- Top of page
- Patients and methods
Objectives To assess if the clinical outcome of patients treated after performing a Rapid Diagnostic Test for malaria (RDT) is at least equivalent to that of controls (treated presumptively without test) and to determine the impact of the introduction of a malaria RDT on clinical decisions.
Methods Randomized, multi-centre, open clinical trial in two arms in 2006 at the end of the dry and of the rainy season in 10 peripheral health centres in Burkina Faso: one arm with use of RDT before treatment decision, one arm managed clinically. Primary endpoint: persistence of fever at day 4. Secondary endpoints: frequency of malaria treatment and of antibiotic treatment.
Results A total of 852 febrile patients were recruited in the dry season and 1317 febrile patients in the rainy season, and randomized either to be submitted to RDT (P_RTD) or to be managed presumptively (P_CLIN). In both seasons, no significant difference was found between the two randomized groups in the frequency of antimalarial treatment, nor of antibiotic prescription. In the dry season, 80.8% and 79.8% of patients with a negative RDT were nevertheless diagnosed and treated for malaria, and so were 85.0% and 82.6% negative patients in the rainy season. In the rainy season only, both diagnosis and treatment of other conditions were significantly less frequent in RDT positive vs. negative patients (48.3%vs. 61.4% and 46.2%vs. 59.9%, P = 0.00 and 0.00, respectively).
Conclusion Our study was inconclusive on RDT safety (clinical outcome in the two randomized groups), because of an exceedingly and unexpectedly low compliance with the negative test result. Further research is needed on best strategies to promote adherence and on the safety of a test based strategy compared with the current, presumptive treatment strategy.
Objectifs: Evaluer si les résultats cliniques des patients traités après l’exécution d’un test de diagnostic rapide de la malaria (TDR) sont au moins équivalents à ceux des contrôles (avec traitement présomptif, i.e. sans test) et déterminer l’impact de l’introduction d’un TDR de la malaria sur les décisions cliniques.
Méthodes: Essai clinique randomisé, multi centrique, ouvert (RCT) à deux bras en 2006 à la fin de la saison sèche et de la saison des pluies dans 10 centres de santé périphériques au Burkina-Faso: un bras avec l’utilisation de TDR avant la décision du traitement et un bras géré cliniquement. Critère d’évaluation primaire: persistance de la fièvre au jour 4. Paramètres secondaires: la fréquence du traitement de la malaria et du traitement antibiotique.
Résultats: 852 patients ont été recrutés durant la saison sèche et 1317 patients fébriles durant la saison des pluies, et randomisés pour être soumis soit au TDR (P_RTD) ou être gérés par présomption (P_CLIN). Pour les deux saisons, aucune différence significative n’a été trouvée entre les deux groupes randomisés ni dans la fréquence du traitement antimalarique, ni dans la prescription d’antibiotiques. Durant la saison sèche, 80,8% et 79,8% des patients avec un TDR négatif ont néanmoins été diagnostiqués et traités pour la malaria, comparés à 85,0% et 82,6% de patients négatifs durant la saison des pluies. Dans la saison des pluies seule, autant le diagnostic et le traitement d’autres conditions étaient significativement moins fréquents chez les patients TDR positifs comparé aux patients négatifs (48,3% vs 61,4% et 46,2% vs 59,9%, p = 0,00 et p = 0,00, respectivement).
Conclusion: Notre étude n’a pas été concluante pour ce qui est de la sécurité du TDR (les résultats cliniques dans les deux groupes randomisés), en raison d’un très faible respect du résultat d’un test négatif, ce qui était inattendu. Des recherches supplémentaires sont nécessaires sur les meilleures stratégies pour promouvoir l’adhésion et sur la sécurité d’une stratégie basée sur un test comparée à la stratégie actuelle de traitement présomptif.
Objetivos: Evaluar si el resultado clínico de pacientes tratados después de habérseles realizado una Prueba Diagnóstica Rápida para malaria (PDR) es equivalente al de los controles (tratados presuntivamente sin habérseles realizado una prueba) y determinar el impacto de la introducción de una PDR de malaria sobre las decisiones clínicas.
Métodos: Ensayo clínico aleatorizado, multicéntrico y abierto, con dos brazos, realizado en el 2006 al final de la estación seca y durante la estación lluviosa, en 10 centros sanitarios periféricos en Burkina Faso. En un brazo se utilizó la PDR antes de tomar la decisión de tratamiento, y en otra se realizó un manejo puramente clínico. El resultado primario: persistencia de fiebre después del Día 4. Resultados secundarios: frecuencia del tratamiento de malaria y de tratamiento con antibióticos.
Resultados: Se reclutaron 852 pacientes en la estación seca y 1317 pacientes febriles durante la estación lluviosa, y se aleatorizaron para ser sometidos a PDR ((P_PDR) o para ser tratados de forma presuntiva (P_CLIN). En ambas estaciones, no se halló diferencia significativa entre los dos grupos aleatorizados en cuanto a la frecuencia de tratamiento antimalárico, ni en cuanto a la prescripción antibiótica. Durante la estación seca, 80.8% y 79.8% de los pacientes con un PDR negativo fueron sin embargo diagnosticados y tratados para malaria, y también lo fueron un 85.0% y 82.6% de pacientes con resultados negativos en la estación lluviosa. Solo en la estación lluviosa, el diagnóstico y tratamiento para otras condiciones fueron significativamente menos frecuentes entre pacientes PDR negativos versus PDR positivos (48.3% vs. 61.4% y 46.2% vs. 59.9%, p = 0.00 and p = 0.00, respectivamente).
Conclusión: Nuestro estudio era inconcluso con respecto a la seguridad de las PDR (resultado clínico en ambos grupos aleatorizados), debido a un bajo cumplimiento, excesivo e inesperado, con los resultados negativos de la prueba. Se requieren más investigaciones sobre las mejores estrategias para promover la adherencia y sobre la seguridad de una estrategia basada en una prueba comparada con la estrategia actual, basada en un tratamiento presuntivo.
- Top of page
- Patients and methods
In recent years, following WHO recommendations, most African countries have adopted treatment protocols for malaria based on artemisinin combination treatments (ACT) (Ogbonna & Uneke 2008). The new protocols have proven to be very effective, but they are also much more expensive than previous regimens. The presumptive treatment of all fevers for malaria, previously a current practice, has therefore being questioned on economical grounds (Pfeiffer et al. 2008). Moreover, presumptive treatment is considered potentially dangerous as it might contribute to selecting for resistant Plasmodium falciparum strains. New guidelines for malaria management recommend a mandatory laboratory test before malaria treatment (WHO 2006). In many African areas without laboratory facilities, the only possibility is the use of a rapid diagnostic test (RDT).
In Burkina Faso the adoption of the new ACT based strategy is very recent, and in 2006 had yet to be implemented. In a context of a highly variable, seasonal transmission, the safety and utility of the introduction of a RDT for malaria was discussed. We decided therefore to study the safety, utility and cost-effectiveness of a RDT based strategy vs. the current, presumptive clinical management in a region where malaria incidence is highest, and where no laboratory facility is available at periphery. We also wanted to assess how the prescribing behaviour of health personnel (nurses) was affected by the availability of the new test, and in particular, the adherence to a negative test result. Safety and adherence are the object of the present paper. The safety did not concern RDT testing in itself, but rather the subsequent prescribing; theoretically, harm could be caused both by a false negative and a false positive result (Bisoffi & Van den Ende 2008): False negatives would not be treated for malaria; false positives would risk to be left without treatment for the true cause of their fever. We aimed to assess if the new strategy would be at least equivalent to the previous (presumptive) one in terms of short term clinical outcome. The issue of adherence to malaria diagnosis has recently been investigated (Reyburn et al. 2004, 2007; Hamer et al. 2007), but most of these studies were not yet published at the time of our field study (2006). The cost effectiveness of RDT based strategies has been questioned due to lower-than-expected compliance of health workers with the (negative) test result (Bisoffi & Van den Ende 2008; Lubell et al. 2008).
We hypothesized that (i) the short-term health outcome should not be worse in the group submitted to RDT; (ii) less frequent malaria treatment, and more frequent treatment for other causes of fever, mainly antibiotics, should be observed in febrile patients with a negative RTD result; (iii) the opposite should be observed in febrile patients with a positive RTD result. Thus our objectives were: to assess if the short-term outcome (day 4) of patients treated after performing a RDT is at least equivalent (not inferior) to that of controls (without RDT) in terms of clearance of fever and of other major symptoms and signs; and to assess the impact of the introduction of a malaria rapid diagnostic tests (RDT) on clinical decisions by health officers.
- Top of page
- Patients and methods
As shown in the flow charts (Figures 1 and 2), 2861 patients were assessed for potential inclusion in the study in the dry season and 3573 in the rainy season. One thousand nine hundred and ninety-five and 2237 respectively did not respond to the inclusion criteria for reasons outlined in the flow charts, but 866 and 1336 were eligible for inclusion, of whom 14 and 19, respectively, refused. The remaining 852 and 1317 patients were recruited and randomized either to the RDT (P_RTD) or presumptive management on clinical grounds only (P_CLIN).
Of 852 febrile patients recruited in the dry season, 404 were submitted to the RDT and 448 treated presumptively. Of 1317 recruited in the rainy season 654 had RDT and 663 presumptive treatment. Demographic and clinical characteristics were evenly distributed in the randomized groups (Table 1).
Table 1. Comparison of randomized groups
| M||517 (49.1%)||538 (48.6%)||0.96|
| F||527 (49.1%)||559 (50.5%)|
|Age (years) (median)||4 Q1 = 1, Q2 = 19||4 Q1 = 1, Q2 = 18||0.98|
|T° (median)||38.3 Q1 = 37.8, Q2 = 39.1||38.3Q1 = 37.7, Q2 = 39||0.21|
|Positive thick film||48.6%||46.2%||0.26|
Follow-up at day 4 was possible for 813 of 852 patients in the dry season (95.4%) and 1282/1317 patients (97.3%) in the rainy season. In the dry season, four deaths (two infants aged 10 and 15 months, respectively, and two adults) were recorded in the P_RDT group vs. three deaths (all infants) in the P_CLIN group (P = 0.71), while in the rainy season one death was reported in each cohort (one infant submitted to RDT and one adult not submitted) (P = 1). Subsequent microscopy showed that in the dry season no fatality was due to malaria (in only one case was the thick film positive, but at a very low parasitaemia of 120 parasites/μl). One of the infants had been treated for malaria only, after a positive RDT test, which was subsequently found to be a false positive. The infant who died in the rainy season had a malaria infection at very high parasitaemia (about 15%) and was appropriately treated after a positive RDT; the adult had no malaria parasite and presumably died of pneumonia.
The rates of persistence of fever (8.2% in both groups in the dry season, 3.9 vs. 3.7% in the rainy season) and of other symptoms (20.1 vs. 20.3% in the dry season, 6.4 vs. 8.5% in the rainy season) were also similar in both groups (Table 2).
Table 2. Clinical outcome at follow-up (day 4th) in the two arms
| ||RDT (%)||Clinical (%)||p|
|Death||4/388 (1.0)|| 3/425 (0.7)||0.71|
|Persistence of fever||32/388 (8.2)||35/425 (8.2)||0.99|
|Persistence of other symptoms||78/388 (20.1)||86/425 (20.2)||0.96|
|Death||1/636 (0.15)||1/646 (0.15)||1|
|Persistence of fever||25/636 (3.9)||24/646 (3.7)||0.83|
|Persistence of other symptoms||41/636 (6.4)||55/646 (8.5)||0.16|
Comparing the seasons as a whole, a significant worse outcome was found in the dry season (Table 3): lower rate of resolution of fever and of other symptoms (8.2% and 20.1%vs. 3.8% and 7.4%, respectively: P = 0.00 and 0.00), and higher death rate (7/813 or 0.9%vs. 2/1282 or 0.16%: P = 0.03).
Table 3. Clinical outcome at follow-up (day 4th) in the two seasons
| ||Dry season (%)||Rainy season (%)||P|
|Death||7/813 (0.9)|| 2/1282 (0.16)||0.03|
|Persistence of fever||67/813 (8.2)|| 49/1282 (3.8)||0.00|
|Persistence of other symptoms||164/813 (20.1)|| 96/1282 (7.4)||0.00|
In the dry season, the rapid test result was positive in 113 of 404 (28%) patients and indeterminate in 4 (not considered in the analysis). In the rainy season, the RDT result was positive in 443 of 650 (68.2%) cases and indeterminate in four (not considered in the analysis). In both seasons, no significant difference was found between the two randomized groups in the frequency of final diagnosis of malaria (83.7%vs. 80.8% in dry season, P = 0.28, and 92.7%vs. 91.9% in rainy season, P = 0.58) and antimalarial treatment (84.2%vs. 80.1% in dry season, P = 0.13, and 92.5%vs. 92.0% in rainy season, P = 0.73) (Table 4).
Table 4. Diagnosis and treatment of malaria and other conditions in the two arms
| ||RDT (%)||Clinical (%)||P|
|Diagnosis of malaria||338/404 (83.7)||362/448 (80.8)||0.28|
|Antimalarial treatment||340/404 (84.2)||359/448 (80.1)||0.13|
|Other diagnosis||257/404 (63.7)||328/448 (73.2)||0.00|
|Antibiotic treatment||229/404 (56.7)||275/448 (61.4)||0.16|
|Diagnosis of malaria||606/654 (92.7)||609/663 (91.9)||0.58|
|Antimalarial treatment||605/654 (92.5)||610/663 (92.0)||0.73|
|Other diagnosis||343/654 (52.4)||344/663 (51.9)||0.83|
|Antibiotic treatment||331/654 (50.6)||334/663 (50.3)||0.93|
In the dry season, an alternative diagnosis was made more often in P_CLIN than in P_RTD (73%vs. 64%, P = 0.00), but with no significant difference in antibiotic prescription (56.7%vs. 61.4%, P = 0.16). In the rainy season, both the frequency of other diagnoses and of antibiotic prescriptions were similar in both groups (52.4%vs. 51.9%, P = 0.83, and 50.6%vs. 50.3%, P = 0.93, respectively) (Table 4).
A further analysis was carried out in the P_RDT arm only, in order to assess how decisions were influenced by the RDT result. As expected, in both seasons the frequency of malaria diagnosis and treatment was significantly higher for positive vs. negative RDT results (Table 5). In the dry season, 92% and 95.6% of patients with a positive RDT were diagnosed and treated for malaria, vs. 80.8% and 79.8% with a negative RDT (P = 0.00 and 0.00). In the rainy season, 96.2% and 98.2% of positive patients were diagnosed and treated for malaria, vs. 85.0% and 82.6% of negative patients (P = 0.00 and 0.00).
Table 5. Malaria diagnosis and treatment according to RDT result
| ||Positive RDT (%)||Negative RDT (%)||P|
|Diagnosis of malaria||104/113 (92.0)||232/287 (80.8)||0.00|
|Antimalarial treatment|| 108/113 (95.6)||229/287 (79.8)||0.00|
|Other diagnosis||75/113 (66.4)||179/287 (62.5)||0.45|
|Antibiotic treatment||69/113 (61.0)||157/287 (54.7)||0.25|
|Diagnosis of malaria||426/443 (96.2)||176/207 (85.0)||0.00|
|Antimalarial treatment||435/443 (98.2)||171/207 (82.6)||0.00|
|Other diagnosis||214/443 (48.3)||127/207 (61.4)||0.00|
|Antibiotic treatment||205/443 (46.2)||124/207 (59.9)||0.00|
In the dry season the two arms did not differ significantly in the frequency of other diagnoses or antibiotic prescription (66.4%vs. 62.5% and 61.0%vs. 54.7%, P = 0.45 and 0.25, respectively) (Table 5). This was not the case in the rainy season, when the diagnosis and treatment of other conditions were less frequent in RDT positive patients (48.3%vs. 61.4% and 46.2%vs. 59.9%, P = 0.00 and 0.00, respectively) (Table 5). The additional training had no apparent effect on the diagnosis and treatment of malaria and of other conditions for RTD negative patients (Table 6).
Table 6. Effect of training: comparison between both seasons regarding diagnosis and treatment of malaria and other conditions
|Diagnosis and treatment||Dry season (%)||Rainy season (%)||P|
|RDT negatives diagnosed as malaria||232/287 (80.8)||176/207 (85.0)||0.28|
|RDT negatives treated for malaria||229/287 (79.8)||171/207 (82.6)||0.50|
|Alternative diagnoses in RDT negatives||179/287 (62.5)||127/207 (61.4)||0.89|
|Antibiotic treatment in RDT negatives||157/287 (54.7)||124/207 (59.9)||0.29|
- Top of page
- Patients and methods
Despite a growing mass of literature on RDT for malaria, we were surprisingly unable to find any single paper on the safety of a RDT based strategy, compared with presumptive malaria management, and very few articles on adherence to RDT results (Hamer et al. 2007; Reyburn et al. 2007; Lubell et al. 2008). Test based strategies might fail their purpose to save unnecessary costs, and be even dangerous, if clear evidence on both aspects is not provided (Bisoffi & Van den Ende 2008; Lubell et al. 2008).
No significant difference was found in the clinical outcome between the two randomized groups (Table 2). Because of the poor adherence to the test result, we were not able to show if the RDT based strategy can be considered safe. Other authors found that malaria infections missed by microscopy and therefore untreated are not associated with mortality risk (Njama-Meya et al. 2007). Similar evidence for RDT is, however, still lacking.
Another potential harm affects RDT false positive patients. In endemic areas the presence of malaria parasites in blood may not reflect a clinical malaria episode (Schellenberg et al. 1994). Thus some febrile, RDT positive patients may be simple carriers of malaria parasites, with another (potentially severe) disease. The harm from a missed treatment, under the influence of a positive malaria test, might not be negligible. In one case in the dry season, a child with a false positive RDT result was treated for malaria only and subsequently died (presumably of pneumonia). The difference in mortality and clinical outcome between the two seasons (Table 3), as was recently found in Burkina Faso by other authors, raises concern (Kynast-Wolf et al. 2006). The difference might be due to a different epidemiological pattern between the two seasons and/or to the fact that a patient was less likely to receive the appropriate treatment for her/his condition in the dry season when malaria is much less frequent.
No significant difference was found in either season between the two randomized groups for clinical decisions concerning malaria diagnosis and treatment. Other potential causes of fever were more frequently diagnosed in the dry season, though the frequency of antibiotic treatment was similar in the two arms (Table 4). In both seasons a positive RDT result was significantly correlated with the decision to treat for malaria (Table 5), but negative patients were also diagnosed as malaria cases in 80% (dry season) to 85% (rainy season) of cases. The expected, higher frequency of alternative diagnoses and treatments after a negative RDT result was only observed in the rainy season (after a second intensive 3-day training session) (Table 5). In general, more than half patients were treated with antibiotics, in both arms and in both seasons (Table 4), and so were about half patients with a positive RDT, despite being almost all diagnosed as malaria cases (Table 5). The so called ‘double diagnosis’ (and treatment) is questioned by Public Health officers as a waste of resources. In individual care, however, nurses often prefer to treat a potentially harmful cause of fever if they cannot rule it out.
Other authors have very recently addressed the adherence issue in African, Anglophone countries, (Hamer et al. 2007; Reyburn et al. 2007; Lubell et al. 2008) while we are not aware of any published study from Francophone Africa as yet. While the above referred studies have generally found a poor compliance with the negative test result, none has shown such a low adherence as in ours. Undoubtedly local concepts of illness influence malaria management (Beiersmann et al. 2007; Some & Zerbo 2007). Moreover, nurses were not compelled to refrain from malaria treatment in case of a negative result. This could be regarded as a major flaw in the study design. Also, ACT were not yet available in most health facilities: cheaper regimens (generally including amodiaquine), were used in most cases. Other authors have found in Kenya that clinical officers tend to reserve ACT for positive cases, and to treat negative patients with cheaper regimens (Zurovac et al. 2008).
During the second training session in the rainy season, it was particularly stressed that a negative test virtually excludes malaria. The result was frustrating. Even more negative patients were treated for malaria than in dry season (Table 6), probably reflecting the conviction that in the rainy season every febrile patient has malaria. Nurses intuitively feel that the pre-test probability of malaria is so high, that the disease remains likely even after a negative test.
This study has a major limitation. As we did not expect such poor adherence, the study failed to fulfil its first objective, that was, to assess a possible difference in clinical outcome between the two arms. One can argue that the study design was flawed, because in order to fulfil the main objective, adherence should have been enforced and strictly supervised. A posteriori, this is obviously true, but we planned to study the safety issue under near-real conditions, rather than in a quasi-experimental context that might not reflect everyday practice.
Some operational constraints must also be acknowledged. Not all clinical officers of the 10 health centres participating in the study were trained, due to logistic problems: while it was agreed that patients should be attended by trained nurses during the study period, this did not always prove possible. In some cases, the study supervisors found that the diagnosis and treatment decision were made before knowing the RDT result. Clearly, if this happened in the context of an intensively supervised study, we may expect it to occur even more in everyday practice. Finally, as the frequency of fever persistence at follow-up was lower than expected in both arms, the sample size would have been inadequate to the primary endpoint in any case.
Safety of a RDT-based strategy (especially for children below 5 years) remains a fundamental issue that should be addressed by future research. If safety is clearly demonstrated, policies to promote adherence will have a better evidence base (Bisoffi & Van den Ende 2008) Evidence in this respect should probably be pooled from different study settings in different countries. More research is also needed on adherence. Policy makers should seriously consider the issue of (non) adherence to diagnostic tests. Operational research should concentrate on effective strategies to promote compliance.
- Top of page
- Patients and methods
We thank the patients who participated in the study, the personnel of the health centres involved and all the investigation staff, as well as all the health staff of the ‘An Ka Heresso’ Project at the time of the field survey: Giuseppe Baracca (who first suggested a study on RDT), Klara Van den Ende, Annalisa Romeo, Bouma Neya, Mamadou Traore and Rosalie Midjour. We also thank Maria Gobbo, Monica Degani and Barbara Paiola for their active collaboration to the training of the investigational staff, Marleen Boelaert for support to data analysis and interpretation and Marco Albonico for critical reading of the manuscript. This study was funded by UNIDEA – UNICREDIT Foundation.