Factors contributing to anaemia after uncomplicated falciparum malaria in under five year-old Nigerian children ten years following adoption of artemisinin-based combination therapies as first-line antimalarials

Background Artemisinin-based combination therapies (ACTs) have remained efficacious treatments of acute falciparum malaria in many endemic areas but there is little evaluation of factors contributing to the anaemia of acute falciparum malaria following long term adoption of ACTs as first-line antimalarials in African children. Methods Malarious <5 year-olds randomized to artemether-lumefantrine, artesunate-amodiaquine or dihydroartemisinin-piperaquine treatments were followed up clinically for 6 weeks. Anaemia was defined as haematocrit <30%; Malaria-attributable fall in haematocrit (MAFH) as the difference between haematocrit 28–42 days post- and pre-treatment; Total MAFH (TMAFH) as the difference between days 28–42 haematocrit and the lowest haematocrit recorded in the first week post-treatment initiation; Drug-attributable fall in haematocrit (DAFH) as the difference between MAFH and TMAFH; Early appearing anaemia (EAA) as haematocrit <30% occurring within 1 week in children with normal haematocrit pre-treatment. Predictors of anaemia pre-treatment, EAA, MAFH or DAFH >4% were evaluated by stepwise multiple logistic regression models. Survival analysis and kinetics of DAFH were evaluated by Kaplan-Meier estimator and non-compartment model, respectively. Results Pre-treatment, 355 of 959 children were anaemic. Duration of illness >2 days and parasitaemia ≤10,000 μL−1 were independent predictors of anaemia pre-treatment. EAA occurred in 301 of 604 children. Predictors of EAA were age ≤ 15 months, history of fever pre-treatment and enrolment haematocrit ≤35%. The probabilities of progression from normal haematocrit to EAA were similar for all treatments. MAFH >4% occurred in 446 of 694 children; its predictors were anaemia pre-treatment, enrolment parasitaemia ≤50,000 μL−1, parasitaemia one day post-treatment initiation and gametocytaemia. DAFH >4% occurred in 334 of 719 children; its predictors were history of fever pre-and fever 1 day post-treatment initiation, haematocrit ≥37%, and parasitaemia >100,000 μL−1. In 432 children, declines in DAFH deficits were monoexponential with overall estimated half-time of 2.2d (95% CI 1.9–2.6). Area under curve of deficits in DAFH versus time and estimated half-time were significantly higher in non-anaemic children indicating greater loss of haematocrit in these children. Conclusion After ten years of adoption of ACTs, anaemia is common pre-and early post-treatment, falls in haematocrit attributable to a single infection is high, and DAFH >4% is common and significantly lower in anaemic compared to non-anaemic Nigerian children. Trial registration Pan African Clinical Trial Registry (PACTR) [PACTR201709002064150, 1 March 2017].


Background
Artemisinin-based combination therapies (ACTs) have been adopted as first-line treatments of uncomplicated Plasmodium falciparum malaria by over 80 countries [1][2][3]. In many of these countries particularly those in southeast Asia, the period of adoption as first-line treatments has exceeded 10 years [4,5]. In Nigeria, artemetherlumefantrine (AL) and artesunate-amodiaquine (AA) were adopted as first-line treatments in 2005 [6].
Anaemia, a consequence of treated and untreated falciparum malaria, is common before, during or after treatment in up to 40% of African children [7][8][9]. It is a major public health problem, and in anaemic patients, has been attributed to repeated infections within a short time frame, bone marrow dyserythropoiesis and increased splenic clearance of infected and non-infected erythrocytes [10][11][12][13]. Anaemia can also occur in African children with apparently asymptomatic infections [14,15].
Despite adoption as first-line treatments in many countries for over a decade, there is little evaluation of the predictors of anaemia before, during or in the 1 week following initiation of treatment in malarious children. In malarious patients, the effects of infections and the drugs used to treat the infections can be evaluated using the following indices: malaria-attributable fall in haematocrit (MAFH), total malaria-attributable fall in haematocrit (TMAFH) [10], or drug-attributable fall in haematocrit (DAFH) [16]. Such evaluations are urgently needed as they can provide useful leads in the diagnosis and management of the anaemia following ACTs, and the extent to which adoption of ACTs as first-line antimalarials has modified the burden of the anaemia associated with malaria infections. From the clinical point of view, worsening of haematocrit levels or falls in haematocrit following ACTs of uncomplicated falciparum malaria in young African children, until proven otherwise, can be considered an adverse event attributable to the drug(s) used [17] drug attributable falls in haematocrit [16]. However, DAFH can be attributed to interaction between the drug, the infection and the host.
For the reasons indicated above, the aims of present study were: using pre-determined cut-offs (haematocrit <30%, MAFH or DAFH >4%), to evaluate the predictors of the following: anaemia at presentation, progression from normal haematocrit at presentation to anaemia within 1 week of starting treatment (early-appearing anaemia [EAA]), and MAFH or DAFH >4% in a group of under 5 year old malarious children who were treated with AL, AA or dihydroartemisinin-piperaquine (DHP), during an open labelled therapeutic efficacy study. An additional aim was to evaluate the kinetics of the disposition of DAFH in anaemic and non-anaemic children following initiation of treatment.

Study locations
The study took place between June 2014 and September 2015. It was part of a programme to monitor antimalarial therapeutic efficacy at eight sentinel sites located in six geographical areas of Nigeria namely, Ogbia, Neni, Ogwa, Numan, Ilorin, Kura, Bodinga and Ibadan in Bayelsa, Anambra, Imo, Adamawa, Kwara, Kano, Sokoto and Oyo States, respectively. In virtually all study locations, malaria is hyperendemic and transmission occurs all year round; however, it is more intense during the rainy season from April to October. The details of the therapeutic efficacy study will be described elsewhere. The dataset presented here have not been previously published elsewhere.

Study procedures
Standardized procedures and protocol were used at all sentinel sites. Briefly, patients were eligible to participate in the study if they were: aged 6-59 months, had symptoms compatible with acute uncomplicated malaria and P. falciparum mono-infections between 2000 and 200,000 μL −1 of blood, no history of antimalarial drug ingestion in the two weeks preceding enrolment, had no evidence of severe malaria [18,19] and parents or guardians gave written informed consent.
Enrolled patients were randomized to AL, AA or DHP treatments for 3 days (day 0-2) as previously described [20]. The day of presentation (day of starting treatment) was regarded as day 0. Thick and thin blood films were obtained from each child as soon as they came to the clinic and the slides were carefully labelled with the patients' codes and air-dried before being Giemsa-stained. Follow-up with clinical and parasitological evaluation was done daily on days 1-3 and 7, and thereafter, weekly for additional 5 weeks. Parasitaemia, asexual or sexual, in thick films was estimated by counting asexual and sexual parasites relative to 500 leukocytes, or 500 asexual or sexual forms whichever occurred first. From this figure, the parasite density was calculated assuming a leukocyte count of 6000 μL −1 of blood [21]. A slide was considered asexual parasite negative if no asexual parasite was detected after examination of 200 microscope fields. Parasite reduction ratio was defined as previously described [22].

Haematological evaluation
Capillary blood obtained from a finger prick was collected before treatment and during follow-up, and was used to measure haematocrit using a microhaematocrit tube and microcentrifuge (Hawksley, Lancing, UK). Anaemia was defined as a haematocrit <30% and was further classified as mild, moderate or severe if haematocrit was 21-29%, 15-20% or <15%, respectively [8,10].

Definition of haematological parameters evaluated
The following parameters described by Price et al. [10] and modified by Sowunmi et al. [16] were used for evaluation ( Fig. 1): before treatment was defined as the difference between the patient's maximum haematocrit value measured on days 28-42 and that on day 0 (pre-treatment). (ii)Total malaria-attributable fall in haematocrit (TMAFH) was defined as the difference between the patient's maximum haematocrit on days 28-42 and lowest haematocrit value recorded in the first week post-treatment initiation.
(iii)Drug-attributable falls in haematocrit (DAFH) was defined as the difference between MAFH and TMAFH, or as the difference between pre-treatment haematocrit and the lowest haematocrit recorded in the first week post-treatment initiation. (iv)Early-appearing anaemia (EAA) was defined as haematocrit <30% occurring within 1 week in patients who were non-anaemic pre-treatment. (v) Anaemia recovery time: Anaemia recovery time (in anaemic children pre-treatment) was defined as time elapsing from initiation of treatment to attainment of haematocrit value ≥30%. In non-anaemic children (pre-treatment) who subsequently progressed to anaemia following initiation of treatment (early-appearing anaemia), anaemia recovery time was defined as the time from the appearance of (onset of anaemia), to recovery from anaemia (until haematocrit rose to ≥30%).

Kinetics of the disposition of drug-attributable falls in haematocrit
The kinetics of the deposition of the deficits in haematocrit from 30% using the lowest recorded haematocrit in the first week of initiating treatment (DAFH) were evaluated as previously described for the general evaluation of kinetics of the disposition of anaemia in malarious children [23,24]. Briefly, in all children, haematocrit values below pre-treatment levels following initiation of treatment and at follow-up were subtracted from pretreatment haematocrit at each time of measurement until haematocrit rose to or above pre-treatment level and the resulting values plotted against time. The final haematocrit when pre-treatment level was reached was therefore zero in all patients with falls in haematocrit Parameters used for estimation of malaria-related falls in haematocrit (adapted from Price et al. [10] and Sowunmi et al. [16] following initiation of treatment. However, the final haematocrit at the time when pre-treatment haematocrit level was reached was assumed to be 0.01% [24]. The areas under the curve (AUC) of deficit in haematocrit from pre-treatment level versus time were obtained by trapezoidal rule using the computer program Turbo Ken (designed by Clinical Pharmacology Group, University of Southampton, United Kingdom) as previously described [23,24]. AUC was also obtained manually by calculating the average haematocrit values between two consecutive time measurements and multiplying it by the time interval between the measurements, and summing up all the values, in a manner similar to that for the numerical estimation of area under a drug concentration-time curve (AUC) [25]. The unit of quantification would be %.d, if haematocrit values were used or g/L.d if haemoglobin values were used. Haematocrit values may be converted to haemoglobin values by dividing by 3 [26]. Semilogarithm plots of deficit in haematocrit from pre-treatment level versus time were plotted. The apparent terminal elimination rate constant (λ) was obtained by leastsquare regression analysis of the post-peak log-linear part of the plot of deficit in haematocrit from pretreatment level versus time. The apparent terminal halftime of DAFH (t 1/2(DAFH) ) was calculated from ln 2/λ (that is, λt = 0.693). Only children in whom haematocrit values were measured at least 3 times following a drugattributable fall in haematocrit post-initiation of treatment were included in this evaluation.

Statistical analysis
Data were analyzed using version 6 of Epi-Info software [27] and the statistical program SPSS for Windows version 20.0 [28]. Variables considered in the analysis were related to the densities of P. falciparum asexual forms, haematocrit status at presentation and the values of DAFH and AUC (DAFH) . Proportions were compared by calculating χ 2 using Yates' correction, Fisher's exact or Mantel Haenszel tests. Normally distributed, continuous data were compared by Student's t test and analysis of variance (ANOVA). Kaplan-Meier estimator and pairwise log-rank tests were used to determine cumulative risk of anaemia occurring within 1 week post-treatment initiation with AL, AA or DHP. Stepwise multiple logistic regression models were used to test the association between: anaemia pre-treatment, normal haematocrit pre-treatment progressing to EAA within 1 week posttreatment initiation, malaria-attributable fall in haematocrit (MAFH) in excess of 4%, or drug-attributable fall in haematocrit (DAFH) in excess of 4%, and factors that were significant at univariate analysis. Correlation between the values of DAFH and AUC (DAFH) in the same patients was assessed by Pearson's correlation coefficient. Agreement between the values of DAFH and AUC (DAFH) was assessed by Bland-Altman analysis [29]. Data were double entered serially using patients' codes and were only analyzed at the end of the study.

Results
Characteristics of children enrolled in the study and parasite positivity following treatment

Factors contributing to anaemia at presentation
Factors associated with anaemia at presentation are presented in Table 2. In a univariate analysis, an age ≤ 15 months, duration of illness >2 days, and enrolment asexual parasitaemia ≤10,000 μL −1 were associated with anaemia at presentation. In a multivariate analysis, duration of illness >2 days and an enrolment asexual parasitaemia ≤10,000 μL −1 were independent predictors of anaemia at presentation ( Table 2).
Factors contributing to progression from normal haematocrit pre-treatment to anaemia within 1 week post -treatment initiation (early-appearing anaemia) Data for evaluation progression from normal haematocrit pre-treatment to an early-appearing anaemia (EAA) within 1 week post-treatment initiation were available in 604 children. Of these children, 301 children (50%) progressed to anaemia within 1 week post-treatment initiation. The proportions of non-anaemic children who subsequently progressed to anaemia within 1 week of starting treatment were similar with all three treatments (101 of 191 (53%) versus 90 of 187 (48%) versus 110 of 226 children (49%) in AL, AA and DHP treatments, respectively P = 0.59). The probabilities of progression to anaemia 1 week post-treatment initiation were also similar in all three treatments (Log-rank statistic = 1.24; P = 0.54, Fig. 2). Factors contributing to progression to anaemia within 1 week are shown in Table 3. In a univariate analysis, an age ≤ 15 months, history of fever at presentation, enrolment haematocrit ≤35% and persistent asexual parasitaemia till 1 or 2 days post-treatment initiation were significantly associated with progression to anaemia within 1 week post-initiation of treatment. In a multivariate analysis, an age ≤ 15 months, history of fever at presentation and enrolment haematocrit ≤35% were independent predictors of progression to anaemia within 1 week (Table 3).
Time-course of haematocrit following artemisinin-based combination treatments Figure 3a shows the time-course of haematocrit in children who were non-anaemic at presentation and those who were anaemic at presentation following ACTs. Following treatment, there was little or no fall in haematocrit in anaemic children compared to non-anaemic children indicating little or no drug-attributable falls in

Malaria-attributable fall in haematocrit
Data for estimation of MAFH were available in 748 children (428 and 320 non-anemic and anaemic children, respectively). In these children, there was no malariaattributable fall in haematocrit following infection in 54 children (49 and 5 non-anaemic and anaemic children, respectively). The proportion of children with no MAFH following treatment was significantly higher in nonanaemic children compared with anaemic children at presentation (P < 0.0001). Varying degrees of malariaattributable falls in haematocrit were seen in 694 children (379 and 315 in non-anaemic and anaemic children, respectively). Overall, mean MAFH was 6.8% (95% CI 6.5-7.2, n = 694) and it was significantly lower in non-anemic compared to anaemic children Factors associated with MAFH in excess of 4% are shown in Table 4. In a univariate analysis, an age ≤ 38 months, anaemia at presentation, enrolment parasitaemia ≤50,000 μL −1 , parasitaemia 1 day after start of treatment and gametocytaemia within 1 week were associated with MAFH >4%. In a multivariate analysis,     anaemia at presentation, enrolment parasitaemia ≤50,000 μL −1 , parasitaemia 1 day after start of treatments and gametocytaemia within 1 week were independent predictors of MAFH >4% (Table 4).

Drug-attributable falls in haematocrit
Data for estimation of DAFH were available in 929 children (588 and 341 non-anaemic and anaemic children at presentation, respectively). In these children, there was no DAFH following treatment in 210 children (82 and 128 non-anaemic and anaemic children, respectively). The proportions of children without DAFH were significantly higher in anaemic compared to non-anaemic children (P < 0.0001, Fig. 4). Varying degrees of DAFH occurred in 719 children (506 and 213 non-anaemic and anaemic children, respectively; Fig. 4). Overall, lowest falls from pre-treatment levels were seen on day 1, 2 or 3 in 232, 217 or 158 children, respectively. Overall, mean DAFH was 5% (95% CI 4.7-5.2, n = 719) and it was significantly higher in non-anaemic compared to anaemic children Factors associated with DAFH in excess of 4% are shown in Table 5. In a univariate analysis, enrolment body temperature ≥ 38°C, history of fever at presentation, fever 1 day post-treatment initiation, enrolment haematocrit ≥37%, enrolment parasitaemia >100,000 μL −1 , no gametocytaemia at presentation and within one week post-treatment initiation were significantly associated with DAFH >4%. In a multivariate analysis, history of fever at presentation, fever 1 day posttreatment initiation, enrolment haematocrit ≥37%, and a b Fig. 4 Drug-attributable falls in haematocrit in anaemic (a) and non-anaemic children (b) following artemisinin-based combination treatments enrolment parasitaemia >100,000 μL −1 were independent predictors of DAFH >4% (Table 5).
Relationship between area under curve of the plot of deficit in drug-attributable falls in haematocrit and value of drug-attributable falls in haematocrit The relationship between AUC (DAFH) and the value of DAFH was evaluated in 432 children. The mean of the ratio of AUC (DAFH) : value of DAFH was 8.8 (95% CI 7.4-10.3). There was a significantly positive correlation between AUC (DAFH) and value of DAFH (r = 0.47, P < 0.0001). Bland-Altman plots of AUC (DAFH) and multiples of values of DAFH are shown in Fig. 6. The limits of agreement between AUC (DAFH) and 9 or 10 multiples of value of DAFH were narrow. At 9 and 10 multiples of value of DAFH, the limits of agreement were −142.1 -148.6% and −147.4 -142.1%, respectively. The bias at multiples of 9 or 10 value of DAFH was statistically insignificant (P = 0.36 and 0.46, respectively) indicating that AUC (DAFH) and 9 or 10 multiples of value of DAFH can be used interchangeably in the same patient. However, there was a statistically significant bias at 8 multiples of value of DAFH (P = 0.01).

Discussion
In this study, we evaluated prospectively in young malarious children, the factors contributing to: anaemia at presentation, anaemia one week after starting treatment in children who were not anaemic at presentation, the burden of anaemia imposed by a single episode of malaria, the falls and the duration of falls in haematocrit attributable to three ACTs. Our results showed: anaemia is common at presentation and in the one week following initiation of treatment; the burden of anaemia after a single episode of infection is high; and drug related falls in haematocrit are significantly more common and more severe in children who were not anaemic at presentation compared with those who were anaemic at presentation.
Compared with factors contributing to anaemia during the early period of adoption of ACTs as first-line antimalarials in the study areas [8], the factors contributing to anaemia ten years after adoption as first-line treatments appeared to have been considerably modified as shown in the results of the present study. The significant modifications include: reduction in age band susceptible to malarial anaemia at presentation from children 6-59 months to children 6-15 months and relatively low gametocyte carriage at presentation or following treatment. However, the prevalence of anaemia at presentation and the burden of anaemia following a single episode of infection appeared to have been little affected. The low gametocyte carriage ten years following adoption as first-line antimalarials can be attributable to rapid clearance of asexual parasitaemia [17,23,30,31] and alteration in gametocyte sex ratios [32,33] which can produce a female bias sex ratio that is less infective to mosquitoes [34]. It has also been postulated that ACTs can mobilize immature gametocytes from bone marrow to peripheral blood where they can be rapidly eliminated by components of ACTs [30,33]. The latter can also prevent development of immature gametocytes which are non-infective to mosquitoes to mature gametocyte which are infective to mosquitoes [30,33]. This is particularly important because all components of ACTs can kill immature gametocytes. It is not apparent from the results of the present study why a low asexual parasitaemia is a predictor of anaemia at presentation. This finding is paradoxical.
Progression to anaemia within one week of treatment initiation in a half of the children who were not anaemic at presentation can be clearly attributable to relatively high DAFH values with all three treatments. These children differ significantly from anaemic children at presentation in whom DAFH values were relatively low and who did not have a worsening of their anaemia in the one week following treatment initiation (Fig. 5). This finding in anaemic children suggests considerable degree of "haematocrit conservation" [31,35]. Taking together, this would suggest the three ACTs may prevent further falls in haematocrit in anaemic children at presentation and in half of non-anaemic children at presentation.
The burden of anaemia imposed by a single episode of infection was relatively high and it occurred in two-third of the young malarious children. In these children, repeated attacks of acute infections within a short time frame can be expected to lead to development of anaemia [16]. Expectedly, anaemia and gametocytaemia at presentation predicted MAFH >4%. Indeed the risk of MAFH >4% was approximately nine and a half fold higher in anaemic compared to non-anaemic children at presentation (Table 4). Taking together with factors predicting anaemia at presentation, it would appear that by reducing the age band susceptible to malaria at presentation from 6 to 59 months to 6-15 months, ACTs have, during the ten years of adoption as first-line treatments, not reduced the burden of anaemia imposed by a single episode of an apparently uncomplicated infection. We have no ready explanation for a presenting parasitaemia ≤50,000 μL −1 being a predictor of MAFH in excess of 4%. The early-appearing anaemia in 50% of non-anaemic children was clearly attributable to DAFH in excess of 4% (see above, paragraph 3). Therefore, all of the predictors of DAFH in excess of 4% were not unexpected (Table 5). However, there is no ready explanation from the results presented for temperature in excess of 37.9°C being a predictor of DAFH in excess of 4%. In this context, there is need to justify the use of DAFH >4% as cut off for predictors of DAFH. Firstly, this cut off has been used in a previous study [16]. Second, a ≥ 5% (5 units) fall in haematocrit from baseline is unlikely to be a random effect and likely to represent a significant fall from baseline in a manner similar to using 95% confidence interval in a two-way analysis of variance [30].
An important question arising from the present study is whether there are differences in the recovery from anaemia attributable to infections (pre-treatment anaemia) and the anaemia which follows treatment of the infections in patients who were not previously anaemic at presentation (early-appearing anaemia). Overall, it would appear there are significant differences as evidenced by the significantly longer anaemia recovery time in children with anaemia primarily due to the infections compared to the anaemia following treatment of the infections. Additionally, pre-treatment anaemia is primarily due to interaction(s) between host and the parasites while the anaemia following treatment is due to interaction(s) between host, parasite and the drug. A striking feature of all non-anaemic children pretreatment who progressed to early-appearing anaemia were: most had pre-treatment haematocrit close to the lower threshold of normal and the critical falls in haematocrit at which all progressed to anemia was a drugattributable fall in excess of 10% (Fig. 4). Perhaps determining threshold for progression of all those developing early appearing anaemia should form routine part of the clinical evaluation of drug-attributable falls in haematocrit following different drug regimens. The pre-treatment haematocrit threshold in the cohort of children with normal haematocrit at presentation appears to be a haematocrit of 40%.
The ratio of the estimated AUC (DAFH) to value of DAFH was approximately nine folds. This ratio represents the a b relationship of a pharmacokinetic estimate to a pharmacodynamic estimate -two methods of measuring DAFH in the same patients employed in the present study. An intriguing feature of the ratio is: it is somewhat similar to the number of half-times required for >99.9% completion of a pharmacokinetic process in a rudimentary one-compartment pharmacokinetic model. A somewhat more intriguing finding is their insignificant agreement when 9 or 10 multiples of value of DAFH and AUC (DAFH) were subjected to Bland-Altman analyses. The analyses indicate both measurements can be used interchangeably in the same patients despite the difference in their units of measurement. The difference in units of measurement would suggest some further application of Bland-Altman analysis when applied to establishing agreement between pharmacokinetic and pharmacodynamic processes in clinical drug studies.
The main limitation of the present study is not quantifying once-infected red blood cells. Quantification of once-infected red blood cells would have enabled the study to distinguish between non-anaemic children at presentation who progressed to early-appearing anaemia from those who did not. Another limitation is that the study did not evaluate the contribution of the background causes of anaemia, to the factors contributing to anaemia pre-and post-treatment initiation.

Conclusions
In conclusion, after ten years of adopting of ACTs as first-line antimalarials in Nigeria, anaemia is common pre-and early post-treatment, anaemia burden of a single infection is relatively high, and drug-attributable falls in haematocrit >4% is common and significantly lower in anaemic compared to non-anaemic malarious children.