Global climate change and malaria

Global climate change and malaria

Reflection and Reaction 4 5 Blower SM, McLean AR. AIDS: modeling epidemic control. Science 1995; 267: 1252–53. Blower SM, Koelle K, Mills J. Health p...

292KB Sizes 0 Downloads 30 Views

Reflection and Reaction

4 5

Blower SM, McLean AR. AIDS: modeling epidemic control. Science 1995; 267: 1252–53. Blower SM, Koelle K, Mills J. Health policy modeling: epidemic control, HIV vaccines and risky behavior. In: Kaplan EH, Brookmeyer R, eds. Quantitative evaluation of HIV prevention programs. New Haven: Yale University Press, 2002: 260–89.

6

7

Blower SM, Schwartz EJ, Mills J. Forecasting the future of HIV epidemics: the impact of antiretroviral therapies and imperfect vaccines. AIDS Rev 2003; 5: 113–25. Blower SM, Moss RB, Fernandez-Cruz E. Calculating the potential epidemic-level impact of therapeutic vaccination on the San Francisco HIV epidemic. AIDScience 2003; 3(21).

Global climate change and malaria We wish to respond to a number of statements made by Paul Reiter and colleagues1 on our article on malaria and global warming,2 in which we model duration of exposure to Plasmodium falciparum malaria and independently validate the model using 3791 presence/absence parasite surveys3 collected across Africa. Although we recognise that an important component of science is open debate, Reiter and colleagues made some inaccurate statements and hence misrepresentations of our work that need to be addressed. For example, Reiter and colleagues comment that we have modelled “merely duration of the transmission season”, which we interpret as “heightened transmission and increased incidence”.1 The first part of this statement is absolutely correct and modelling duration (and timing) of transmission season (for the first time at a continental scale) is exactly what we set out to achieve. The latter part of the statement regarding increased incidence is inaccurate as nowhere in the paper is incidence interpreted in the light of changes in personmonths of exposure under global climate-change scenarios. The relation between population exposure and disease incidence is not straightforward and there are many contributing factors. To make such inference from our model—which is concerned with spatiotemporal population exposure—would not be valid. Reiter and colleagues state that the model was based on a “mere 15 African locations”.1 This is incorrect. The relation between climate and malaria is complex and for this reason, four phases of model development were used to derive the final model. As discussed in the paper, the initial 15 studies were used to provide crude (first pass) climatic thresholds (within established biological ranges), which were subsequently refined by comparing various iterations of the model against historical published and unpublished maps and clinical case data. The final phase of model development involved extensive consultation and dialogue with experts from throughout Africa regarding areas of agreement, false negatives, false positives, and season duration. All four phases of this iterative develop256

ment process contributed uniquely to the final model. A comparison between the model and historic maps and clinical case data used in the model building process is shown for southern Africa as an example (figure). None of the 3791 presence/absence parasite surveys were used in any way in the model development process, but were withheld to facilitate a true independent accuracy assessment of the final model after development. Similarly, Reiter and colleagues then cite as a “greater failing” of the model our reliance on parasite ratio studies as the relations between parasite prevalence, clinical disease, and transmission season length are unlikely to be linear. Given that “parasite prevalence” was not used at all in the paper and that we make no attempt to infer prevalence from the predicted duration of transmission season, this constitutes another inaccurate statement by Reiter and colleagues. To reiterate, we used only presence/absence data from parasite surveys to independently validate the final model (after the development phase was fully complete) and do not use parasite prevalence in any way. That the relations mentioned above are not linear is undoubtedly true and is the subject of ongoing research in our programme and many others. Reiter and colleagues question the use of contemporary population estimates on the grounds that populations are projected to increase and a greater proportion of people are projected to be living in urban areas over the coming century.1 To include population projections in already uncertain climate projection scenarios would make it difficult to disaggregate the effects of climate from those of population and would result in an “accumulation of uncertainties”.4 Yet this question illustrates exactly the utility of such a model. We chose to model specific climate and population scenarios, but the transparency and reproducibility of the model means that it can be used as a baseline against which to evaluate changes in exposure under any combination of climate and/or population scenarios. http://infection.thelancet.com Vol 5 May 2005

Reflection and Reaction

A

B

.. Harare Harare Zimbabwe

Tsumeb Maun

.. Windhoek Windhoek

Maun

.. Bulawayo

Namibia

Limpopo river

Botswana

.. Bulawayo

Namibia N N AA M M II BB II AA .. Windhoek Windhoek

.. Johannesburg Johannesburg

.. Johannesburg Johannesburg

.. Durban Durban n a ce nO

.. Durban Durban

SSouth S OO UU TTAfrica HH AA FF RR II CC AA

an

S OO UU TTAfrica SSouth HH AA FF RR II CC AA

Oc e

Risk Risk Low Risk isk

Limpopo river

BBotswana B OO TT SS W W AA NN AA

.. Gaborone Gaborone

.. Gaborone Gaborone

rnational undaries ds rs

.. Harare Harare Zimbabwe

.. Tsumeb

ian

dia

Cape Town

0

200

Ind

In

400

Cape Town

Kilometres Namibia (malaria risk, 1995) Low risk High risk

Botswana (annual malaria cases)

South Africa (malaria risk, 1938)

50–500 500

Moderate risk High risk Continuous

0

200

400

Kilometres

Zimbabwe (expert opinion) Seasonal Perennial potential

No transmission 3 months transmission 3–6 months transmission 6 months transmission

Figure: Comparison between the historic maps and case data (A) and seasonality model (B) for southern Africa

Reiter and colleagues intimate that the spatiotemporal sensitivity of 63% is too “low” (this figure comprising a spatial sensitivity of 90% and a temporal sensitivity of 70%). The nature of the current validation process means that a long-term mean model is being used to predict the outcome of short-term, small-scale surveys based on fluctuating climate and thus constitutes a very harsh test of the model because of disparate temporal and spatial scales between model and surveys. That the model was able to spatially predict malaria occurrence (based on the surveys) with a sensitivity of 90% in our view is good. In addition, one would expect the true temporal sensitivity to be further lowered by uncertainty regarding rates of parasite clearance. For the reasons above, and from a control planning perspective, a month’s tolerance was deemed to be adequate. A less harsh and fairer test of the accuracy of the model would be possible if accurate annual climatic data existed for each survey for the year in which the survey was undertaken. The high spatiotemporal specificity of 96% within a month’s tolerance (89% with no tolerance) means that the model is powerful in predicting absence of malaria. Furthermore, if the sensitivity and specificity were weighted by http://infection.thelancet.com Vol 5 May 2005

expected presence/absence of malaria in Africa (eg, based on distributional maps), the overall spatiotemporal agreement would be considerably higher. Reiter and colleagues deem person-months of exposure to be an unsuitable outcome measure as “an extension from 1 to 4 months will have more impact than from 10 to 12 months”. Yet this is exactly the point; proportional increases in person-months of exposure are likely to be larger in areas with shorter transmission seasons. We fail to see why this should be problematic. In addition, the person-months of exposure figures should be read in conjunction with overall population exposure and extent to build an overall exposure profile for the country. Malaria is a complex adaptive system reliant on a number of social and environmental factors and there are likely to be many unforeseen and unusual biological feedback mechanisms under various scenarios of global climate change. Models require simplifying assumptions to be made but nevertheless can generate useful insights into possible scenarios. The famous pioneering work of MacDonald5 used a model to show that control measures directed against the malaria vector would pay greater dividends than those directed at the larvae. This 257

Reflection and Reaction

resulted in a change in emphasis in control measures.6 Although we have no doubts that the model will be improved on, the research constitutes the first validated continental-scale transmission model to evaluate both the spatial and temporal aspects of population exposure. Climate change science has a history of political influence, but it is important that an objective standpoint be maintained. The greatest value of our work is not the future malaria projections (around which there is considerable uncertainty), but rather the derivation of a transparent and reproducible baseline model.

FT and BS are at the Malaria Research Lead Programme, Medical Research Council, Durban, South Africa; FT is also at the Africa Centre for Health and Population Studies, Mtubatuba, South Africa. Correspondence to: Dr Frank Tanser, Africa Centre for Health and Population Studies, PO Box 198, Mtubatuba, 3935, South Africa. Tel +27 35 550 7555; [email protected] 1 2 3

4

5

Frank Tanser, Brian Sharp

6

Reiter P, Thomas C, Atkinson P, et al. Global warming and malaria: a call for accuracy. Lancet Infect Dis 2004; 4: 323–24. Tanser FC, Sharp B, Le Sueur D. Potential effect of climate change on malaria transmission in Africa. Lancet 2003; 362: 1792–98. Mapping Malaria Risk in Africa (MARA). Towards an atlas of malaria risk in Africa: first technical report of the MARA/ARMA collaboration. Durban: MARA/ARMA, 1998. McMichael A, Martins P, Kovats R, Lele S. Climate change and human health. In: Elliot P, Wakefield J, Best N, Briggs D, eds. Spatial epidemiology: methods and applications. New York: Oxford University Press, 2000: 444–61. MacDonald G. The epidemiology and control of malaria. London: Oxford University Press, 1957. Harrison G. Mosquitos, malaria and man. New York: Clarke, Irwin and Co, 1978.

Global climate change and malaria In their response to our commentary,1 Paul Reiter and colleagues2 have focused on the uncertainties of predicting malaria epidemics at subnational scale. We acknowledge that Reiter and co-authors have a wealth of knowledge and first-hand experience in the field of malaria epidemiology, with a firm, if somewhat reductionist, theoretical foundation. By contrast, we focus on a different question: the likely long-term influence of global climate change on malaria transmission potential at the broad geographic scale. Here, the “classic components of science—unbiased observation and systematic experimentation”2 are not practical options, since anthropogenic climate change is a one-off event, mostly still located in the future. Nor can we afford to postpone policy decisions until the likely outcomes are clearer, since to do so risks serious and potentially irreversible effects.3 This situation calls for a response in the form of another “classic component” of science: methodological innovation. There is a pressing need to extend existing scientific methods if the potentially severe health impacts of complex, long-term issues such as global climate change are to be anticipated and averted. Indeed, several of Reiter’s co-authors have made important contributions in this field.4–6 There is room for scientific debate about the relative importance of environmental and social factors underlying the spatial distribution of malaria, and the year-toyear stability, or instability, of malaria incidence. However, it is not necessary to be a malariologist to appreciate that 258

climate is a fundamental influence on the distribution of malaria transmission at large spatial and temporal scales. Recent evidence reinforces this conclusion. The historic northern limit of malaria transmission corresponds approximately to the northern summer (July) 15°C isotherm, suggesting that local transmission is unlikely once temperatures fall below this figure.7 The current limits of transmission are almost entirely within the latitudes of 30°N and 30°S, corresponding approximately to the 15°C northern winter (January) isotherm and the 15°C southern winter (July) isotherm. A recent study concluded that “biologic characteristics of diverse vector mosquitoes interact with climate to explain much of the regional variation in the intensity of transmission”.5 Existing models of the potential effects of climate change on vector-borne diseases require simplifying assumptions to be made. This requirement makes their forecasts uncertain8 but it does not mean they are necessarily misleading. Future models of the potential effects of global climate change on vector-borne disease should build on recent findings, and attempt to account more fully for the important effects of non-climate factors. In the short term, social and demographic factors will have more important effects on malaria than climate change.6,7 Yet today’s decisions affecting the world’s climate will also have major implications for human health later this century.9 Simon Hales, Alistair Woodward http://infection.thelancet.com Vol 5 May 2005