Contents

Chapter 1: About MARA/ARMA

Chapter 2: Focus A - Empirical Data Collection

Chapter 3: Focus B - Modelling

Chapter 4: Related Projects

Chapter 5: Conclusions and Outlook

References
Appendix 1: MARA/ARMA Publications
Appendix 2: Environmental Data Sets
Individuals Who Contributed to MARA/ARMA

Preface

At a time when the fight against malaria is regaining interest among the international community, and when major new initiatives such as the "African Initiative on Malaria" or "Roll Back Malaria" are being established, I would like to congratulate the report’s authors and collaborators for their contribution to this piece of work.

The development of the final products (the detailed malaria endemicity maps, and in particular maps identifying regions with epidemic malaria potential) required the field collection of a vast amount of information. It is not always easy to find existing information. The MARA/ARMA products will allow us to better understand the distribution of malaria in Africa, at both the country and regional levels. This is a formidable planning and management tool now available to malaria control programmes.

Beyond disease distribution maps, readers will find in this first technical report the outline of a unique epidemiological and climatological database accessible to various audiences, including health programme managers, epidemiologists, medical doctors, and researchers. Even better, the malaria transmission models that are being developed will allow users to foresee, with a minimum of information, the transmission dynamics in a given zone and thereby to take appropriate measures in anticipation.

It is my wish that the work which has been initiated will be completed and, above all, that all national malaria control programmes will make good use of it.

Dr Ebrahim M. Samba
Regional Director
WHO-AFRO

Préface

Au moment où la lutte contre le paludisme reprend de l'intérêt au sein de la communauté internationale et où se concrétisent des initiatives d'envergures, telles que l' "Initiative Africaine pour la lutte contre le paludisme au 21 ème siècle"  et le "Roll Back Malaria", je voudrais féliciter les auteurs et tous ceux qui sont impliqués dans ce projet, pour le travail accompli.

En effet, le résultat final que sont entre autres les cartes détaillées sur l'endémicité du paludisme et surtout la cartographie des zones à potentiel épidémique a nécessité la collecte d'une masse très importantes d'informations sur le terrain. Il n'est pas toujours facile de retrouver ces informations. Les produits du projet MARA/ARMA permettront de mieux connaître la distribution du paludisme en Afrique, dans un pays, dans une région. En cela, ils sont de formidables outils de planification et de
gestion des programmes antipaludiques.

Au delà de la cartographie, les lecteurs trouveront en ce premier rapport technique, la description d’une base de données climatologiques et épidémiologiques unique, utilisable par les planificateurs des programmes de santé, les épidémiologistes, les médecins et les chercheurs, entre autres. Mieux encore, les modèles statistiques développés sur la transmission du paludisme permettront aux utilisateurs, de pouvoir prévoir, avec un minimum d'informations, la dynamique de la transmission du paludisme dans une zone donnée et ainsi anticiper les mesures à prendre.

Il reste à espérer que le travail qui a débuté soit conduit à son terme, et surtout que tous les programmes nationaux de lutte contre le paludisme en fassent la meilleure utilisation possible.

Dr Ebrahim M. Samba
Directeur Régional de l'OMS pour l'Afrique

Acknowledgements

The production of this document has been sponsored by the South African Medical Research Council.

Recently, the MARA/ARMA collaboration has been the recipient of a Multilateral Initiative on Malaria (MIM) award.

A special thanks to Colleen Fraser for her significant contribution in maintaining the central MARA/ARMA database, and to Carrin Martin for logistical support.

Many individuals have contributed to the success of MARA/ARMA so far. A list of individuals who have contributed data or helped locate information is contained in Appendix 3.
 

This technical report constitutes the first summary document prepared by the MARA/ARMA collaboration. It aims to give an overview of activities and progress. Although great care has been taken in its preparation, it is likely to be incomplete and incorrect in some points.  We welcome any input, both on this document and on the MARA/ARMA initiative in general. Comments can most easily be sent using the attached questionnaire. Comments from malaria control workers and malaria researchers in endemic areas are especially welcome.

List of Abbreviations

Glossary

Foreword

The National Malaria Research Programme (SAMRC) GIS effort was started in 1989 with very limited funding and with the focus of mapping malaria risk in South Africa, towards more targeted control. With the continued support of the programme leader, Dr Brian Sharp, the division has developed a core expertise in GIS and database management. It was this expertise which was recognised by Dr Don de Savigny following the IDRC sponsored meeting on GIS for Health and the Environment in Sri Lanka, 1994 (de Savigny and Wijeyaratne, 1995). Subsequent discussions between himself and Dr Bob Snow concerning the need for an initiative such as MARA/ARMA resulted in two small workshops funded by the WHO/TDR Task Force on ITNs and IDRC respectively. The conceptual proposal was developed at an initial meeting consisting of Fred Binka, Bob Snow, Christian Lengeler and myself. The second workshop at which the full proposal was developed, included a larger group of 15 people (entomologists, epidemiologists, geographers).

The subsequent growth and success of MARA/ARMA must however be attributed to the dedication of the co-investigators and data co-ordinators at the regional centres and their data sources.  I would specifically like to mention the initial momentum provided by Dr Bob Snow in terms of the development of the data proforma and Colleen Fraser for capturing this into the stand alone data application. This was a critical first step in getting the collaboration operational and in ensuring standardisation. Subsequent important steps were the search for baseline digital data sets and the training of regional staff in Durban. There are many others to thank but I would like to mention Marlies Craig who has shown exponential ability in terms of her acquisition of much needed raster GIS skills and Carrin Martin for consistent logistical support. It was a privilege to work with African scientists of the calibre of the MARA/ARMA regional co-ordinators and co-investigators.

I believe that MARA/ARMA and this 1st preliminary product is a good example of how GIS can be usefully employed in health in a cost-effective manner and is an outstanding example of a networked collaboration. It contains all the essential components including a robust database and an understanding of disease determinants and their spatial scale.

The future of MARA/ARMA is now secure with the recent grant of the Multi-lateral Initiative for Malaria (MIM) as well as the co-funding of the Highland module by the TDR task force for Health and the Environment. A measure of recognition is the recent decision by Roll Back Malaria (RBM) to place a staff member at the co-ordinating centre in Durban, to support RBM activities. The central challenge that lies ahead for the collaboration is the dissemination of existing and new products and GIS skills that will contribute significantly to new initiatives against malaria, such as MIM and RBM. In the research arena MARA/ARMA will need to work more closely with climatologists towards a better understanding of unstable malaria and
its prediction.

In conclusion Health GIS is a research field which is in its infancy and constitutes an approach which has little applicability in certain diseases. There is however no doubt that in the case of environmental vector borne diseases such as malaria, a GIS approach is highly appropriate

David leSueur
Principal Investigator
MARA/ARMA collaboration

Executive Summary

The distribution, transmission intensity and clinical consequences of malaria in Africa vary greatly across the continent. Africa experiences a complete spectrum of malaria epidemiology ranging from intense perennial transmission to unstable, epidemic-prone areas. This has implications for the planning, targeting and implementation of control activities at continental, national and regional levels. Despite the importance of these facts, a continental perspective of  where (distribution), how much, (transmission intensity), when (seasonality), why (environmental determinants) and who is affected (populations at risk) does not currently exist.

The Mapping Malaria Risk in Africa/Atlas du Risque de la Malaria en Afrique (MARA/ARMA) collaboration was created to fill this gap and to establish a continental database of the spatial distribution of malaria in Africa. To achieve this aim it relies on two complementary approaches:

A. The formation of a continental database of available malariometric data representing precisely geographically positioned survey data from published and unpublished sources in 44 countries.

The largest possible number of existing malaria survey data are being collected from published articles, university theses, ministry reports and unpublished work by research and other institutions. Parasite ratio data in children are by far the most commonly available data in Africa and MARA/ARMA has therefore focussed on them. Data on other transmission indicators such as entomological inoculation rates and infant parasite surveys are also being collected. Incidence rates derived from routine statistics will only be used in the few countries with reliable health information systems. In order to carry out this large data collection exercise MARA/ARMA has established five regional (and two sub-regional) centres, each with a data collector. These centres have also been equipped with Geographic Information System (GIS) skills and digital data sets necessary to geo-reference the collected data.

B. The development of environmentally determined models of continental limits of transmission risk, in order to supplement the data collection of empirical data in areas where no such data exist.

Malaria distribution, seasonality, frequency and transmission intensity is being investigated at various spatial scales. This approach has provided an immediate starting point for describing malaria epidemiology on a continental and national scale.

Both approaches are complementary and the empirical data have been used to develop and validate models in a number of countries. The malaria maps have been combined with other data sets such as population and administrative boundaries, thus providing a fundamental resource for planners (district-national-regional level), donors and researchers.

MARA/ARMA is a unique example of a Pan-African collaboration of researchers and control managers. The first products of the MARA/ARMA collaboration are now being distributed. The collection of basic malaria data and risk maps are providing national governments, donors and international agencies with a more empirical basis for evidence-based, strategic planning for malaria control. Complementary projects aiming to describe highland malaria and to establish a malaria vector map for Africa are also part of the MARA/ARMA collaboration. Through a closer collaboration with malaria control programmes and scientists on the continent the collaboration will build upon these experiences to assist the global efforts to Roll Back Malaria into the next millennium.

Résumé

La distribution et l'intensité de la transmission du paludisme sont loins d'être homogènes en Afrique. L' on trouve en Afrique un large spectre de situations épidémiologiques, de la malaria pérenne de haute intensité à la malaria instable épidémique.  Cela est d'importance pour la planification, la hiérarchisation et l'implantation des activités de lutte aux niveaux continental, national et local. Malgré l'importance de ces aspects géographiques notre compréhension actuelle de la distribution (où ?), des déterminants (pourquoi?), de la quantification (combien?) et de la saisonalité (quand?) de l'endémicité palustre est très imparfaite.

L'objectif principal de la collaboration Mapping Malaria Risk in Africa / Atlas du Risque de la Malaria en Afrique (MARA/ARMA) est d'établir une base de données à l'échelon continental sur la distribution du risque de la malaria. Deux approches principales sont utilisées pour atteindre cet objectif.

A. Une collection aussi complète que possible de données paludométriques existantes pour le continent est constituée pour 44 pays. Chaque point est défini géographiquement de manière précise.

Les sources utilisées sont les publications scientifiques, les thèses universitaires, les rapports de ministères et les documents produits par des organisations de recherche ou de développement. Les données de prévalence parasitologique sont de loin les données les plus largement disponibles et la collaboration MARA/ARMA s'est concentrée sur ces dernières. Toutefois, d'autres données telles que des données de transmission entomologique, et les taux de parasitémie chez les nourrissons sont aussi collectées. Les données sur l'incidence des cas de malaria provenant des statistiques de routine des ministères de la santé n'ont été collectées que dans un petit nombre de pays où ces statistiques sont réputées fiables. Afin de conduire ce large exercice de collecte, la collaboration MARA/ARMA a établi cinq centres régionaux (et deux centres annexes) avec dans chacun de ces centres un collecteur employé à plein temps. Ces centres ont été équipés avec des Systèmes d'Information Géographiques (SIG) et le personnel formé à leur utilisation.

B. La deuxième approche repose sur une modélisation de facteurs climatiques visant surtout à supplémenter la collecte des données paludométriques dans les zones où ces données ne sont pas disponibles.

La distribution de la malaria, sa saisonalité et l'intensité de la transmission ont ainsi p? être étudié a diverses échelles géographiques.  Ceci a permis de donner  rapidement une première description de la transmission de la malaria à l'échelle du continent .

Ces deux approches sont complémentaires et les données paludométriques ont été utilisées pour valider le modèle climatique dans quelques pays.  Ces cartes ont été combinées avec d'autres  données (démographiques et administratives) pour produire de cartes intégrées utilisables dans le cadre des activités de lutte a divers échelons.

MARA/ARMA est un exemple unique de collaboration panafricaine entre des chercheurs et des personnes travaillant dans des programmes de contrôle.  Les premiers produits de cette collaboration commencent à être distribués. Les cartes malariométriques vont ainsi fournir aux programmes de contrôle gouvernementaux, aux donateurs ainsi qu'aux agences internationales une base de données utilisable pour une planification plus rationnelle des resources et activités. Deux projets additionnels sur la malaria d'altitude ("highland malaria") et sur la cartographie des anophèles vecteurs font aussi partie de la collaboration MARA/ARMA.

A travers une collaboration plus étroite entre les chercheurs et les programme de contrôle l'expérience accumulée servira à soutenir les efforts actuellement en cours pour améliorer le problème du paludisme en Afrique.

Chapter 1.  About MARA/ARMA

Introduction

Malaria - the Problem

Fortunately the situation has recently been receiving renewed attention by national and international health and funding organizations. In particular, the recent call by the new Director-General of the World Health Organization, Dr Brundland, to "Roll back malaria" is encouraging in this respect. The development of continental to district planning tools is therefore very timely.

New Hope for Malaria Control

Many factors affect the choice of malaria control methods in a region: endemicity, vector species and behaviour, seasonality, disease patterns, health services factors and more. Since all these factors are not distributed equally across the continent, accurate, relevant and timely information on them is required before malaria control can be planned and resources allocated
properly.

For this purpose, maps offer an ideal way of displaying complex information in a way that is intuitively understandable and instructive. Some factors, such as the availability of health and malaria control services and existing infrastructures, can be simply mapped to give a visual representation of information already available in non-spatial format. For other factors, such as
malaria endemicity, disease patterns, vector distribution or seasonality, spatial methods may be used to produce seamless data surfaces (where information is available for every point in space). Such continuous data coverages provide information in a format that is relevant both at the continental scale, where it provides a quick overview of the situation, down to the national and sub-national level, where it provides detailed guidance and answers specific questions. In both cases problems can be tackled in a systematic and informed way.

Mapping Malaria

The Need for Maps

Secondly, in areas of endemic malaria, the pattern of severe malaria disease has been shown to vary according to the intensity of the transmission (Snow et al., 1997). In areas with lower levels of endemicity the disease pattern was found to be dominated by cerebral disease forms in older children (over 2 years of age) while in areas of very high transmission the disease pattern was dominated by severe malaria anaemia in young children and infants. This age-dependance of malaria disease according to the intensity of malaria transmission has great practical importance for the preventive and curative services since both the target age group and the clinical care need to be adapted .

Thirdly, the protective effect of control measures such as ITNs might be related to initial level of transmission intensity, although the total number of lives saved by the intervention in the four recently completed trials in Africa was remarkably similar in all settings (Lengeler, 1998). Other measures such as larviciding might only be viable in highly seasonal or fringe areas where the mosquito populations are sufficiently unstable to be affected by such interventions. Vaccines may become particularly effective in areas of very high endemicity, where the initial boost of immunity offered by the vaccine will then be sustained by the immuno-response to frequent inoculations in high-transmission zones. In-door insecticide spraying has to be applied according to the seasonality of malaria transmission - both in terms of timing and choice of insecticide. A chemical effective for nine months is useless in areas of perennial transmission.

Mapping - Past and Future

In spite of the enormous problem malaria is causing in Africa, there is a scarcity of basic data and a lack of understanding of the situation. Details on malaria risk and severity, and fundamental perspectives of where (distribution) why (environmental determinants) how much (transmission intensity) and when (seasonality) malaria occurs, do not exist. No-one has attempted to define which populations are truly exposed to risk of malaria. In addition, the lack of accurate diagnostic methods and reporting systems has resulted in unreliable records of malaria-caused death and sickness from routine sources.

We therefore need to rethink how to define endemicity and how we may map malaria risk in order to better support planning and programming of malaria control. In view of the renewed interest in the control of malaria in Africa such maps needs to be drawn up systematically for the entire continent.

Providing timely and adequate data is an essential pre-requisite of evidence-based planning which is increasingly applied to all areas of public health. This necessity is at the core of the Mapping Malaria Risk in Africa/Atlas du Risque de la Malaria en Afrique (MARA/ARMA) Collaboration. The purpose of this technical report is to present the MARA/ARMA collaboration and its initial products, and to solicit feed-back from potential users.

MARA/ARMA: an International Collaboration

From the start the MARA/ARMA collaboration has been an international and inter-disciplinary undertaking (with biologists, medical doctors, statisticians, geographers), initiated by a group of malaria researchers active in Africa, with close links to national malaria control programmes. Today we have a real opportunity to develop a dynamic atlas of disease risk and severity through international collaboration. This has been due to the availability and affordability of GIS software, Internet connectivity in Africa, a spirit of international collaboration, growing emphasis on evidence-based planning and increasing availability of
global data sets (population, climate, satellite imagery, etc.)

Double Focus

A   Empirical Data Collection

For the data collection process, Africa was divided into functional regions, with five regional centres and two sub-centres responsible for 5-7 countries in their region (Figure 1). The regional centres are located at existing institutions throughout the continent (see list on the inside of the cover), each with a full-time data co-ordinator supervised by a co-investigator. Since it was necessary to geo-reference all collected data, the data co-ordinators were trained in the use of GIS and equipped with the necessary hardware, software and digital data sets. Although the MARA/ARMA activities are located at certain institutions, the initiative itself is non-institutional and runs in the spirit of an open collaboration involving countries and people within Africa who want to participate and contribute.

In addition to the malaria data, data on vectors and on epidemic highland malaria were also collected in the frame of closely related initiatives.

B  Modelling

The main rationale behind the modelling was:
(1) malaria data is distributed unevenly both within and between countries, and (2) the data points on their own are not sufficient to complete the picture of malaria endemicity for various reasons, which are discussed in Chapter 3. We need both the point malaria data as well as the continuous coverages of environmental factors which affect malaria transmission, and then we need to extrapolate, using spatial models, to complete the picture.

The factors that determine the distribution and severity of malaria are diverse and complex, but climate can be considered the major determinant. Temperature and rainfall limit malaria to the warm, humid regions of Africa, where the mosquitoes and parasites can breed and develop, and transmission can occur. Such climate data are increasingly available in the form of data surfaces that can be manipulated in a GIS and can be used to make predictions about data-poor areas. Both approaches - the data collection and the modelling - are run in parallel and the empirical data have already been used to develop and validate environmental models of endemicity in a number of countries.

Products

Products will be presented in more detail in the following chapters. They include an atlas of malaria risk in Africa, poster-size maps, books and pamphlets, digital maps and malaria databases on CD and the Internet, as well as publications in scientific journals. Training in both GIS and the use of malaria data has been an important component of the collaboration from the start and this will be further strengthened during the next phase, with support from the Multilateral Initiative on Malaria (MIM).

Objectives

1. To carry out a comprehensive collection of available malaria data for the African continent;

2. To highlight areas of no or sparse data;

3. To spatially define factors which exclude malaria (eg. absence of population, high altitude, deserts) in order to delineate zones where malaria transmission is unlikely to occur;

4. To define malaria risk categories in terms of climatic and environmental data and to develop models able to predict malaria risk over the whole continent;

5. To map areas at risk of epidemic malaria;

6. To develop a base-map of malaria risk in Africa, at the level of second administrative unit (“district”) which integrate geographic, population and environmental factors;

7. To make continental and national risk maps available to national, regional and international organisations;

8. To contribute to the training of malaria control staff in the areas of GIS with a view to using it in evidence-based health planning.
 

Chapter 2. Focus A - Empirical Data Collection

What Data?

Disease Incidence

There are exceptions to this and routine data from countries such as Namibia, Botswana, Zimbabwe and South Africa have been used. Malaria incidence data collected in the frame of special surveys - for example in the frame of an intervention trial - have also been included into the MARA/ARMA database.

Entomological Inoculation Rate

Parasite Prevalence (Parasite Ratio)

Malaria is described as endemic when there is a measurable incidence both of cases and of natural transmission over a succession of years. The most widely cited method to describe levels of endemicity was based on spleen rates and was proposed during the malaria conference in Kampala in 1950 (WHO, 1951). Subsequently the definition was revised, based
on PR (Metselaar and Van Theil, 1959):

               Hypoendemic:   PR in 2-9 year olds <10%
               Mesoendemic:   PR in 2-9 year olds 11 to 50%
               Hyperendemic:  PR in 2-9 year olds constantly >50%
               Holoendemic:   PR in infants constantly > 75%

Although widely used this classification creates artificial groups from a natural continuum and hides important differences between localities. Several other attempts have been made to classify endemicity (Gill, 1938; Lysenko and Semashko, 1968). This has involved the definition of malaria "paradigms" based on parasite and vectors species, level of transmission, population, social-, behavioural- and economic characteristics, health infrastructure, use of drugs, influence of development projects, climate and geography. The categories were derived so as to create a malaria classification that would fit with available intervention tools. Unfortunately most of the required information is not available widely in Africa.

In the frame of the MARA/ARMA collaboration we have decided to use PR as the main malaria indicator because of its wide availability and reliability. Validation against other indicators (such as EIR and available incidence data) may lead to a re-definition of the classical endemicity cut-offs.

The age categories used by Metselaar & Van Thiel(1959), i.e. 2-9 years for all categories except for holoendemic areas, may be meaningful, but they remain arbitrary. Omumbo et al. (1997) investigated whether it mattered which age category was used to calculate endemicity. Even though the PRs varied between different age categories, (the difference between the 1-5 and the 6-9 year categories was most pronounced), only rarely (8% of the time) did they fall in different classes of endemicity. On the basis of these results it was decided to concentrate on PRs in children below 10 years, excluding infants, irrespective of the age groupings each survey used.

Other effects on PR needed to be considered: for example the effects of season, which often causes dramatic differences between wet season PR and dry season PR. In addition to the within-year variability there also exists clear inter-annual and longer-term variability in many parts of Africa. To keep track of this, information on date and season needed to be recorded, as well as any other information about prevailing conditions at the time of the survey.

Differences in PR over small distances have been well described in a number of settings (Cattani et al., 1986; Jambulingham et al., 1991; Sharp and leSueur, 1996; Smith et al., 1995). They are dependent not only on random chance distributions of infections, but also on systematic variability such as distances from breeding sites, human behaviour, type of housing and many more. This kind of information is only rarely recorded, and MARA/ARMA operates at a level where such small area variation is not accounted for, but forms part of the "background noise", that may partially obscure larger-scale relationships with environment. Nevertheless, information on the presence of swamps, irrigation or rice cultivation was recorded if given in the report.

Survey types and sampling method have also been considered, since fever surveys and clinic based surveys tend to over-estimate the community infection prevalence. School surveys, which inevitably only include children well enough to attend school, and often include older children in which immunity may reduce infections, tend to under-estimate the PR. The ideal survey design was therefore based on a true random sampling of children to reflect the geographic and age-distribution of a well defined community.

It was not known how much data would be found, so no exclusions were made on any of the above criteria at the collection stage, but all details were recorded so that quality-coding of the data would be possible at a later stage. The diagrammatic model of the MARA/ARMA data base (Figure 2) shows all information that is being collected, if reported.

Historical Malaria Maps

Data Collection

How Are the Data Collected?

A data proforma was used for capturing the information in the original reports. It consists of different sections which can be assembled in different ways to cater for incorporation of different types and quantities of malaria data, depending on the reference source. This allows for different types of surveys, undertaken in different geographic locations, at different times to be transcribed to one standard proforma. Operating procedures guide the data collectors through the process of extracting data from reports and publications.

A relational database (Figure 2) was then designed in Microsoft Access™ to accommodate the full complexity of all data relationships. The structure permits future growth, incorporation of new data entities, and a completely flexible means of combining selected data for analysis. To this effect, a stand-alone application conforming to the proforma was created in MSAccess / VisualBasic™ to ensure standardisation. A certain amount of data validation is performed automatically at the point of entry by the system. Verification and further validation is completed at the Durban co-ordinating centre and queries forwarded to the data coordinators.

Geo-referencing the Data

Search Strategy and Data Sources

Secondly, because unpublished data is rarely in a format that allows it simply to be mailed, a personal visit by the data coordinator was needed to get access to this source of information. The regional data coordinators started to visit all identified institutions likely to hold unpublished documents in the countries under their control. This included the relevant ministries, universities and research institutions. This long-term work, which is very cost- and time intensive, is still ongoing.

Finally, international archives in Africa and in Europe (WHO Geneva, Paris, Antwerpen, Lissabon) are being searched and all identified documents abstracted. As of mid-1998, this intensive search has allowed to identify 1014 individual reports / data sources. Of these, 29% were from scientific journals, 45.5% were Ministry of Health reports, 19.5% were other reports from international bi-or multilateral agencies,  NGOs, mission hospitals or personal communication and 5.9% were found in postgraduate theses (Figure 3). The large amount of data from unpublished sources confirms the need for country visits to locate unpublished records.

From these various data sources, 2529 prevalence ratios conducted on children under 10 years of age between 1926 and 1997, 147 records of malaria incidence and 52 entomological inoculation rates were extracted. A map of all survey points collected to date is shown in Figure 4. One report may contain several data points, one data point being one set of results unique in source, space and time, but may still be broken down into different age categories. The abundance of collected surveys is remarkable, considering the logistic constraints in the collection process. So far some countries have been well covered, while in others data is very sparse or difficult to access. Once the collection process is completed, remaining data gaps will need to be  filled by modelling malaria risk on the basis of other data-rich regions where conditions are similar.

Chapter 3. Focus B - Modelling

Modelling: Why?

Modelling: What?

Modelling: How?

What Do We Know?

Rainfall provides mosquito breeding sites and increases humidity, which enhances mosquito survival, and the relationship between mosquito abundance and rainfall has been illustrated repeatedly. Unfortunately the relationship is not direct: a specific amount of rain does not lead to a specific mosquito density. By examining rainfall patterns in known malaria and non-malarious areas, it appeared that an average of 80mm per month, for at least three to five months, was a reasonable requirement for annual malaria transmission(Craig et al, 1998).

Malaria Distribution Model

Perhaps the most fundamental question is distinguishing malarious from non-malarious areas. A great many factors affect malaria transmission, but at the continental level the limiting factors are climatic. These control the development of both parasites and mosquitoes and also mosquito survival. Together they determine whether transmission is likely to occur or not.

Recently spatial temperature and rainfall data surfaces for Africa, at a resolution of about 5km2, have become available (Hutchinson et al., 1995). These consist of extrapolations from weather station data (from over 60 years) and elevation data, and are suitable for modelling the long term presence or absence of malaria.

Defining Malaria Distribution

Going by what is known about the relationships (Figure 6) one can set limits, above which we are sure that transmission can take place, and another limit, below which transmission is more or less impossible. In this way one is expressing the suitability of climate for malaria transmission, and defining the areas where malaria transmission can be expected to occur.

Parasite development ceases at 16°C, but transmission below 18°C is unlikely because very few adult mosquitoes survive the 56 days required for sporogeny at that temperature, and because mosquito abundance is limited by long larval duration. At 22°C sporogeny is completed in less than three weeks and mosquito survival is certainly great enough (15%) for the transmission cycle to be completed. Thus temperature below 18°C was considered unsuitable, and above 22°C, suitable for stable transmission. The upper limit of temperature suitability is determined by vector survival, since sporogeny takes less than a week. Temperatures of above 32°C have been reported to cause high vector population turnover, weak individuals and high mortality (D. leSueur, PhD Thesis, University of Natal, 1991; R. Maharaj, PhD Thesis, University of Natal, 1995), and thermal death occurs at 40°C. In terms of rainfall, 80mm per month was considered suitable, 0mm unsuitable.

The climate data maps were assigned new, fuzzy, values, based on their suitability for sporogonic development and the occurrence of rainfall, which indicate how suitable (on a scale of 0 to 1) the climate is. Since both temperature and rainfall have to be favourable at the same time of the year to allow transmission, the suitability maps were combined to calculate which of the two was more limiting each month. Furthermore, suitable conditions have to continue for a certain time period, long enough for the transmission cycle to be completed. In the hot north African countries the rainfall season only needs to last for three months for mosquito populations to grow enough to permit transmission, while in the rest of Africa the required period was set at five months. Finally, all areas with frost in winter were masked from the distribution model because frost eliminates mosquitoes populations. The details are given in Craig et. al. (1998).

Because the climate data are a long-term mean, one is calculating climatic suitability in the average year, and thus the suitability for stable, or annual, transmission.

What Does the Model Show?

Estimating Populations At Risk

Table 1  Estimates of total and childhood (0-4 years) populations living in areas 50% and 90% suitable for stable malaria, and consequent estimates of annual malaria deaths among African children for 1990 (Snow, et. al., 1998a).

Climatic suitability  (Transmission stability)	1990 total population exposed to infection risk	1990 childhood population exposed to mortality risk	Lower mortality estimate:  4.6 per 1000 p.a.	Median mortality estimate:  8.0 per 1000 p.a.	Upper  mortality estimate:  10.3 per 1000 p.a.
50%	360,243,292	66,338,541	305367	530,708	683,286
90%	293,955,320	54,331,548	249925	434,652	559,615

Seasonality

Both the duration and the start and end of the malaria season are important to malaria control. The duration of the season will affect the dynamics of transmission with longer seasons allowing more intense transmission and higher levels of infection in the human population. Knowing the duration of the transmission season is important in terms of ensuring that suitable control strategies and tools are used (e.g. in an area with 9 months of transmission, impregnation of nets needs to be carried out just prior to the onset of the season and with an insecticide with a residual effect of at least 9 months). As with distribution, the primary factors determining the onset and termination of transmission are climatic.

In the highly seasonal areas, where the annual rainfall and temperature ranges are great, mosquito populations annually drop to levels where transmission cannot be sustained. As rain sets in and/or temperature rises, the vector populations must recover from their annual out-of-season lows to levels where transmission can again take place. This requires a period of highly suitable climate, especially if the season lasts only for a few months. On the other hand, where climate is stable all year, mosquito populations do not go through the same marked annual growth and decline cycles, but persist at more stable levels. A slow steady population turnover is possible even at temperatures close to the lower threshold (around 18°C ), if these temperatures persist all year. It is therefore understandable that summers in highly seasonal areas need to be hotter than those in stable regions to produce mosquito populations capable of transmitting malaria.

Months of Transmission

Malaria At the Periphery

The question of epidemic malaria concerns the occasional malaria outbreaks in normally malaria-free regions or, in its wider sense, a particularly severe season in a normally low-risk area. Factors that cause epidemics range from unusually prolonged or favourable climatic conditions, to man-made alterations to the physical environment that allow mosquito breeding where it was not possible before (irrigation, dam or canal constructions, deforestation) and the movement of people into and from malarious regions.

To explore the impact of inter-annual fluctuations on epidemic occurrence, annual climate surfaces were commissioned from the Climatic Research Unit, Norwich (CRU/SAMRC, 1998). These surfaces can be used to investigate the climate patterns over time. Epidemics will need to be investigated at a sub-continental level, because of major differences between north, equatorial and southern African climate patterns. To start off with, a continental malaria distribution model, based on a five-month transmission season, was created for each year from 1951 to 1995 (the years for which data was available). The total number of years out of 45, in which climatic suitability according to this model was above 10%, were calculated (Figure 14).

The outcome of this model is in some ways similar to the original distribution model (Figure 7). This is not surprising, because if conditions are suitable in the average year, transmission can occur frequently, but if conditions are unsuitable in the average year, transmission would occur very infrequently, if at all. However, much work still has to be done before we have a useful product that reflects fundamental differences across the continent and that has been tested against records of epidemics.

Statistical Models

At this level we are looking at malaria in terms of infection rates and how these relate to climatic and environmental factors at a national / regional scale. This level is still too coarse to consider small-scale factors that cause differences in transmission between or within seemingly homogenous zones. At the regional level we are looking for statistical correlations between empirical data and certain environmental patterns, which can then be used to extrapolate to areas for which no data are available. Spatial statistics will play an increasing role in this aspect, but for the moment the models are purely statistical.

Kenya

In the second step this model (Figure 15) was used to estimate malaria mortality and morbidity among Kenyan children. To this end published and unpublished information about mortality and morbidity at different levels of endemicity was extracted and summarized. National population data from 1080 4th level administrative units was available from the 1989 census, and was projected to 1997. The population data was then combined with the information on mortality and morbidity and the endemicity model, to estimate how many children are affected in areas of high, medium, low or unstable malaria endemicity. In this way it was estimated that daily, 72 and 400 children below the age of five either die or develop clinical malaria respectively (Snow et al., 1998). Despite several limitations, such an approach goes beyond "best guesses" to informed estimates of the geographic burden of malaria and its fatal consequences in Kenya.

Mali

It was interesting to find that the association between malaria prevalence and distance to water was inverse U-shaped. The highest malaria prevalence was experienced by populations at an intermediate distance from water, and not by those closest to water. This was explained by the fact that people living very close to water make use of bed nets, simply because of the tremendous nuisance of mosquitoes in those places.

The statistical model derived from the data was then applied back to the environmental data surfaces to predict the malaria prevalence for the rest of Mali. Finally, the predicted malaria risk was converted into one of 4 categories: high (>70%), medium (30-70%), low (10-30%) and very low risk (<10%),  (Figure 16). The prevalence for seventy of the 101 surveys fell within the predicted bands. Further analysis is being carried out on the spatial structure of the data, to see whether there may be other environmental factors that significantly influence the distribution of malaria risk.

Experience From the Regional Models

Why were different techniques used? Firstly, it illustrates the explorative nature of the work at the moment. Sample sizes and distribution of available data affect the choice of the tests to be used. Further, the fact that there are major ecological differences between different parts of the continent, means that the best statistical approach will need to be selected for each  region. Kenya and Mali are very different: conditions vary considerably within Kenya, where we find two distinct rainy seasons in some regions and one season or even continuous rain in others, both desert and tropical forests, coastline and very high mountains. Mali on the other hand is fairly uniformly hot, with no high mountains, and with one pronounced rainy season. These differences affect whether or not a variable such as “June rainfall” has the same positive or negative effect on PR in all parts of the country, how it interacts with other variables, such as “June temperature” or “May rainfall” and whether or not it can be combined with another variable,  such as “July rainfall”, to reduce the number of variables.

As the data collection nears completion, endemicity models will be created for all malarious countries. We anticipate that different techniques will be applied in different countries or ecological regions, depending on the amount, type and quality of available empirical data, and perhaps also depending on the ecological conditions. During the next stage of the project more effort will be invested into examining and / or developing possibly new statistical and spatial techniques, to deal with the unique problems presented by the MARA/ARMA data set. It is worth noting that in any event the accuracy of the final models will be dependent on and limited by the abundance and accuracy of the data - both malaria data and environmental data.

Chapter 4. Related Projects

Highlands Malaria

Why is Highland Malaria Important?

Epidemic malaria in highland areas represents a significant public health problem. Historically, low risk of infection in highland areas has created little functional immunity in local populations, resulting in relatively high case mortality in adults and children during epidemics.  At the same time, national malaria control programmes have not been well equipped to identify and respond to epidemics. There is, therefore, a need for increased scientific understanding of the epidemiology of highland malaria, as well as greater capacity in epidemic surveillance and response.  The HIMAL project, which was the product of a TDR workshop on highland malaria in Addis Ababa in 1996, is designed to address these issues.

What is HIMAL?

After the current work of HIMAL is complete, the project will focus on collecting prospective malaria data through improved surveillance in highland areas.  In particular a workshop to be held towards the end of 1998, and which will include local scientists and control programme managers, will address the design of sustainable surveillance systems, that include capacity building in GIS and epidemic early warning.

Vector Data Collection

Distributional data are important because the species do not share the same behavioural characteristics, nor are they equal in their ability to transmit malaria. An. funestus and An. gambiae s.s. are largely associated with humans and their habitations, preferring to feed on people and rest inside houses. An. arabiensis, on the other hand, will feed on humans or cattle, rest indoors or outdoors, and is slightly less efficient at transmitting malaria. As a result, different control strategies are needed to combat different species. For instance, indoor spraying with residual insecticides will be effective against An. funestus and An. gambiae but only partially effective against An. arabiensis.

Maps of the distribution of An. funestus and the combined An. gambiae complex were produced by Gillies and de Meillon thirty years ago (Gillies and de Meillon, 1968). This source has been updated for the An. gambiae complex and Figure 18a-d shows the distribution of the member species. The records are extracted mainly from published literature from 1962 onwards. Earlier records, in which identification of the species are based on salinity tolerance, are also included, as well as identifications based on cross-mating experiments. Blank areas do not mean that mosquitoes are absent, but that no species identifications have been published from those areas.

Chapter 5. Conclusions and Outlook

MARA/ARMA Products for Malaria Control Programme Managers

At a regional level such distribution maps could help neighbouring countries to discuss and plan their malaria problem, in a coordinated way, in order to optimize resources. Malaria issues along common borders, as for example between South Africa and Moçambique, are likely to become more numerous with the advent of large scale control, especially if control takes place in one country and not in another. The spread of drug resistance or vector resistance to insecticides, and many other practical issues could also be added to regional maps to give a better insight into the problems faced by managers. Finally, at continental level the mara/arma database, maps, and population-at-risk figures should provide a useful planning tool for international agencies, especially in view of the renewed interest in malaria control.

The present technical report aims to give a general overview of the MARA/ARMA activities to-date to a general audience. Our hope is that this will help programme managers and implementers as well as international agencies to get a clearer idea of the possibilities offered by mapping in general and the approaches developed by the MARA/ARMA collaboration in particular. We welcome any inputs and suggestions regarding the present work as important contributions to the future of this undertaking and we invite you to fill in and send us the questionnaire in this document. We hope that this instrument will prove useful for the needs of the disease control community.

MARA/ARMA Products for Malaria Researchers

The Way Ahead

Where is the MARA/ARMA collaboration going from here? Firstly, the data collection will still be continuing for at least two more years until all avenues have been exhausted. Several countries have not yet been visited and some of the collected data have not been abstracted. No data co-ordinator has been employed so far for southern Africa, and this has hindered the data collection in this region. The database is the major resource in this project and the collection process has to take a high priority until its completion.

Secondly, environmental malaria models for the whole continent will be further developed and refined. This should lead to better overall malaria maps. The model could also be a useful start for predicting the possible extension of malaria as a result of global climate changes. At national level, the statistical modelling process started in Kenya and Mali will be extended to other countries and regions. Statistical methods will be revised, and spatial statistics will need to be incorporated in future. This will increase the number of detailed endemicity maps, as well as maps showing where and which populations are at different levels of risk.

The final product of the MARA/ARMA collaboration will be an atlas of malaria risk for the whole continent, both in a book version and in digital format, that will contain country maps of endemicity, seasonality, as well as available vector distribution maps. It is envisaged that other related data, such as drug resistance or bed net use could also be included, when available. Eventually, the electronic version will be placed on the Internet for general use. The aim is that the digital atlas will allow for constant updating, extracting, querying and refining of malaria risk distribution in Africa.

Finally it is hoped that this collaboration serves as a model for other large-scale disease information systems in Africa and in other developing countries.

References

Clyde, D. F. 1967 Malaria in Tanzania, London, Oxford University Press.

Craig, M. H., R. W. Snow and D. le Sueur.1998. A climate-based distribution model of malaria transmission in Africa. Parasitology Today, in press.

CRU/SAMRC. 1998. African Monthly climate data 1951-1995, Edition 1, CD-ROM. Climatic Research Unit, University of East Anglia, Norwich, NR4 7TJ, United Kingdom.

de Meillon, B. 1934. Observations on Anopheles funestus and Anopheles gambiae in the Transvaal. Publications of the South African Institute for Medical Research 6: 195.

Deichmann, U. 1996. Africa population database, Internet. National Center for Geographic Information and Analysis, United Nations Environment Programme, World Resources Institute, www.grid2.cr.usgs.

DeSavigny, D and P. Wijeyaratne (eds). 1995. GIS for Health and the Environment. Proceedings of an International Workshop, Colombo, Sri Lanka, 5-10 September 1994.

Detinova, T. S. 1962. Determination of the epidemiological importance of populations of Anopheles maculipennis by their age composition. In: Age Grouping Methods in Diptera of Medical Importance, with Special Reference to Some Vectors of Malaria, World Health Organization, Geneva.

FAO. 1991. FAO - Artemis NOAA AVHRR NDVI Image Bank Africa 1981-1991, CD-ROM. Food and Agriculture Organisation (FAO) of the United Nations Remote Sensing Centre, NASA Goddard Space Flight Centre (GSFS), Africa Real Time Environmental Monitoring Information System (ARTEMIS).

GDE Systems Inc. 1995. Geoname Digital Gazetteer Version 1, CD-ROM.

Gill, C. A. 1938 The seasonal periodicity of malaria and the mechanism of the epidemic wave, London, Churchill.

Gillies, M. T. and B. de Meillon. 1968 The Anophelinae of Africa South of the Sahara, Johannesburg, The South African Institute for Medical Research.

Hutchinson, M. F., H. A. Nix, J. P. McMahan and K. D. Ord. 1995. Africa - A topographic and climatic database, Edition 1,CD-ROM. Center for Resource and Environmental Studies, Australian National University, Canberra, ACT 0200, Australia.

Jambulingham, P., S. S. S. Mohapatra, P. Govardhini, L. K. Das, A. Manoharan, S. P. Pani and P. K. Das. 1991. Microlevel epidemiological variations in malaria and its implications on control strategy. Indian Journal of Medical Research [A] 93: 371-378.

Leeson, H. S. 1931 Anopheline Mosquitos in Southern Rhodesia, London, The London School of Hygiene and Tropical Medicine.

Lengeler, C. 1998. Insecticide-treated bednets and curtains for malaria control (Cochrane review). The Cochrane Library, Issue 3, 1998. Oxford: Update Software.

Lysenko, A. Y. and I. N. Semashko. 1968. Geography of malaria. In: Lebedew, A. W. (ed.) . Medical Geography, Academy of Sciences USSR, Moscow.

Macdonald, G. 1957 The Epidemiology and Control of Malaria, London, Oxford University Press.

Martens, W. J. M. 1997. Health Impacts of Climate Change and Ozone Depletion: An Eco-Epidemiological Modelling Approach. Doctoral thesis, Maastricht University.

Metselaar, D. and P. M. Van Theil. 1959. Classification of malaria. Tropical and Geographical Malaria 11: 157-161.

Microsoft. 1998. Encarta World Atlas, CD-ROM. Microsoft Corporation.

Nelson, G. S. 1959. Atlas of Kenya Edition 1. Surveys of Kenya, Crown Printers, Nairobi.

Omumbo, J. A., J. Ouma, B. Rapuoda, M. H. Craig, D. le Sueur and R. W. Snow. 1997. Mapping malaria transmission intensity using geographical information systems (GIS); an example from Kenya. Annals of Tropical Medicine and Parasitology 92: 7-21.

Sharp, B. L. and D. le Sueur. 1996. Malaria in South Africa - the past, the present and selected implications for the future. South African Medical Journal 86: 83-89.

Smith, T., J. D. Charlwood, W. Takken, M. Tanner and D. J. Spiegelhalter. 1995. Mapping the densities of malaria vectors within a single village. Acta Tropica 59: 1-18.

Snow, R. W., M. H. Craig, U. Deichmann and D. le Sueur. 1998. A continental risk map for malaria mortality among African children. Parasitology Today, in press.

Snow, R. W., E. Gouws, J. A. Omumbo, B. Rapuoda, M. H. Craig, F. C. Tanser, D. le Sueur and J. Ouma. 1998. Models to predict the intensity of Plasmodium falciparum transmission: applications to the burden of disease in Kenya. Transactions of the Royal Society of Tropical Medicine and Hygiene, in press.

Snow, R. W., J. A. Omumbo, B. Lowe, C. S. Molyneux, J. O. Obiero, A. Palmer, M. W. Weber, M. Pinder, B. Nahlen, C. Obonyo, C. Newbold, S. Gupta and K. Marsh. 1997. Relation between severe malaria morbidity in children and level of Plasmodium falciparum transmission in Africa. Lancet 349: 1650-1654.

Wilson, D. B. 1956. Atlas of Tanzania Edition 3. Survey Division, Department of Lands and Surveys, Government Printers, Dar es Salaam.

World Health Organization. 1951. Report of the malaria conference in equatorial Africa, Kampala, 1950 (Report). Technical Report Series 38, WHO, Geneva.

World Health Organization. 1994. World malaria situation in 1991. Bulletin of the World Health Organization 72: 160-164.

World Resources Institute. 1995. Africa Data Sampler, Edition 1, CD-ROM.

Zadeh, L. A. 1965. Fuzzy sets. Information and Control 8: 338-353.
 

le Sueur, D., F. Binka, C. Lengeler, D. de Savigny, R. W. Snow, T. Teuscher and Y.T. Touré. 1997. An atlas of malaria in Africa. Africa Health 19: 23-24.

Omumbo, J. A., J. Ouma, B. Rapuoda, M. H. Craig, D. le Sueur and R. W. Snow. 1998. Mapping malaria transmission intensity using geographical information systems (GIS); an example from Kenya. Annals of Tropical Medicine and Parasitology 92: 7-21.

Craig, M. H., R. W. Snow, and D. le Sueur. 1999. A climate-based distribution model of malaria transmission in Africa. Parasitology Today. In press.

Snow, R. W., M. H. Craig, U. Deichmann, and D. le Sueur. 1999. A continental risk map for malaria mortality among African children. Parasitology Today. In press.

Coetzee, M., M. H. Craig, and D. le Sueur. 1999. Mapping the distribution of members of the Anopheles gambiae complex in Africa and adjacent islands. Parasitology Today. Submitted.

Snow, R. W., E. Gouws, J. A. Omumbo, B. Rapuoda, M. H. Craig, F. C. Tanser, D. le Sueur, and J. Ouma. 1998. Models to predict the intensity of Plasmodium falciparum transmission: applications to the burden of disease in Kenya. Transactions of the Royal Society of Tropical Medicine and Hygiene 92: 601-606.

Bagayoko, M., I. Kleinschmidt, N. Sogoba, M. H. Craig, D. le Sueur and Y. T. T. Touré. 1999. Mapping malaria risk in Mali. In preparation.

Kleinschmidt, I., G. P. Y. Clarke, M. Bagayoko, M. H. Craig and D. le Sueur. 1999. A spatial statistical approach to malaria mapping. In preparation.

Appendix 2: Environmental Data Sets

Climate data: A topographic and climate database (CD-ROM)
M.F. Hutchinson, H.A. Nix, J.P. McMahon and K.D. Ord, 1995, Centre for Resource and Environmental Studies, The Australian National University, Canberra, ACT 0200, Australia

Topographic data: African Data Sampler (CD-ROM)
World Resources Institute (WRI), 1995, 1709 New York Ave., NW, Washington, DC 20006, USA

Population data: African population database (public domain)
U. Deichmann (uwe@ncgia.ucsb.edu), 1996, National Centre for Geographic Information and Analysis , University of California, Santa Barbara, CA 93106
http://grid2.cr.usgs.gov/globalpop/africa

Populated places: Geoname Digital Gazetteer v.1 (CD-ROM)
GDE Systems, Inc., 1995, P.O. Box 509009, San Diego, CA 92150-9009
http://www.GDEsystems.com

Elevation data: Global DEM (public domain)
EROS Data Centre (EDC), 1996, U.S. Geological Survey
http://edcwww.cr.usgs.gov/landdaac/gtopo30/gtopo30.html

NDVI: Artemis NOAA AVHRR NDVI Image Bank Africa 1981-1991 (CD-ROM)
Food and Agriculture Organisation (FAO) of the United Nations Remote Sensing Centre; 1991, NASA Goddard Space Flight Centre (GSFS); Africa Real Time Environmental Monitoring Information System (ARTEMIS), NASA Goddard Space Flight Centre, Greenbelt, MD 20771, USA

Global land cover classification (public domain)
R.S. DeFries, M. Hansen, J.R.G. Townshend and R. Sohlberg, 1998, Laboratory for Global Remote Sensing Studies, Department of Geography, 2181 LeFrak Hall, University of Maryland at College Park, College Park, MD 20742-8225, USA
http://www.geog.umd.edu/landcover/8km-map.html

Vector Data
Europe: G. Davidson, R. Page (London School of Hygiene and Tropical Medicine) • C. Ravaonjanahary, P. Carnevale (WHO) • Angola: H. Ribeiro • Benin: A.S. Badawi, G. Pichou, S. Sales • Botswana: R. Abdulla-Khan • Burkina Faso: R. Subra, J. Coz, J. Hamon • Cameroon: B. Colussa, A. Garcia Morilla, L.F. Delfini, P. Capravi, E.P. de Grimaldo, H. Bailly-Choumara • Central African Republic: A.S. Badawi • Chad: G. Morcos • Cote d'Ivoire: J. Coz, J. Hamon, P. Carnevale • Ethiopia: N. Rishikesh, Y. Mekuria, S.C. Luen, R.L. Turner, G.B. White, V. Ariaratuam, B. Feinstein, M. Fettene • Gabon: A.S. Badawi • Gambia: J.H. Bryan, A.W.R. McCrae, W.F. Snow, M.T. Gillies, J.F. Invest • Ghana: L.F. Delfini, G. Kudicke, R. Iyengar • Guinea: A.S. Badawi, G.E. Bakri, V. Pansini, V. Bespiatov • Kenya: J.E. Hudson, A.G. Carmichael, M.H.M. Abulcader, G.P. Joshi, R.R. Fritz, R.B. Highton, J.A. Chandler, J.H. Bryan, S. Digo, A.W.R. McCrae • Liberia: R.S. Bray, G. Pichou, B.K. Mason, Dr le Du, Per Hedman, B. Colussa • Madagascar: G. Chauvet, S. Laventure • Mali: Y. Toure, J. Coz • Mauritania: G. Pichou, J.A. Huddleston • Mauritius: C.M. Courtois, C.C. Draper, F. Gebert, A.R. Gapaul • Mozambique: J. de Sousa, A. du Silvo Carvalhosa, J. Clarke, C.L. Hatch, M. de C. Pereira, G.E. Bakri, Dr le Du, N. Cuamba • Niger: J. Coz • Nigeria: N. Rishikesh, V. Ramakrishna, R. Elliott, G.E. Bakri, R.B.I. Otitoju, S. Dike, M.W. Service, C.D. Ramsdale, C. Cywinski, N.A. Aslam, J.S. Dodge, P. Rosen, A.M. Robertson, G. Shidrawi, J. Clarke, C.V. Foll, F.R.S. Kellet • Reunion: R. van de Vyver, R. Girod • Saudi Arabia: G. de Almeida • Senegal: M. Sarr, G. Morcos, D. Goethals, J. Coz • Sierra Leone: J. Storey, O.J. Beltran • Somalia: A.M. Haridi, A.A.M. Djeloutik • South Africa: G. van Eeden, S.J. Miles, A. Smith, A. Gericke • Swaziland: S.K. Sobti, D.M. Eckard, V. Ramakrishna, P.M. Mathews, J.J.P. la Grange • Sudan: R.L. Turner, S.R. Chowdhury, J. Akiyama, M.A. Akood, A.M. Haridi • Tanzania: A.E.P Mnzava, J.E. Hudson, G.B. White, F.M. Bushrod, A. Smith, F. Mosha, J. Clarke, N. Kolstrup, P. Wegesa, M.T. Gillies, G. Pringle, G.D. Chetty, S.K. Sobti, Y.S. Kim, G.E. Bakri, J.A. Odetoyinbo, G.M. Versi, D. Goethals, J. Storey, C. Shiff, E. Temu • Togo: G. Houel, G.E. Bakri, S. Adrieu, G. Pichou, J. Brengues, S. Sales, R. Iyengar, A. Geller, V. Pansini • Uganda: G. Shidrawi, E. Ouori, G.B. White, A.W.R. McCrae, L.G. Mukwaya • Yemen: R.L. Koumetsou, S.A. Smith • Zambia: J. Hadjinicolaou, H.E. Paterson, A.J. Shelley • Zimbabwe: J. Hadjinicolaou, C.A. Green, R.H. Hunt, H.E. Paterson, J. Govere.

Back to top

MARA/ARMA home