Neuroepidemiology in Tropical Health

Neuroepidemiology in Tropical Health covers major neurological diseases of relevance in tropical settings and examines the specificities of epidemiology of neurological diseases in the context of tropical countries that face many challenges when compared to the developed world. Part One focuses on methods and their eventual specificities, and how such methods, like sampling, can be adapted for specific scenarios. Parts Two and Three discuss environmental factors and their consequences for neurology in the tropical world, as well as large geographical areas and their specificities. Finally, Part Four presents relevant neurological diseases in in-depth chapters.

This invaluable information will help readers recognize the various neurological conditions presented, with the inclusion of their aetiologies and treatment in tropical areas. The book therefore fills a gap in the neuroepidemiology literature, with chapters written by an international collection of experienced authors in the field.

• Highlights differences and similarities between neuroepidemiology in tropical areas and temperate zones with a focus on methods and underlying factors
• Covers environmental factors in the tropical world and their consequences for neurology
• Chapters include references (key articles, books, protocols) for additional detailed study
• Includes wide topics of neurological disease in the tropics, not only infectious diseases, but also nutrition and public health

• Language: English
• Publisher: Academic Press
• Release date: Sep 21, 2017
• ISBN: 9780128046258

Book preview

2015;72:1295–1303.

Part I

Fundamental Concepts

Outline

Chapter 1 Methodological Challenges of Neuroepidemiological Studies in Low- and Middle-Income Countries

Chapter 1

Methodological Challenges of Neuroepidemiological Studies in Low- and Middle-Income Countries

Farid Boumediene¹,², Benoît Marin¹,²,³ and Pierre-Marie Preux⁴,    ¹INSERM UMR 1094 NET, Limoges, France,    ²University of Limoges, Limoges, France,    ³CHU Limoges, Limoges, France,    ⁴Institute of Epidemiology and Tropical Neurology, University of Limoges, Limoges, France

Abstract

Neuroepidemiological studies are challenging in low- and middle-income countries. This chapter focuses on methodological problems in studies conducted in these countries and attempts to give the reasons for their limitations. Regional conditions and environmental factors must be given careful consideration in the research design because of the importance of understanding the challenges of living in these environments. Existing information on neurological disorders is often not readily accessible; there is a lack of census data; and migratory patterns into cities make enumeration and sampling even more challenging. As there is usually no well-developed healthcare system, a door-to-door screening process is often the only way to identify those with neurological disorders. The questionnaires and study designs should be adapted from standardized protocols, pre-tested and validated in local conditions. The involvement of all health actors is mandatory to obtain strong evidence which could change practices in a durable way. Neuroepidemiological studies allow us to increase knowledge by applying validated approaches, and are used as decision tools as well as for the elaboration of advocacy/policy documents.

Keywords

Low- and middle-income countries; methodology; neurological disorders; epidemiology; questionnaires

Outline

1.1 Introduction 3

1.2 Study Design and Feasibility in the Field 4

1.3 General Context 6

1.3.1 Geographical Difficulties 6

1.3.2 How to Move Forward? 6

1.3.3 Systematic Review and Meta-Analysis 7

1.3.4 Socioeconomic and Sociocultural Factors 7

1.4 Difficulties in the Availability and Mobilization of Data 7

1.4.1 Medical Data 7

1.4.2 General Population Census 7

1.4.3 Cartography and Geographical Information Systems 8

1.5 Study Design 8

1.5.1 Geographical and Logistical Challenges 8

1.5.2 Demographic and Follow-Up Issues 8

1.6 Clinical and Regulatory Issues 8

1.6.1 International Ethic Rules 8

1.6.2 Suspected Cases Ascertainment 9

1.6.3 Diagnostic and Treatment 9

1.7 Design-Specific Issues 10

1.7.1 Cases Census and Sample Studies 10

1.7.2 Involvement of Institutions and Key Stakeholders 10

1.7.3 Medical Examinations, Biological Specimens and Laboratory Tests 10

1.7.4 Monitoring and New Technologies 10

1.8 Valorization: Scientific and Public Health Issues 10

1.8.1 International Acknowledgments of the Works Conducted in Tropical Areas 10

1.8.2 Advocacy—Policy and International Cooperation 11

1.9 Conclusion 11

References 11

1.1 Introduction

Most of the resources used to deal with neurological disorders around the world are spent in high-income countries. The medical demography (number of specialists) is highly correlated with the level of development of a country: most low- and middle-income countries (LMICs) have a very low number of neurologists, usually only in the cities (even sometimes only in the capital). Therefore, the departments of neurology are situated in the central hospital of these cities and most people with neurological disorders (PWND) do not receive appropriate care.¹ This is known, for example, in epilepsy as the treatment gap (TG).

Since the early 2000s, researchers have identified inadequate case ascertainment in a population, failure to diagnose neurological diseases and lack of appropriate care (which could include medical, surgical and/or social intervention) as the main reasons accounting for the TG. In addition, the TG can further be caused simply by a lack of treatment (unavailability of a drug in an area), as well as failure to continue treatment once started, due to sociocultural or economic factors. The latter is often referred to as the secondary TG.

Whatever the neurological disease concerned, there were only a few incidence studies of neurological disorders in LMICs. Prevalence studies used different study designs and are hardly comparable, usually identifying only main pathologies with the most explicit clinical signs. There were also very few unbiased case–control studies and natural history studies. Several studies mentioned the size of the TG but few analyzed it thoroughly.

Consistently applicable and valid epidemiological methods (descriptive, analytic and interventional) are needed in LMICs to generate comparable data on the burden of neurological disorders, outcome of interventions, risk factors and treatment. Systematic reviews and meta-analyses using such well-executed primary studies would be relevant to all involved in the care of neurological disorders. Epilepsy is probably the most studied neurological disorder and will be taken as a model in this chapter. It illustrates perfectly the various situations which could be transposed to the other neurological pathologies.

1.2 Study design and Feasibility in the Field

Standards of epidemiology are the same everywhere. It should not be considered that degraded methods are relevant for LMICs because of the difficulties in performing research. This is helpful neither for researchers, nor for the quality of research performed in those areas and is definitively not an ethical consideration.

Epidemiology can be divided in two branches: experimental and observational, depending on whether the exposure is assigned by the investigator. Observational studies can be either descriptive or analytic.

The main goal of descriptive epidemiology is to estimate the prevalence (number of subjects with a given characteristic—i.e., epilepsy—divided by the total number of subjects at risk in the underlying population) or rates (number of subjects developing a given characteristic—i.e., epilepsy (incidence rate) or a death (mortality rate)—divided by the total number subjects at risk in the underlying population for a given period (person-time)). To be very clear, the term prevalence rate is totally wrong and should not be used.

Sample size should be sufficient to provide precise estimates of the 95% confidence interval. The design should include this estimation of the number of needed subjects. This of course should be carried out before starting the study. Even in a descriptive study, this calculation is relevant. For example, one could estimate the number of subjects to get an accurate precision of a proportion. Different types of calculations are available for other designs.

There should not be a confusion between this number (which is a key number for the study) and the feasibility to include this number. If the researchers can not include this minimal number of subjects, they should resign, or change their objective or design. Of course, the higher the number, the higher the challenges, because the area in which the inclusion will be performed will expand and the logistical constraints will increase. Another issue for the feasibility is the duration of the study. The period of inclusion should not be too long since the investigators will progressively resign with weariness.

Accurate sampling methods are of utmost importance to include a representative sample of the target population. In LMICs, it is not uncommon to lack recent and valid census data, and in this context, advanced sample methods (e.g., cluster sampling) can be used.

Descriptive epidemiology should not be considered as leading to low-impact studies (by comparison to experimental studies). Well-conducted original descriptive epidemiology might change the consideration of the scientific community’s appreciation of certain health issues that were previously under-considered. For example, sustained efforts of researchers to estimate the high epilepsy burden in LMICs led the International League Against Epilepsy to launch a global campaign against epilepsy called Out of the Shadows.

In terms of design, a cross-sectional study (a snapshot) will be performed to estimate a prevalence. Gold standard methodology in this case is a door-to-door case ascertainment that necessitates investigating the status of the disease in all the subjects present in the area on the day (or during the period) of investigation. This might be a huge challenge, especially for diseases with a low prevalence. In some cases, two-phase design can be used: (1) screening of all the subjects (to identify suspected cases), and then (2) in the suspected cases, confirmation of the diagnosis by the gold standard.

A longitudinal study should be performed to estimate rates. The major challenge in LMICs is the maintenance of a cohort for a relevant predefined period; that in some cases might be long (i.e., the follow-up of the subjects). This major difficulty in conducting cohorts is reflected in the field of epilepsy, by the fact that a recent meta-analysis identified 46 prevalence studies while only 8 incidence studies were available in sub-Saharan Africa.² It is also sometimes possible to reconstitute cohorts in the past, called retrospective or historic cohorts.

Analytic epidemiology refers to studies aiming to identify associations between an exposure and an outcome. Two study designs can be used: case–control studies (leading to odds ratio estimation) and exposed/non-exposed cohort study (leading to relative risks estimation).

The case–control approach lies in the inclusion of cases (patients with the disease) and controls (patients proven to be without the disease). Exposure is then retrospectively assessed in both groups. The main methodological issues in case–control studies are: (1) the difficulty to recruit a sample of controls that provides a good picture (representative) of the subjects without the disease, and above all (2) the validity of exposure estimation (especially when no historical files are available and when a recall bias could be important). These two issues might lead to biased results. Given these limits, the level of evidence given by case–control studies is considered as low.

An exposed/non-exposed cohort study lies in the inclusion of exposed (subjects with the exposure at baseline) and non-exposed subjects (subjects proven to be without the exposure at baseline) who are followed for a relevant period, driven by the pathophysiological underlying mechanism and latency. As for the descriptive cohorts, the main issue is the follow-up of the subjects. Given that exposure is assessed at baseline and that the occurrence of the disease is followed-up, the level of evidence of analytic cohort studies is superior to case–control studies.

As a rule of thumb, case–control studies are best suited for a first evaluation of an association, when a long period of latency between exposure and the disease is expected, for rare diseases, for frequent exposure, and when the time constraints to deliver the answer is short. Conversely analytic cohort studies are best suited for providing a confirmation to a suspected association with highest level of evidence, for frequent disorders and rare exposures. At this stage, in the field of epilepsy for example, we face a lack of analytical studies. For sub-Saharan Africa, Ba-Diop et al. identified only eight analytic studies, only of case–control design.²

An experimental study consists of comparing an outcome in relation to the exposure to an intervention (drug, medical device, complex health intervention) that is assigned by the investigator. In practice, randomization is used to allocate the intervention. Results from experimental studies represent the highest level of evidence. The most famous design is the randomized two parallel arms control trial in which a sample of subjects is randomized to receive either the standard intervention (or a placebo, when none standard is available) or the experimental intervention (to be assessed). The main methodological components are individual randomization, blinding and intent-to-treat analysis that can lead to initial and sustained comparability between groups. When individual randomization is not possible (e.g., when the intervention includes the health system organization or when a contamination bias between subjects is possible), randomization can be performed on groups of subjects (clusters). The drawbacks of this approach are the increasing sample size and the complexity of statistical analysis.

A review dedicated to randomized controlled trials focused on neurological disorders in developing countries³ pointed that (1) there has been an exponential evolution of the number of clinical trials in the recent years (64% of trials in between 2004 and 2014), (2) the Asian continent contributed significantly to the trials performed in LMICs (48%), followed by Africa (36%) and Latin America (16%), (3) there is a fairly good coverage of pathologic fields including non-communicable diseases, (4) also there is an increasing diversity of intervention types (therapeutic 72%; preventive 17%; rehabilitative 10.5%; diagnostic 0.5%) (5) besides there is a lack of early-phase trials (phases I and IIa), and (6) while the methodological quality of the trials has improved with time, there is still a tremendous need for improvement of some critical methodological issues. The development of clinical research units in LMICs is key for local researchers to be able to apply the highest standards in terms of methods, data-management and statistical analysis.

In some case, for ethical, political or logistical constraints, the randomization might be impossible, even cluster randomization. This is case for example when the intervention is complex (i.e., based on a modification of the organization). In this context, researchers can use quasi-experimental design based on the comparison of groups whose constitution is not random-based (e.g., choice of the investigator or location of the subject). Design can be single difference impact estimates which rely on the comparison between outcomes of a group with the intervention as compared to a control group without. This design is limited by the confusion driven by the possible lack of comparability between both groups at baseline. Another design consists of comparing the outcome before and after the intervention in a single group. This design is limited by the possibility for the outcome to evolve naturally or be impacted by any coincidental events. Another more robust design, that combines both previous approaches, is difference-in-differences which compares the changes in outcome over time between treatment and comparison groups to estimate the impact of the intervention. This design allows the initial difference between groups to be removed.

The absence of comparability between groups at baseline for known and unknown confounding factors can highly impact the results of quasi-experimental studies. Advanced statistical methods such as propensity score and instrumental variables are available to try to take this potential source of bias into account. While these approaches are useful, their application has been proved to be suboptimal in the literature.⁴ Besides, residual confounding is difficult to exclude.

1.3 General Context

1.3.1 Geographical Difficulties

There are common characteristics in the LMIC environment, in particular in the tropical zone. First, of course, the climate is specific. There is often poor sanitation, which promotes the spread of infections. Environmental toxins are common, and malnutrition is frequent. Scarce resources also lead to poor access to care and greater inequality in access.

As an example, meta-analysis of prevalent studies of active epilepsy and lifetime epilepsy throughout the world found a higher prevalence in rural areas of LMICs, a lower one in urban parts of LMICs, and the lowest prevalence was found in higher-income countries (HICs).⁵ The authors felt that access to healthcare was an important determinant of these results. This distribution, particularly explicit for epilepsy, could certainly be applied to many neurological diseases. As another example, population-based cross-sectional and case–control studies of active convulsive epilepsy were carried out in five centers in sub-Saharan Africa.⁶ To reduce heterogeneity of findings, the authors used the same methods and definitions between regions. Interestingly, they found that heterogeneity could be accounted for by markers of birth trauma, exposure to a range of parasites and other factors including malnutrition.⁶ The relationship between epilepsy and undernutrition is a complicated one. To ascertain which came first, one would need a long-term cohort study.⁷

There is inadequate public health sector and systems. There are many specific sociocultural characteristics that may by themselves be associated with neurological diseases. The population is young, with a low life-expectancy compounded by the HIV/AIDS pandemic. There is currently rapid and uncontrolled urban and suburban growth, again creating a mismatch in healthcare provision. Political factors, i.e., unstable governments, multiple levels of decision-making and the low priority given to health programs, worsen the situation. Curative and preventive neurological health services are often very weak or lacking.

It is important for any investigator to assess this big picture, not only from the point of view of designing a good study, but also to better place neurological disorders in the context of local difficulties. Carrying out epidemiological research in these settings is then very challenging for a variety of reasons—methodological, logistic, political, economic, ethical and often the low-perceived value of such works.

1.3.2 How to Move Forward?

Before any research, a review of literature is essential to draw up the state of the art of the problematic studied. However, studies of neurological disorders in the tropics are not numerous, and these gaps are amplified when one wishes to find original articles on the investigated country. Studies published in indexed journals are rare in the field of tropical neurology, and it is often difficult to obtain historical information (usually retrospective studies are impossible due to non-computerized medical records and lack of conceptual data models). The few existing studies are often difficult to compare because their methods are not standardized.²,⁸

Documentary research should go beyond indexed journals in which publications, which are very selective and do not always provide the detailed information sought. Medical or pharmaceutical theses in medicine or pharmacy, or PhD theses in public health (or in a specific discipline) represent an important source of detailed information. For over 30 years, the Institute of Neurological Epidemiology and Tropical Neurology has been supplying an original database on scientific productions dealing with neurological disorders, mainly on the African continent (http://www.unilim.fr/ient/base-bibliographique-de-l-ient/). There are also specific databases to be exploited depending on the pathologies investigated: for example, in epilepsy, it is pertinent to consult the Wan Fang (online) database which archives the Chinese and English journals that deal with studies conducted in mainland China (www.wanfangdata.com). It should be noted that journals with no impact factor (to date) are also emerging on the international scene to promote the results of original studies (e.g., Neurology Asia, http://www.neurology-asia.org/).

Finally, ministerial documents are often overlooked as they are a valuable source of information. The case of epilepsy in Lao PDR is explicit: when Tran et al. initiated the first on-site investigations in 2004, they did not find any published data.⁹ However, through further research, they identified a survey conducted by the Ministry of Health in the mental health unit of the Mahosot Hospital (located in the capital Vientiane). This survey was mentioned in a report stored in the Ministry.

1.3.3 Systematic Review and Meta-Analysis

There are a few systematic reviews and meta-analyses concerning some of the above-mentioned issues. Most of them have some or all the following problems: insufficient scrutiny of the existing literature, methodological differences in the included primary studies, (including differences in case ascertainment, various definitions or criteria of diagnosis) which could lead to heterogeneity, lack of standardization on age or sex, inclusion of samples that may not be representative of the general population.

The meta-analyses of incidence studies,¹⁰ prevalence studies⁵ and TG in epilepsy¹¹,¹² have demonstrated a high degree of unexplained heterogeneity. This is due to the methodological differences between primary studies but also could be due to which studies to include or exclude from the meta-analysis. Problems with meta-analyses are greatest when there are few primary studies, as with incidence studies, and where there is wide variation between estimates of these studies.¹⁰,¹³

1.3.4 Socioeconomic and Sociocultural Factors

Consideration of sociocultural factors and economic status is fundamental to conducting investigations in LMICs. These two themes are discussed in detail in other chapters of this book, but the constraints and limitations that these two dimensions entail in the conduct of neuroepidemiological research programs should be mentioned.

Indeed, in so-called traditional societies, beliefs and perceptions about neurological diseases very often lead to a marked stigma because of the clinical expression of these pathologies. Thus, affected people generally do not want their neighbors or co-workers to know about their state of health. The consequences are dramatic because this directly constitutes a barrier to diagnosis and treatment seeking, especially in rural areas where anonymity is more difficult to preserve.

In addition, the economic difficulties are known to be associated with the treatment deficit, although this phenomenon still deserves specific research to be better understood. Indeed, sometimes, people seeking care in traditional systems afford much more than they would have to in a medical system.

Given the large share of the informal sector in LMICs, income information is often difficult to determine because it is irregular and a large majority of people prefer not to disclose it. How to deal with the issues of education and socioeconomics is fundamental in understanding the difficulties of access to care. In 2012, Cooper et al. found that only 7% of poverty indicators were commonly measured by standardized validated methods.¹⁴ Conversely, in 2012, Mbuba et al. showed that assessments based on socioeconomic status were very common for neglected diseases, but were much less frequent for neuroepidemiological investigations.¹⁵

It is therefore strongly recommended to involve in the research programs disciplines of human and social sciences such as ethnology, anthropology, sociology and of course health economics.

1.4 Difficulties in the Availability and Mobilization of Data

1.4.1 Medical Data

Once epidemiological research in the tropics is initiated, the question of mobilizing information about known patients arises very quickly. In most LMIC countries, the health system is organized per a pyramidal system where the top is constituted by the central hospitals and the base by the primary health centers. The mobilization of useful information is very uneven depending on where you are in the pyramid: easier and often computerized in the upper part (central hospitals of the capital and provincial hospitals) and very difficult to access for in the lower part (district hospitals and primary health center). Although there is feedback from the bottom to top of the pyramid, the latter are generally aggregated (and therefore not very exploitable because they aim at quantitative assessments by groups of diseases for health surveillance or the supply of decentralized pharmacies). Regardless of the level of the pyramid where the research is located, useful information could be found in patient files in paper format. This could considerably impact the access and collection of data, computerization time, and supposed first that a safe storage of these files was