 ), systolic blood pressure (<70 mmHg or >270 mmHg), diastolic blood pressure (<30 mmHg or >150 mmHg), total cholesterol (<1.75 mmol/L or >20.00 mmol/L), high-density lipoprotein (HDL)-cholesterol (<0.40 mmol/L or >5.00 mmol/ L), and fasting glucose (<2.50 mmol/L or >30.00 mmol/L) (Supplementary data are available at IJE online p. 07). Notably, BMI was computed based on measured weight and height in all but one Mexican cohort, 22 yet the self-reported weight and height closely correlated with measured estimates. 23 At the time of writing, i.e. first datalock on 29 March 2020, 32 cohorts with data on cardiovascular outcomes have been pooled. Seven additional cohorts have joined the CC-LAC but are currently working to provide cardiovascular outcomes, thus these 7 cohorts have not been pooled yet. To date, the CC-LAC is therefore a consortium of 39 cohorts in 13 LAC countries ( Figure 1). Among those cohorts we currently have pooled, there are two multi-country studies: Centro de Excelencia en Salud Cardiovascular para América del Sur (CESCAS) Cohort (Argentina, Chile and Uruguay) 24 and the 10/66 Study (Cuba, Dominican Republic, Peru, Venezuela, Mexico and Puerto Rico), 25,26 whereas the other cohort studies were based in only one country. The largest sample size is with The Mexican Teachers' Cohort (  32 and MONICA-VITÓ RIA study (Brazil). 33 The pooled dataset including all cohorts is not designed to be fully representative of the populations of regions and countries in which they have been conducted. Nevertheless, the results provided by this consortium will be informative for most of the LAC region, where large, longitudinal and multi-country studies remain insufficient in the field of cardiovascular diseases risk prediction. This consortium aims to advance the regional scientific evidence by overcoming the limitations of individual cohorts, e.g. studies with small sample size or limited number of events, while building upon a strong collaborative network of investigators.
How often have they been followed up?
The CC-LAC has pooled the baseline assessment and the latest follow-up available for each cohort; for the purpose of this consortium all subjects have been followed once, i.e. at baseline and one follow-up. All pooled individuals have the outcomes of interest, either non-fatal or fatal cardiovascular events (or censored). At the time of the first datalock, the 32 pooled cohorts had a mean follow-up time to the first cardiovascular non-fatal/fatal event of 8.50 (median ¼ 8.80) years, ranging from <1.0 to 27.7 years. What has been measured? Table 1 shows the number of observations per pooled cardio-metabolic risk factor. In addition to these variables, some cohorts have collected additional anthropometric measurements (e.g. waist/hip circumference or skinfold thickness) and laboratory data (e.g. Glycated hemoglobin (HbA1c), triglycerides, low-density lipoprotein (LDL)-cholesterol). Table 2 depicts the cardio-metabolic risk factors available across the pooled cohorts, and Supplementary data pp. 17-18, available as Supplementary data at IJE online, show further information on these variables at followup rounds. In addition to clinical cardio-metabolic risk factors, we have pooled two relevant socio-demographic indicators: place of residence (urban or rural) and schooling (years of education).
Two outcomes of interest have been harmonized and pooled, these are non-fatal and fatal cardiovascular events: haemorrhagic and ischaemic stroke, myocardial infarction (including revascularization), and mortality due to these conditions, as well as sudden death. Where relevant, these outcomes have been adjudicated or extracted from reliable sources, such as clinical records or death registries. Details on the ascertainment methods of these outcomes are provided in Supplementary data pp. 8-16, available as Supplementary data at IJE online.
At the time of the first datalock, data from 174 180 individuals have been pooled: 171 937 (98.7%) have complete follow-up or were censored (e.g. lost-to follow-up or death by other causes), 578 (0.3%) have had a first nonfatal cardiovascular event and 1665 (1.0%) have experienced a fatal cardiovascular event (Table 1).

What has it found? Key findings and publications
This is the first of a series of anticipated outputs and therefore constitutes a key initial publication to provide a general overview of this regional long-needed endeavour. This cohort profile aims to inform the international research and clinical community about our ongoing efforts and inform them of our anticipated forthcoming outcomes. Our ongoing work includes: (i) age-specific risk estimates of cardio-metabolic risk factors on cardiovascular nonfatal and fatal outcomes; (ii) comparative risk assessment of current versus ideal levels of cardio-metabolic risk factors based on regional age-specific risk estimates and survey data; and (iii) a novel risk score for cardiovascular events based on LAC hazard estimates, and countryspecific re-calibration with LAC population-based estimates of risk factors. Although other relevant efforts have provided cardiovascular risk charts for LAC, 34,35 they all relied on hazard estimates from cohorts outside LAC which may not accurately apply to LAC populations.
Besides scientific publications and other relevant outputs, CC-LAC has found that several cohort studies have been conducted in LAC and provide strong data to advance the knowledge of cardiovascular health in LAC. Probably, these cohort studies were not included in previous international cohort data pooling efforts because of language barriers or because LAC cohorts are rather young, i.e. started mostly in the 2000s. Moreover, CC-LAC has demonstrated that it is feasible to conduct a large-scale cohort data pooling effort across LAC. We encourage other research groups working in different fields across health sciences (e.g. infectious diseases, climate change and maternal/child health) to also embark on similar efforts. .

What are the main strengths and weaknesses?
The main strength of the CC-LAC is its regional scope and the large sample size. As discussed before (see 'Why was the cohort set up?' section), LAC has not participated in other global cohort data-pooling efforts to study cardiovascular risk factors and outcomes, leaving room to study cardio-metabolic risk estimates in this region. This work is therefore the first-ever cohort data-pooling study in LAC, which will ultimately produce solid evidence to guide clinical and public health practice in LAC. Previous data-pooling efforts in LAC have only focused on cross-sectional or survey data, and have shown relevant differences in relation to high-income countries. 36,37 CC-LAC will complement and move these efforts to the next level, achieving larger impact in LAC healthcare practice, research and policy. Moreover, the fact that most of the pooled cohorts have begun in recent years is also a strength. They will provide evidence based on contemporary estimates, reflecting the present epidemiological scenario. The CC-LAC has managed to harmonize key cardio-metabolic risk factors that were measured following consistent protocols, e.g. repeated blood pressure measurements or standard laboratory methods for biomarkers. Lastly, with the large sample size, other relevant outcomes, such as diabetes risk, may also be investigated.
Data pooling comes with methodological challenges. Understanding the heterogeneity among cohorts in terms of population and sampling methodology, as well as variables collection and ascertainment, is relevant to find the best way to pool and harmonize available cohort data. Additional sources of heterogeneity, such as different levels of non-response rates, would be a limitation of any cohort data-pooling project. Future results should be interpreted considering all this and other limitations specific to each research question.
Like other cohort pooling studies, among the weaknesses we should point out the heterogeneity in the definition of cardiovascular events. Although pooled cohorts have followed adjudication processes, or these outcomes were based on death certificates, data were sometimes registered as 10th revision of the International Statistical Classification of Diseases and Related Health Problems (ICD-10) codes or in broader categories (e.g. cardiovascular deaths). This prevents us from studying some specific outcomes, such as haemorrhagic versus ischaemic stroke. This is not a criticism of our cohorts, but a call to health and vital registries authorities as well as science funders in LAC. Opportunities for additional follow-up rounds of established cohort studies, or to conduct follow-up rounds of large projects that may have been initiated as cross-sectional surveys could be supported. In this regard, for example, the PERU MIGRANT Study was established as a cross-sectional project, 38 though   with support from UPCH and intentional funders (e.g. GloCal Health Fellowship Program from the University of California Global Health Institute), the investigators managed to conduct two follow-up rounds. 39 In Jamaica, the Jamaica Healthy Lifestyle Study, a repeated crosssectional survey on non-communicable diseases risk factors has recently received National Institutes of Health (R01) funding to expand the cross-sectional work into a true longitudinal cohort. 40 Finally, another limitation is the relative scarcity of data from the Caribbean. The CC-LAC will provide relevant information pooling all available cohorts from this region. We strongly believe that this cohort profile, and forthcoming publications, will allow us to identify additional sources from this region or will stimulate interest among researchers from the Caribbean to conduct local and multi-country cohort studies.
The lack of novel or more sophisticated cardiometabolic risk factors may be a limitation as well; these could include inflammation markers (e.g. high-sensitive Creactive protein) or lipid biomarkers (e.g. apolipoproteins). Some of the pooled cohort studies have, nevertheless, stored blood samples from their baseline rounds. We strongly believe that CC-LAC will enhance visibility of LAC cohort studies and promote regional collaboration. In the future, additional risk factors may be measured in some cohorts or new follow-up rounds may be conducted to overcome current limitations.
Can I get hold of the data? Where can I find out more? Data availability will be prioritized among CC-LAC members. New LAC cohort studies are welcomed to join the CC-LAC. Expressions of interest in additional collaborations will be received and handled by the CC-LAC steering committee. For further details on our work, please contact the corresponding author.

Cohorts Consortium of Latin America and the Caribbean (CC-LAC)
Profile in a nutshell • We have pooled cardio-metabolic risk factors: anthropometrics, blood pressure, lipid and diabetes biomarkers, and non-fatal (stroke, myocardial infarction, revascularization) and fatal (stroke, myocardial infraction, revascularization, sudden death) cardiovascular outcomes.
• Currently, data are not available for collaborations outside the consortium.