Study design
A prospective, single-blind evaluation of two RDTs compared to slide microscopy.
Study site
The study was carried out in the Gerihun community health centre (CHC) in Bo district, one of five CHCs in which MSF is working. Bo district, situated in south-east Sierra Leone, is a hyperendemic malaria region where transmission is perennial. In 2008, of the 417,576 consultations in all MSF health facilities, 181,711 (43.5%) had uncomplicated malaria confirmed by RDT. Severe malaria is the principal cause of morbidity in the area, accounting in 2008 for 54.3% (3733/6875) of all admissions of children under five years of age in the paediatric department of the MSF-run referral hospital (MSF, internal source, 2008/2009). For its catchment population of 152,000 people MSF offers free malaria diagnosis, treatment and prevention in a setting of free primary heath care.
Enrolment of study participants
All children under five years of age consulting Gerihun CHC for fever and/or history of fever in the previous 72 hours were screened for clinical suspicion of malaria according to routine CHC protocols. Children were included in the study if they satisfied all of the following screening criteria: age between two and 59 months, suspected malaria defined as fever (axillary temperature > 37.5°C), and/or history of fever in the 72 hours prior to attending Gerihun CHC, no signs of severe disease and/or clinical danger, no treatment for malaria administered within the previous four weeks, and signed, informed consent by responsible caregivers.
Patients were included for follow-up of the CareStart™ three-line P.f/PAN-pLDH test if they satisfied all of the following criteria in addition to the initial inclusion criteria: positive CareStart™ three-line P.f/PAN-pLDH test for at least one malaria species and positive malaria blood smear at Day 0 (day of inclusion), place of residence less than one hour walking distance from Gerihun CHC, and able to attend follow-up for up to 28 days.
Sample size
To detect Plasmodium falciparum with a sensitivity of 90%, 138 blood smear positive patients were required (alpha error 0.05, precision 5%). Similar calculations applied to the specificity, so an additional 138 blood smear negative patients needed to be tested. To allow for an estimated 5% of patients with incomplete information, patients were continuously recruited until a minimum of 145 positive and 145 negative blood smears had been included. The study was not powered to evaluate detection of malaria species other than Plasmodium falciparum, or mixed infections that included Plasmodium falciparum.
A convenient sample size of 145 patients with a positive CareStart™ three-line P.f/PAN-pLDH test result and positive malaria blood smear at Day 0 (day of inclusion) was chosen to describe the time taken for the false-positive CareStart™ three-line P.f/PAN-pLDH tests to become negative.
Study procedures
On the day of inclusion, demographic and clinical information were recorded, thick and thin blood smears prepared, and the two malaria RDTs (CareStart™ three-line P.f/PAN-pLDH test and Paracheck-Pf®) performed. Treatment was based on the results of the blood smear, taken as the gold standard reference diagnosis. All positive patients were treated for free according to the following treatment scheme recommended by World Health Organization (WHO), the national and MSF protocol: artesunate 4 mg/kg body weight and amodiaquine 10 mg/kg body weight once a day on Days 0-2. Patients eligible for the follow-up part of the study had all been treated with three doses of artesunate and amodiaquine, administered at the study site, and were observed over the first 30 minutes for immediate vomiting. Failure to complete the full course of treatment at the study site resulted in exclusion from the follow-up study part.
Follow-up of CareStart™ three-line P.f/PAN-pLDH false-positive tests
Patients with a positive CareStart™ three-line P.f/PAN-pLDH test and a negative blood smear on the last day of treatment, Day 2, were followed from Day 2 with repeated CareStart™ three-line P.f/PAN-pLDH tests and blood smears at 7-day intervals until the CareStart™ three-line P.f/PAN-pLDH test results were negative or a maximum of 28 study inclusion days were reached.
Evaluation of malaria rapid diagnostic tests (RDTs)
Both malaria RDTs were of traceable quality (standard supplier, used within the shelf life) and had a guaranteed history of proper storage and transport conditions. All RDTs were performed and interpreted according to the manufacturers' instructions.
1) CareStart™ three-line P.f/PAN-pLDH (batch number AI8IL, manufacturer's catalogue number G0121, AccessBio, New Jersey, USA) is an individually packaged test cassette, diagnosing Plasmodium infections by pLDH detection, distinguishing between Plasmodium falciparum and the other malaria species Plasmodium vivax, Plasmodium malariae, or Plasmodium ovale. It requires 5 μl of whole blood to be collected with a pipette provided by the test kit. Test results need to be read after 20 minutes.
2) Paracheck-Pf® (batch numbers 31930 and 31932, manufacturer's catalogue number 30301025, Orchid Biomedical System, Goa, India) is also an individually packaged test cassette, diagnosing Plasmodium falciparum infections by HRP2 detection. It requires 5 μl of whole blood to be collected with a sample applicator provided by the test kit. Test results need to be read after 15 minutes.
Test results were read by two independent readers, who received training before the beginning of the study. The readers were taught the procedure and interpretation of the RDT results. The study did not start until the study supervisor was satisfied that the readers were familiar with handling and reading of both tests. Preparation and reading of the RDTs were continually reviewed under the observation of the study supervisor. The first reading was performed at the time specified by the manufacturer. The person doing the first reading was a trained laboratory technician and also prepared the two RDTs beforehand. As the two manufacturers provided no maximum time before reading the results, the second reading was performed less than 10 minutes after the first. The second reading was done by a trained layperson. The readers were blinded to each other's results and to that of the blood smear. For each RDT the result was classified as negative, positive (differentiating for the CareStart™ three-line P.f/PAN-pLDH test Plasmodium falciparum or other Plasmodium species) or invalid. All tests without a control line were considered invalid by the first reader and were repeated.
Laboratory procedures
Thick and thin blood smears and RDTs were performed from the same finger-prick of blood. Thick and thin smears were prepared on the same slide and were air-dried. The thin blood smears were fixed with methanol and the thick smears left unfixed. Each slide was then stained with 10% Giemsa solution for ten minutes. All blood smears were examined microscopically under oil immersion (× 1,000 magnification). The thick smears were used for diagnosis of Plasmodium species and for parasite-density counting. Smears were considered negative if no parasites were seen in 100 oil-immersion fields. For positive smears, the number of parasites was counted against 200 white blood cells (WBC), or 500 WBC for low-density infections. Parasite density was calculated assuming 8,000 WBC per microlitre. The thin smears were examined to confirm the parasite species for positive samples. Gametocyte presence or absence on all slides was also recorded, and species identification and gametocyte-density counts (number of gametocytes/1,000 WBC) performed.
All inclusion slides were double-read, blinded, by experienced technicians (one in the MSF project in Sierra Leone and one in MSF Austria that has staff with a special focus on laboratory quality). A third reading was performed in case of discordance, such as positive/negative discordance for asexual stages, species discordance for asexual stages, asexual density discordance (difference in parasitaemia ≥50%), and/or positive/negative gametocyte discordance. Third readings were performed by the Department of Infectious and Tropical Diseases at the London School of Hygiene and Tropical Medicine and were taken as the definitive results for these slides being disconcordant between the first and the second reader.
Assessment of ease of use of RDTs
At the end of the field part of the study, the ease of use of both RDTs was compared using quantitative (storing conditions) and qualitative criteria. The study laboratory technicians who performed the RDTs were asked to rank the tests independently in order of preference where "2" corresponded to their most preferred test and "0" to their least preferred in each of the following categories: Ease and safety of taking blood, ease of adding reagents, ease of interpretation and quality of instructions. The trained laypersons who were performing second readings of tests were asked to rank the tests independently based on the ease of interpretation, and for any additional comments where "2" corresponded to their most preferred test and "0" to their least preferred.
Data management and analysis
All data were recorded directly on an individually numbered case report form. Data were entered into EpiData 3.0 software (The EpiData Association, Odense, Denmark). Data cleaning was done to check for inconsistencies in data entry and responses. Data analysis was conducted using Stata 8.0 (Stata Corporation, College Station, Texas, USA) and SPSS 11.0 (SPSS. Chicago, USA).
Baseline characteristics (demographic, clinical, and parasitological) were analysed using descriptive statistics (means, inter-quartile ranges and values). Sensitivity was defined as the proportion of true malaria cases (positive blood smears) that were correctly identified by positive RDTs. Specificity was the proportion of true negative malaria cases (negative blood smears) that were correctly identified by negative RDTs. Positive predictive value was the proportion of true malaria cases (positive blood smears) among the total number of positive RDTs. Negative predictive value was the proportion of true negative malaria cases (negative blood smears) among the total number of negative tests.
Sensitivity, specificity, positive and negative predictive values of each test were calculated for RDTs performed at the day of inclusion and estimated using microscopy as the reference standard. Sensitivity, specificity, positive and negative predictive values were also stratified by category of parasitaemia (<100, <1,000, ≥1,000 parasites/μl). The first reading of the RDTs was used for these calculations. A 95% confidence interval (95% CI) was given for each parameter. Proportions (paired data) were compared using the McNemar test. Agreement between readers of each RDT was assessed using the kappa coefficient. A Kappa coefficient greater than 0.8 was considered as a measure of good agreement.
Ethical issues
Approval was received from the Ethics Review Board of MSF and the Research and Ethics Committee of the Ministry of Health and Sanitation of Sierra Leone. Informed, written consent was sought from responsible caregivers of all children participating in the study. The caregivers were provided with an information sheet and had the study purpose explained in their own language by the study personnel (Mende, Creole or English). Participation was entirely voluntary.