Travel Medicine and Infectious Disease (2014) 12, 726e732
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevierhealth.com/journals/tmid
Comparative benefit of malaria chemoprophylaxis modelled in United Kingdom travellers* Stephen Toovey a,*, Keith Nieforth b, Patrick Smith b, Patricia Schlagenhauf c, Miriam Adamcova d, Iain Tatt d, Danitza Tomianovic d, Gabriel Schnetzler d a
Pegasus Research, Basel, Switzerland d3 Medicine, Parsippany, NJ, USA c University of Zurich Centre for Travel Medicine, WHO Collaborating Centre for Travellers’ Health, Institute for Social and Preventive Medicine, Zurich, Switzerland d F. Hoffmann-La Roche, Basel, Switzerland b
Received 6 March 2014; received in revised form 30 July 2014; accepted 7 August 2014
Available online 28 September 2014
KEYWORDS Mefloquine; Atovaquoneproguanil; Doxycycline; Chloroquine; Prophylaxis
Summary Background: Chemoprophylaxis against falciparum malaria is recommended for travellers from non-endemic countries to malarious destinations, but debate continues on benefit, especially with regard to mefloquine. Quantification of benefit for travellers from the United Kingdom (UK) was modelled to assist clinical and public health decision making. Methods: The model was constructed utilising: World Tourism Organization data showing total number of arrivals from the UK in countries with moderate or high malaria risk; data from a retrospective UK Clinical Practice Research Datalink (CPRD) drug utilisation study; additional information on chemoprophylaxis, case fatality and tolerability were derived from the travel medicine literature. Chemoprophylaxis with the following agents was considered: atovaquone-proguanil (AP), chloroquine with and without proguanil (C P), doxycycline (Dx), mefloquine (Mq). The model was validated for the most recent year with temporally matched datasets for UK travel destinations and imported malaria (2007) against UK Health Protection Agency data on imported malaria. Results: The median (mean) duration of chemoprophylaxis for each agent in weeks (CPRD) was: AP 3.3 (3.5), C P 9 (12.1), Dx 8 (10.3), Mq 9 (12.3): the maximum duration of use of all regimens was 52 weeks. The model correctly predicted falciparum malaria deaths and gave a robust estimate of total cases e model: 5 deaths from 1118 cases; UK Health Protection Agency: 5 deaths from 1153 cases. The number needed to take chemoprophylaxis (NNP) to
* The abstract of this paper was presented as a poster at the 8th European Conference on Tropical Medicine and International Health in Copenhagen, Denmark, September 10e13, 2013. * Corresponding author. E-mail address: [email protected] (S. Toovey).
http://dx.doi.org/10.1016/j.tmaid.2014.08.005 1477-8939/ª 2014 Published by Elsevier Ltd.
Benefit of malaria chemoprophylaxis
727
prevent a case of malaria considered against the ‘background’ reported incidence in non-users of chemoprophylaxis deemed in need of chemoprophylaxis was: C P 272, Dx 269, Mq 260, AP 252; the NNP to prevent a UK traveller malaria death was: C P 62613, Dx 61923, Mq 59973, AP 58059; increasing the ‘background’ rate by 50% yielded NNPs of: C P 176, Dx 175, Mq 171, AP 168. The impact of substituting atovaquone-proguanil for all mefloquine usage resulted in a 2.3% decrease in estimated infections. The number of travellers experiencing moderate adverse events (AE) or those requiring medical attention or drug withdrawal per case prevented is as follows: C P 170, Mq 146, Dx 114, AP 103. Conclusions: The model correctly predicted the number of malaria deaths, providing a robust and reliable estimate of the number of imported malaria cases in the UK, and giving a measure of benefit derived from chemoprophylaxis use against the likely adverse events generated. Overall numbers needed to prevent a malaria infection are comparable among the four options and are sensitive to changes in the background infection rates. Only a limited impact on the number of infections can be expected if Mq is substituted by AP. ª 2014 Published by Elsevier Ltd.
1. Introduction Falciparum malaria is a progressive and potentially lethal disease in patients who do not possess some degree of preexisting immunity to Plasmodium falciparum, the causative parasite; such immunity is usually acquired through having grown up in, and having continued to reside in, a malaria endemic region [1]. Thus, most travellers from developed countries will be at elevated risk of serious and potentially fatal illness should they visit a malaria endemic region. Emigrants from malarious regions settled in non-malarious countries lose protective immunity over time, and hence are at increased risk of clinical malaria should they visit malarious destinations, typically upon making a return visit to their countries of origin. Settled immigrants do however retain some residual semi-immunity and are less likely to die from malaria than non-immune travellers [2]. Antimalarial chemoprophylaxis is accordingly recommended for travellers from malaria free countries who visit malarious regions, to prevent development of acute malaria and its complications, including severe disease and death [3]. All medication is associated with the risk of developing adverse events (AEs), and these risks are quite well characterised for the antimalarial chemoprophylactic agents in current use: mefloquine, atovaquone-proguanil, doxycycline, chloroquine with and without proguanil [4], however, to enable a more complete assessment of the benefit-risk ratio for antimalarial chemoprophylaxis, quantification of benefit would be helpful. To this end we have modelled the benefits of chemoprophylaxis for travellers from the United Kingdom visiting moderate and high-risk malarious destinations. As travellers to low-risk malaria destinations are often recommended stand-by emergency medication rather than chemoprophylaxis, we excluded such destinations from our datasets [5].
2. Materials and methods The model attempts to track the flow of travellers from the United Kingdom to moderate and high-risk malaria
destinations in calendar year 2007, the latest year for which complete data sets for all model variables were available, and to assess the benefit conferred by the use of chemoprophylaxis. The data sources utilised to populate the model are detailed below. The numbers of travellers at travel related risk of malaria exposure were obtained from the United Nations World Tourism Organization (UNWTO) dataset, “Data on Outbound Tourism (2012)” [6]. Destination countries were then cross referenced to country risk category from the US CDC malaria risk tables [7]. For the purposes of this model, countries on the CDC list were reclassified by a malariologist as high, low, or no risk destinations for malaria. In the case of countries such as South Africa, which are mostly malaria free, but which do contain only localised high risk malarious regions, the risk for the whole country was set to ‘no risk’ in order to not overinflate “high risk” exposure. To ascertain the number of UK travellers who sought advice and were assessed by health care professional prior to departure, numbers were obtained from those reported in a survey of departing passengers conducted in 2003 at Heathrow Airport, London, UK, and from the results of a field study of UK travellers [8,9]. The allocation of travellers to each of the four chemoprophylactic drug groups, mefloquine, atovaquoneproguanil, doxycycline, chloroquine and proguanil was determined from the results of a separate study by Blo ¨chliger et al. of prescribing patterns in UK general practice, conducted using the UK Community Practice Research Database (formerly known as the UK General Practice Research Database) [10]. The split between agents, derived from the absolute number of travellers prescribed each chemoprophylactic agent is as follows: mefloquine 15.3%, atovaquone-proguanil 65.6%, doxycycline 14%, and chloroquine with and without proguanil 5.1%. Malaria infection and death rates in UK travellers for the calendar year 2007 were obtained from published UK Health Protection Agency data [11]. The number of reported cases of malaria occurring in UK users of each chemoprophylactic agent was based upon the analysis of
728 Zuckerman et al., which also reported for the calendar year 2007 [12]. Likely AE rates associated with the use of each of the chemoprophylactic agents of interest were obtained from a double-blind, randomised study of Swiss, German and Israeli travellers, thought to be of acceptably similar profile to UK travellers, an equivalent UK study not being available [4]. Mild AEs were not included in the model, only those categorised as either moderate, requiring medical attention, or leading to cessation of usage: given the seriousness of malaria, mild AEs were thought to be generally tolerable and not disruptive of travel. In addition, as the percentages applied were obtained from a clinical study, a setting in which AEs are solicited and which may overestimate the ‘real world’ AE incidence, a further analysis of the AE burden was undertaken. This second analysis considered only those AEs that required medical attention. For both analyses, the AE rate for each regimen, regardless of infection status, was multiplied by the NNP infection, generating the AE burden for each case of malaria prevented, for each regimen, this number being a measure of the benefit-risk ratio for each agent. Overall mortality was taken as published by the UK Health Protection Agency [11], and was applied to users of chemoprophylaxis, and the adjusted total of non-users, to generate the number of fatalities for each of these two groups; these were then summed to provide the total number of malaria deaths. The model was a decision-tree type model, implemented in Microsoft Excel. The model incorporated data-derived event probabilities, including the number of individuals receiving prophylaxis, infection rates with and without prophylaxis, infection-related morbidity and mortality, and drug-related adverse event rates. The model was populated with the actual measured or reported values from the abovementioned sources, to generate a base case scenario, with the following inputs entered into the model: the total number of UK travellers to risk destinations was set at 1,768,210 [6], with the fraction of this total using chemoprophylaxis set at 0.78 [9]; the fraction of chemoprophylaxis users taking each of the four regimens was: mefloquine 0.153, atovaquone-proguanil 0.656, doxycycline 0.14, chloroquine with and without proguanil 0.051 [10]; the fraction of travellers not using malaria chemoprophylaxis because they had been advised against usage was set at 0.37 [8]. The infection rate amongst users of each of the four chemoprophylactic regimens was calculated from the known number of cases in the UK and published utilisation data, applying the usage ratios derived from the Blo ¨chliger drug utilisation study [10e12]. The infection rate amongst non-users of chemoprophylaxis was adjusted for those deemed not at actual risk of becoming infected [8]. Infections per group were summed to provide the total number of cases. Infection rates per person-week of exposure were calculated for each chemoprophylaxis regimen, being the quotient of the total number of infections calculated for that regimen, and the product of the total number of regimen users and mean duration of usage. For each of the four agents, the number of travellers that would need to be prescribed the agent to prevent one death from malaria, and the number needed to prevent one
S. Toovey et al. clinical malaria infection were calculated, being the reciprocal of the difference in rates with adjusted nonusers of chemoprophylaxis, expressed as NNP death and NNP infection respectively. Although the adopted approach is reasonably ‘grounded in reality’ inasmuch as the model was based on real world data, there are nevertheless likely to be differences between the four chemoprophylaxis groups with respect to their destinations, as well as other unmeasured variables, including variations in background malaria intensity, with malaria transmission rates not being geographically uniform within countries. Additionally, the proportion of travellers not using chemoprophylaxis subsequent to taking professional advice could vary between groups and destinations. We therefore also examined the impact of varying rates of malaria infection on the NNP for each agent, by generating a range of values that may be more reflective of the risk spectrum within destination countries: a range of values from 80% to 200% of the actual estimated infection rate in non-users of chemoprophylaxis deemed at risk of infection was utilised. A further sensitivity analysis was undertaken to examine what the impact of increasing and decreasing mefloquine’s share of total chemoprophylactic prescriptions might have on the total predicted number of malaria infections and deaths. We additionally investigated the impact of a higher malaria death rate in infected travellers, examining what the impact on mortality would be if the UK rate were increased to the 3% mortality reported by Germany for nonimmune tourists for the period 1993e2004 [13], as well as a higher rate of 4% that might reflect likely level of care in some lesser experienced centres.
3. Results The model correctly predicted the total number of malaria deaths in the UK in 2007, yielding a result of five deaths, which matched exactly the official HPA statistic. The model predicted the total number of cases to be 1118, which is a very close estimate (97%) of the 1153 actual malaria infections recorded by the HPA in 2007. The NNP infection and NNP death for each of the four agents, based upon actual drug prescription patterns within the UK, are presented in Table 1; the impact of changes in the malaria transmission, or ’background rate’ on NNP infection and NNP death are similarly included in Table 1. As would be expected, both of these numbers fall as the background rate increases, and vice versa. Altering the prescription pattern within the model, by swapping mefloquine and atovaquone-proguanil usages, had no impact on the number of predicted deaths but did yield a 9.8% increase in the number of infections, from 1118 to 1228. The impact of substituting atovaquone-proguanil for all mefloquine usage, i.e. setting mefloquine usage to zero and atovaquone-proguanil to 80.9% of total usage, was similar, with predicted deaths remaining at 5, and predicted infections decreasing slightly (2.3%) from 1118 to 1092. These results are summarised in Table 2. Increasing the mortality rate from the actual observed rate in the UK (
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevierhealth.com/journals/tmid
Comparative benefit of malaria chemoprophylaxis modelled in United Kingdom travellers* Stephen Toovey a,*, Keith Nieforth b, Patrick Smith b, Patricia Schlagenhauf c, Miriam Adamcova d, Iain Tatt d, Danitza Tomianovic d, Gabriel Schnetzler d a
Pegasus Research, Basel, Switzerland d3 Medicine, Parsippany, NJ, USA c University of Zurich Centre for Travel Medicine, WHO Collaborating Centre for Travellers’ Health, Institute for Social and Preventive Medicine, Zurich, Switzerland d F. Hoffmann-La Roche, Basel, Switzerland b
Received 6 March 2014; received in revised form 30 July 2014; accepted 7 August 2014
Available online 28 September 2014
KEYWORDS Mefloquine; Atovaquoneproguanil; Doxycycline; Chloroquine; Prophylaxis
Summary Background: Chemoprophylaxis against falciparum malaria is recommended for travellers from non-endemic countries to malarious destinations, but debate continues on benefit, especially with regard to mefloquine. Quantification of benefit for travellers from the United Kingdom (UK) was modelled to assist clinical and public health decision making. Methods: The model was constructed utilising: World Tourism Organization data showing total number of arrivals from the UK in countries with moderate or high malaria risk; data from a retrospective UK Clinical Practice Research Datalink (CPRD) drug utilisation study; additional information on chemoprophylaxis, case fatality and tolerability were derived from the travel medicine literature. Chemoprophylaxis with the following agents was considered: atovaquone-proguanil (AP), chloroquine with and without proguanil (C P), doxycycline (Dx), mefloquine (Mq). The model was validated for the most recent year with temporally matched datasets for UK travel destinations and imported malaria (2007) against UK Health Protection Agency data on imported malaria. Results: The median (mean) duration of chemoprophylaxis for each agent in weeks (CPRD) was: AP 3.3 (3.5), C P 9 (12.1), Dx 8 (10.3), Mq 9 (12.3): the maximum duration of use of all regimens was 52 weeks. The model correctly predicted falciparum malaria deaths and gave a robust estimate of total cases e model: 5 deaths from 1118 cases; UK Health Protection Agency: 5 deaths from 1153 cases. The number needed to take chemoprophylaxis (NNP) to
* The abstract of this paper was presented as a poster at the 8th European Conference on Tropical Medicine and International Health in Copenhagen, Denmark, September 10e13, 2013. * Corresponding author. E-mail address: [email protected] (S. Toovey).
http://dx.doi.org/10.1016/j.tmaid.2014.08.005 1477-8939/ª 2014 Published by Elsevier Ltd.
Benefit of malaria chemoprophylaxis
727
prevent a case of malaria considered against the ‘background’ reported incidence in non-users of chemoprophylaxis deemed in need of chemoprophylaxis was: C P 272, Dx 269, Mq 260, AP 252; the NNP to prevent a UK traveller malaria death was: C P 62613, Dx 61923, Mq 59973, AP 58059; increasing the ‘background’ rate by 50% yielded NNPs of: C P 176, Dx 175, Mq 171, AP 168. The impact of substituting atovaquone-proguanil for all mefloquine usage resulted in a 2.3% decrease in estimated infections. The number of travellers experiencing moderate adverse events (AE) or those requiring medical attention or drug withdrawal per case prevented is as follows: C P 170, Mq 146, Dx 114, AP 103. Conclusions: The model correctly predicted the number of malaria deaths, providing a robust and reliable estimate of the number of imported malaria cases in the UK, and giving a measure of benefit derived from chemoprophylaxis use against the likely adverse events generated. Overall numbers needed to prevent a malaria infection are comparable among the four options and are sensitive to changes in the background infection rates. Only a limited impact on the number of infections can be expected if Mq is substituted by AP. ª 2014 Published by Elsevier Ltd.
1. Introduction Falciparum malaria is a progressive and potentially lethal disease in patients who do not possess some degree of preexisting immunity to Plasmodium falciparum, the causative parasite; such immunity is usually acquired through having grown up in, and having continued to reside in, a malaria endemic region [1]. Thus, most travellers from developed countries will be at elevated risk of serious and potentially fatal illness should they visit a malaria endemic region. Emigrants from malarious regions settled in non-malarious countries lose protective immunity over time, and hence are at increased risk of clinical malaria should they visit malarious destinations, typically upon making a return visit to their countries of origin. Settled immigrants do however retain some residual semi-immunity and are less likely to die from malaria than non-immune travellers [2]. Antimalarial chemoprophylaxis is accordingly recommended for travellers from malaria free countries who visit malarious regions, to prevent development of acute malaria and its complications, including severe disease and death [3]. All medication is associated with the risk of developing adverse events (AEs), and these risks are quite well characterised for the antimalarial chemoprophylactic agents in current use: mefloquine, atovaquone-proguanil, doxycycline, chloroquine with and without proguanil [4], however, to enable a more complete assessment of the benefit-risk ratio for antimalarial chemoprophylaxis, quantification of benefit would be helpful. To this end we have modelled the benefits of chemoprophylaxis for travellers from the United Kingdom visiting moderate and high-risk malarious destinations. As travellers to low-risk malaria destinations are often recommended stand-by emergency medication rather than chemoprophylaxis, we excluded such destinations from our datasets [5].
2. Materials and methods The model attempts to track the flow of travellers from the United Kingdom to moderate and high-risk malaria
destinations in calendar year 2007, the latest year for which complete data sets for all model variables were available, and to assess the benefit conferred by the use of chemoprophylaxis. The data sources utilised to populate the model are detailed below. The numbers of travellers at travel related risk of malaria exposure were obtained from the United Nations World Tourism Organization (UNWTO) dataset, “Data on Outbound Tourism (2012)” [6]. Destination countries were then cross referenced to country risk category from the US CDC malaria risk tables [7]. For the purposes of this model, countries on the CDC list were reclassified by a malariologist as high, low, or no risk destinations for malaria. In the case of countries such as South Africa, which are mostly malaria free, but which do contain only localised high risk malarious regions, the risk for the whole country was set to ‘no risk’ in order to not overinflate “high risk” exposure. To ascertain the number of UK travellers who sought advice and were assessed by health care professional prior to departure, numbers were obtained from those reported in a survey of departing passengers conducted in 2003 at Heathrow Airport, London, UK, and from the results of a field study of UK travellers [8,9]. The allocation of travellers to each of the four chemoprophylactic drug groups, mefloquine, atovaquoneproguanil, doxycycline, chloroquine and proguanil was determined from the results of a separate study by Blo ¨chliger et al. of prescribing patterns in UK general practice, conducted using the UK Community Practice Research Database (formerly known as the UK General Practice Research Database) [10]. The split between agents, derived from the absolute number of travellers prescribed each chemoprophylactic agent is as follows: mefloquine 15.3%, atovaquone-proguanil 65.6%, doxycycline 14%, and chloroquine with and without proguanil 5.1%. Malaria infection and death rates in UK travellers for the calendar year 2007 were obtained from published UK Health Protection Agency data [11]. The number of reported cases of malaria occurring in UK users of each chemoprophylactic agent was based upon the analysis of
728 Zuckerman et al., which also reported for the calendar year 2007 [12]. Likely AE rates associated with the use of each of the chemoprophylactic agents of interest were obtained from a double-blind, randomised study of Swiss, German and Israeli travellers, thought to be of acceptably similar profile to UK travellers, an equivalent UK study not being available [4]. Mild AEs were not included in the model, only those categorised as either moderate, requiring medical attention, or leading to cessation of usage: given the seriousness of malaria, mild AEs were thought to be generally tolerable and not disruptive of travel. In addition, as the percentages applied were obtained from a clinical study, a setting in which AEs are solicited and which may overestimate the ‘real world’ AE incidence, a further analysis of the AE burden was undertaken. This second analysis considered only those AEs that required medical attention. For both analyses, the AE rate for each regimen, regardless of infection status, was multiplied by the NNP infection, generating the AE burden for each case of malaria prevented, for each regimen, this number being a measure of the benefit-risk ratio for each agent. Overall mortality was taken as published by the UK Health Protection Agency [11], and was applied to users of chemoprophylaxis, and the adjusted total of non-users, to generate the number of fatalities for each of these two groups; these were then summed to provide the total number of malaria deaths. The model was a decision-tree type model, implemented in Microsoft Excel. The model incorporated data-derived event probabilities, including the number of individuals receiving prophylaxis, infection rates with and without prophylaxis, infection-related morbidity and mortality, and drug-related adverse event rates. The model was populated with the actual measured or reported values from the abovementioned sources, to generate a base case scenario, with the following inputs entered into the model: the total number of UK travellers to risk destinations was set at 1,768,210 [6], with the fraction of this total using chemoprophylaxis set at 0.78 [9]; the fraction of chemoprophylaxis users taking each of the four regimens was: mefloquine 0.153, atovaquone-proguanil 0.656, doxycycline 0.14, chloroquine with and without proguanil 0.051 [10]; the fraction of travellers not using malaria chemoprophylaxis because they had been advised against usage was set at 0.37 [8]. The infection rate amongst users of each of the four chemoprophylactic regimens was calculated from the known number of cases in the UK and published utilisation data, applying the usage ratios derived from the Blo ¨chliger drug utilisation study [10e12]. The infection rate amongst non-users of chemoprophylaxis was adjusted for those deemed not at actual risk of becoming infected [8]. Infections per group were summed to provide the total number of cases. Infection rates per person-week of exposure were calculated for each chemoprophylaxis regimen, being the quotient of the total number of infections calculated for that regimen, and the product of the total number of regimen users and mean duration of usage. For each of the four agents, the number of travellers that would need to be prescribed the agent to prevent one death from malaria, and the number needed to prevent one
S. Toovey et al. clinical malaria infection were calculated, being the reciprocal of the difference in rates with adjusted nonusers of chemoprophylaxis, expressed as NNP death and NNP infection respectively. Although the adopted approach is reasonably ‘grounded in reality’ inasmuch as the model was based on real world data, there are nevertheless likely to be differences between the four chemoprophylaxis groups with respect to their destinations, as well as other unmeasured variables, including variations in background malaria intensity, with malaria transmission rates not being geographically uniform within countries. Additionally, the proportion of travellers not using chemoprophylaxis subsequent to taking professional advice could vary between groups and destinations. We therefore also examined the impact of varying rates of malaria infection on the NNP for each agent, by generating a range of values that may be more reflective of the risk spectrum within destination countries: a range of values from 80% to 200% of the actual estimated infection rate in non-users of chemoprophylaxis deemed at risk of infection was utilised. A further sensitivity analysis was undertaken to examine what the impact of increasing and decreasing mefloquine’s share of total chemoprophylactic prescriptions might have on the total predicted number of malaria infections and deaths. We additionally investigated the impact of a higher malaria death rate in infected travellers, examining what the impact on mortality would be if the UK rate were increased to the 3% mortality reported by Germany for nonimmune tourists for the period 1993e2004 [13], as well as a higher rate of 4% that might reflect likely level of care in some lesser experienced centres.
3. Results The model correctly predicted the total number of malaria deaths in the UK in 2007, yielding a result of five deaths, which matched exactly the official HPA statistic. The model predicted the total number of cases to be 1118, which is a very close estimate (97%) of the 1153 actual malaria infections recorded by the HPA in 2007. The NNP infection and NNP death for each of the four agents, based upon actual drug prescription patterns within the UK, are presented in Table 1; the impact of changes in the malaria transmission, or ’background rate’ on NNP infection and NNP death are similarly included in Table 1. As would be expected, both of these numbers fall as the background rate increases, and vice versa. Altering the prescription pattern within the model, by swapping mefloquine and atovaquone-proguanil usages, had no impact on the number of predicted deaths but did yield a 9.8% increase in the number of infections, from 1118 to 1228. The impact of substituting atovaquone-proguanil for all mefloquine usage, i.e. setting mefloquine usage to zero and atovaquone-proguanil to 80.9% of total usage, was similar, with predicted deaths remaining at 5, and predicted infections decreasing slightly (2.3%) from 1118 to 1092. These results are summarised in Table 2. Increasing the mortality rate from the actual observed rate in the UK (
