Monthly sulfadoxine/pyrimethamine-amodiaquine or dihydroartemisinin-piperaquine as malaria chemoprevention in young Kenyan children with sickle cell anemia: A randomized controlled trial

The Enhancing Preventive Therapy of Malaria in children with Sickle cell anemia in East Africa (EPiTOMISE) trial was an open-label, parallel assignment randomized trial. The trial protocol was approved by ethics committees of Moi University (0001907) and Duke University (Pro00077428), registered with clinicaltrials.gov↗ (NCT03178643↗) and the Pan-African Clinical Trials Registry (PACTR201707002371165), and approved by the Kenyan Pharmacy and Poisons Board (ECCT/17/08/06). The Duke Clinical Research Institute oversaw the trial, with technical and infrastructural support from the Duke Global Health Institute and Moi University School of Medicine. Written informed consent in English, Kiswahili, or Dholuo was obtained from the parents or guardians of all children. Independent study monitoring was performed biannually; a medical safety monitor reviewed all severe adverse events (SAEs); and a Data Safety and Monitoring Board (DSMB) convened by the National Heart Lung and Blood Institute (NHLBI) provided biannual review of progress and safety. SP-AQ was supplied free of charge to the trial by Guilin Pharmaceuticals, which had no role in the design, conduct, analysis, reporting, or decision to report the results. This study is reported as per the Consolidated Standards of Reporting Trials (CONSORT) guideline (S1 Checklist).

EPiTOMISE was conducted in Homa Bay, Kenya, a town of historically high malaria transmission; in 2012 to 2013, 46% of children with fever presenting to Homa Bay County Hospital had malaria [16]. Children between 12 months and 10 years of age with homozygous hemoglobin S (HbSS) disease confirmed by hemoglobin electrophoresis and who met additional eligibility criteria were enrolled from a routine SCA clinic at Homa Bay County Hospital and randomized in 1:1:1 ratio to receive either daily Proguanil, monthly SP-AQ, or monthly DP in an open-label fashion. Randomization was performed through the electronic data capture program using a block allocation with a block size of 9. No stratification criteria were used.

The inclusion criteria for enrollment were as follows: age greater than 12 months and less than 10 years at enrollment; current attendance at or willingness to attend the study clinic; residence in either Homa Bay County or the Rongo or Awendo subcounties of Migori County; confirmed hemoglobin genotype of HbSS by electrophoresis or high-performance liquid chromatography (HPLC); no immediate, apparent, or reported plans to relocate residence in the next 2 years; ability to take oral medication and be willing to adhere to the medication regimen or caregiver willingness to give the medical regimen as prescribed; and ability and willingness of parent or legally authorized representative (LAR) to give informed consent. From May 2019 were added the criteria at screening of hemoglobin concentration ≥6.5 g/dL and alanine aminotransferase (ALT) ≤50 U/L. The exclusion criteria were as follows: taking routine antimalarial prophylaxis for another indication (including co-trimoxazole for HIV infection); known allergy or sensitivity to sulfadoxine, pyrimethamine, amodiaquine, proguanil, dihydroartemisinin, piperaquine, artemether, lumefantrine, penicillin (if under 5 years old), or derivatives of these compounds; known chronic medical condition other than SCA (i.e., malignancy, HIV) requiring frequent medical attention; currently participating in another clinical research study, or having participated in one in the prior 30 days; living in the same household as a previously enrolled study participant; chronic use of medications known to prolong the QT interval in children; Fridericia’s corrected QT interval (QTcF) >450 milliseconds (ms); or receipt of a transfusion of red blood cells in the 120 days prior to screening.

Following randomization to chemoprevention regimen, children were dispensed standard doses of each study drug at monthly follow-up visits, which were self-administered by caregivers; dosing (see S1 Text) was weight-based for Proguanil and DP and age-based for SP-AQ. Both SP-AQ (SPAQ-CO) and DP (D-Artepp) were acquired from a manufacturer with WHO prequalification (Guilin), the former supplied without charge by the manufacturer; Proguanil was manufactured by Cosmos Ltd (Nairobi, Kenya). Per local guidelines, children under 5 years also received daily penicillin at standard age-based doses, and all children were enrolled in the Kenya National Hospital Insurance Fund. Adherence to administration of prevention regimen was assessed at each monthly follow-up visit by trial staff via structured queries.

Caregivers were encouraged to call trial staff and report for an acute visit if the child was unwell. At these and at routine monthly visits, children with fever >38°C or a history of subjective or objective fever in the prior 24 hours were tested for malaria with a rapid diagnostic test (RDT) detecting the parasite antigens HRP2 and pLDH (SD Bioline Malaria Ag P.f/Pan COMBO, Alere). Children with positive RDT results were treated with a standard weight-based course of artemether-lumefantrine.

At each monthly visit, we performed a complete review of systems and physical exam, assessed concomitant medications, and collected an interval history, including interval tests or treatments for malaria with verification when possible by review of outside records, self-reported painful crises and dactylitis, and SAEs. Venous blood was collected every 3 months for complete blood count, serum creatinine (sCr), and ALT and every 6 months for hemoglobin electrophoresis; clinical laboratories met ISO 15189 requirements. From May 2019, hemoglobin concentration was also measured at each visit using a point-of-care test (HemoCue Hb 301). At each acute care and routine visit, capillary blood was collected as a dried blood spot (DBS) for post hoc molecular detection of Plasmodium falciparum. In children allocated to DP residing in Homa Bay town, we repeated an electrocardiogram (ECG) 4 to 6 hours following the third dose of each monthly DP course.

All children were followed for 12 months after randomization. The primary outcome was clinical malaria incidence, defined as subjective or objective fever with the presence of P. falciparum parasites measured by a malaria RDT or by severe malaria using a standard definition [17]. Secondary parasitologic outcomes were hospitalization for malaria, light microscopy–positive malaria, unconfirmed malaria, and asymptomatic parasitization. The main hematologic outcome was the incidence of painful events, defined as an episode of pain lasting 2 hours or more without an obvious cause, and additional secondary hematologic outcomes were dactylitis, defined as pain or tenderness with or without swelling of the hands or feet, transfusion of blood products, severe anemia (defined as hemoglobin concentration <5.5 g/dL), and acute chest syndrome. See S1 Text for full outcome definitions. Safety outcomes included SAEs, ALT elevation, anemia, and, in DP recipients, prolongation of the QT interval corrected for heart rate by Fridericia’s method (QTcF). P. falciparum was detected in all DBS using a real-time PCR assay [18]. In selected participants, glucose-6-phosphate dehydrogenase (G6PD) was genotyped by PCR amplification and bidirectional Sanger sequencing across nucleotides 202 and 376 that encode the most common A- deficiency variant in East Africa [19,20], and polymorphisms in CYP2C8*2 were similarly genotyped (see S1 Text for molecular methods) [21,22].

To estimate the sample size, we assumed 3.7 episodes per patient-year (PPY) using the malaria rate in 2011 by children enrolled in the RTS, S/AS01 vaccine trial in Siaya, Kenya [23], which historically has had similar transmission to Homa Bay [24], and a similar expected distribution of episodes shifted by treatment effect factors. Using simulations, we found enrolling 65 patients per treatment arm would provide >90% power to detect a reduction of 40% in the DP arm, approximately 40% power to detect a reduction of 20% in the Proguanil arm, and >90% power to detect a reduction of 40% in either arm. Allowing for 20% loss to follow-up, a sample size of 246 (82 per group) provided at least 90% power, assuming that each comparison between experimental arm and control would be tested at the 0.0269 level (Dunnett’s correction) in order to preserve a nominal alpha level of 0.05.

The analysis was initially planned using the Intention-To-Treat (ITT) population. This was changed to the As-Treated (AT) population during the finalization of the statistical analysis plan owing to the unanticipated crossover from October 2018 to May 2019 from the SP-AQ group to Proguanil; this crossover resulted from the temporary suspension of SP-AQ administration at the request of the DSMB to allow assessment of 4 deaths among SP-AQ recipients. Participants who were still in active follow-up assigned to the SP-AQ arm in May 2019 were crossed back over to SP-AQ at that time. Analyses using the According-to-Protocol (ATP) were also evaluated to compare patients treated as randomized (e.g., did not crossover).

Results are presented by treatment arm using the mean, standard deviation (SD), median, 25th and 75th percentiles (Q1, Q3), and range as appropriate for continuous variables, as well as the count and percentage for non-missing data for categorical variables.

Primary and secondary incidence rates (IRs), or the number of episodes PPY at risk, for each experimental arm were compared to the control using a generalized regression model with a negative binomial (NB) distribution to allow for interdependence between multiple episodes. This model structure allowed the use of the actual follow-up time as the offset variable and a robust sandwich variance estimate. Model distributions considered included the NB, Poisson and zero-inflated Poisson, and NB with the distribution chosen based on the residual plots. Model results are presented using the incidence rate ratio (IRR) with 95% confidence interval (CI). Prespecified subgroup analyses explored the impact of changes to protocol and access to care prompted by the COVID-19 pandemic and other covariates on clinical malaria and painful events by including an interaction term in the model; the main pandemic-related protocol change was that monthly follow-up visits alternated between in-person and telephone visits. Additional covariates at enrollment included age, hydroxyurea use, sex, hospitalization in prior 12 months, and hemoglobin concentration. Secondary dichotomous outcomes are presented as cumulative incidence (CI) rates per treatment arm and compared to the control using Fine and Gray’s method to allow death to be a competing risk and censoring of patients lost to follow-up within the ITT and ATP populations. Comparisons are presented using hazard ratio (HR) with 95% CI. The AT population was not evaluated for comparing cumulative IRs due to multiple crossing of treatment arms with events possibly occurring in each crossed arm. S1 Analysis describes all comparisons of the monthly groups to the daily Proguanil group; analyses comparing monthly DP to monthly SP-AQ were added later.

Each primary and secondary outcome comparison used an α = 0.0269 per Dunnett’s test for multiple comparisons to the same control group. Tables were generated from complete case data only. No imputation of baseline or endpoint data was done. All analyses were conducted using SAS 9.4 (SAS Institute, Cary, NC, USA).

The trial was conducted between January 23, 2018, and December 15, 2020 (Fig 1A). A total of 315 children were screened and consented, and 246 were enrolled and randomized. Overall, 53.3% (n = 131) were boys; the mean age was 4.6 ± 2.5 years; 83.7% (n = 205) reported receiving routine sickle cell care; 69.0% (n = 169) reported being treated for malaria in the prior 12 months; and the mean hemoglobin concentration was 7.9 ± 1.2 g/dL. Baseline characteristics were similar across the 3 groups (Table 1). Between October 2018 and May 2019, the DSMB temporarily suspended administration of SP-AQ owing to the observation that the first 4 participant deaths occurred among children receiving SP-AQ. Pending the implementation of enhanced safety monitoring, addition of eligibility criteria, and adoption of a stopping rule, the SP-AQ group temporarily crossed over to Proguanil; SP-AQ dispensing resumed in May 2019. Overall, of 83 children randomized to SP-AQ, 31 (37.3%) crossed to Proguanil, a mean of 155 ± 82 days after randomization, and 17 of these (54.8%) remained on study in May 2019 and subsequently restarted SP-AQ. As a result of this unexpected protocol change, primary efficacy analyses were performed on the AT population, defined as the population of all participants with treatment assignment reflecting the actual treatment received during the month under observation, and safety analyses performed on the ATP population. Overall 8.1% (n = 20) of children were withdrawn or lost to follow-up.

Owing to substantial local malaria transmission, county authorities began a campaign in February 2018 of indoor residual spraying with primiphos-methyl. Each yearly campaign lasted 6 weeks and was reinitiated in January 2019 and February 2020 (Fig 1B).

Adherence to prescribed chemoprevention was overall high (Table A in S1 Text), although more guardians reported difficulty getting children to take SP-AQ (27.5%) than Proguanil (3.7%) or DP (6.3%) and forgetting to administer Proguanil (16%) than SP-AQ (2.5%) or DP (0%).

We recorded 9 primary outcome malaria events overall, yielding a clinical malaria event IR of 0.04 events/PPY at risk (Table 2). In the AT population, the rates were not significantly different to the Proguanil group (0.03/PPY) in both the recipients of SP-AQ (0.09/PPY; IRR: 3.05; 95% CI: 0.36 to 26.14; p = 0.39) and DP (0.04/PPY; IRR: 1.36; 95% CI: 0.21 to 8.85; p = 0.90). Similar malaria rates were observed in the ITT population (Table 2), and when comparing DP to SP-AQ recipients (Table B in S1 Text).

We recorded 209 self-reported painful events, yielding a rate of 4.30 events/PPY at risk (Table 2). In the AT population, compared to the rate of painful events in the Proguanil group (4.42/PPY), painful events were not significantly different in the recipients of SP-AQ (4.28/PPY; IRR: 0.97; 95% CI: 0.70 to 1.35; p = 0.96) and DP (4.18/PPY; IRR: 0.95; 95% CI: 0.65 to 1.38; p = 0.92). Similar rates and differences were observed in the ITT population, and when comparing DP to SP-AQ recipients.

Among secondary parasitologic outcomes, DP reduced asymptomatic parasitization by P. falciparum by 79% (IRR: 0.21; 95% CI: 0.08 to 0.56; p = 0.002). Most other secondary parasitologic outcomes were rare and did not vary significantly between chemoprevention groups (Table 3); results in the ITT population were similar (Table C in S1 Text). Among secondary hematologic outcomes, we recorded 81 episodes of dactylitis (0.84 episodes/PPY), and the incidence of dactylitis was reduced relative to that in Proguanil recipients (1.08 episodes/PPY) in the recipients of DP (0.51/PPY; IRR: 0.47; 95% CI: 0.23 to 0.96; p = 0.038). Compared to DP recipients, SP-AQ recipients had similar incidences of asymptomatic P. falciparum parasitization (IRR 2.23; 95% CI: 0.68 to 7.29; p = 0.23) and dactylitis (IRR 1.66; 95% CI: 0.77 to 3.56; p = 0.24) (Table B).

In the AT population analyses of prespecified subgroups on main hematologic outcomes, rates of painful events did not vary between chemoprevention recipients within groups (Fig 2, Table D in S1 Text). These subgroups also did not modify the association of DP with reduced rates of dactylitis, which was observed irrespective of age, sex, baseline hemoglobin, or use of hydroxyurea.

There were 102 SAEs overall in the AT population, in whom 41.9% (90/215) experienced at least 1 SAE, yielding an IR of 0.91 SAEs/PPY at risk. The incidence of SAEs was similar in recipients of Proguanil (0.90 events/PPY; 95% CI: 0.64 to 1.26), SP-AQ (1.10/PPY; 95% CI: 0.78 to 1.54), and DP (0.82/PPY; 95% CI: 0.54 to 1.24) (Table 4). The most commonly reported SAEs were hospitalizations resulting from a painful crisis (91 events), from anemia (33 events), and malaria or sepsis (12 events each) (Table E in S1 Text). One SAE was judged by the safety monitor to be related to chemoprevention regimen: an ALT elevation to 517 in a recipient of SP-AQ, which normalized following cessation of drug. We recorded 72 all-cause hospitalizations in the AT population, yielding an overall rate of hospitalization of 0.67/PPY at risk (Table 4); the prevalence and the rate of hospitalization was similar between groups. Ten children died: 2 receiving daily Proguanil, 7 receiving SP-AQ, and 1 receiving DP (Table 4); relative to Proguanil (cumulative IR [CIR] 2.2%), the risk of death was elevated in SP-AQ recipients (CIR 11.3%; HR 5.44; 95% CI 0.92 to 32.11; p = 0.064), although this did not reach statistical significance and was not elevated in DP recipients (CIR 1.3%; HR 0.61; 95% CI 0.04 to 9.22; p = 0.89). Similarly, relative to DP recipients, the risk of death was elevated in SP-AQ recipients (HR 8.86; 95% CI: 0.74 to 106.55; p = 0.10), although this did not achieve statistical significance (Table B in S1 Text).

Among laboratory events, neither the change in hemoglobin concentration from baseline nor the incidence of severe anemia, defined as a drop of more than 2 g/dL from baseline, differed between chemoprevention groups (Table 4). Similarly, there were no differences between groups in the change in baseline nor the incidence of elevated values of ALT or of sCr. Follow-up neutrophil and platelet counts in the ITT population are presented in Tables F and G in S1 Text. Among the 7 children who died while receiving SP-AQ, 2 were G6PD deficient (2 hemizygote boys), and 1 was heterozygote for a CYP2C8*2 allele associated with reduced metabolism of AQ (Table H in S1 Text).

Among DP recipients, we enrolled 12 children in a substudy of QTcF measurements following the third dose of their monthly courses of DP. Among 134 follow-up QTcF measurements in these children, we recorded 16 (11.9%) measurements of QTcF >450 ms, which occurred in 50% (6/12) of children, with a maximum QTcF of 479 ms (Figure A in S1 Text). There was 1 (0.7%) case of a prolongation of >50 ms QTcF compared to baseline to 52 ms (Table 4, Table I in S1 Text). Mean (SD) QTcF changes from baseline peaked at 17 (13) ms at 6 months and then declined to 8 (16) ms after 12 months of administration.