Risk factors of chloroquine resistance in Plasmodium falciparum malaria

ELSEVIER

ACTA TROPICA Acta Tropica 61 (1996) 293-306

Risk factors of chloroquine resistance in Plasmodium f alciparum malaria Folke I. Hess a.,, Aniello Iannuzzi d, Judson Leafasia c, David Cowdrey b, Hans D. Nothdurft a, Frank Von Sonnenburg a, Thomas L6scher ~, Karl H. Rieckmann b a Department of Tropical Medicine, University of Munich, Leopoldstrasse 5, 80802 Munich, Germany b Army Malaria Research Unit, MILPO, Ingleburn, NSW, 2174, Australia c Solomon Islands Medical Training andResearch Institute, P.O. Box 349, Honiara, Solomon Islands d Royal Prince Alfred Hospital, Camperdown, NSW, 2050, Australia Accepted 11 March 1996

Abstract Objective: To identify patient-related risk factors of chloroquine resistance. Design: A case control study. Subjects: Plasmodiumfalciparum infected school children were followed prospectively for 7 days for the detection of chloroquine resistance. Cases were 38 individuals with chloroquine resistant infections. Controls were 125 individuals with chloroquine sensitive infections. Cases were compared with controls with respect to previous or current study factor levels. Subjects were recruited from randomly selected schools which were stratified for area. Study location was in North Guadalcanal, Solomon Islands. Outcome measure: Treatment failure of chloroquine in standard dosage (25 mg/kg). Follow-up period was 7 days. Results: Logistic regression resulted in 5 independent significant predictors of chloroquine resistance, obtained simultaneously with the diagnosis of malarial infection: (i) Young age (odds ratio (OR) for age < 7 years: 7.1; 95% confidence interval (CI): 2.5-25.0; OR per year increase after the age of 5 years: 0.8; 95% CI: 0.6-0.9). (ii) High parasite density (OR for > 1000/#1: 5.0; 95% CI: 2.0-10.6; OR per 500 parasites/#l increase: 1.3; 95% CI: 1.1-1.7). (iii) Normal spleen size (OR: 4.0; 95% CI: 1.5-10.8). (iv) Malnutrition (OR: 4.9; 95% CI: 1.8-13.2). (v) Presence of gametocytes in the thick smear (OR: 3.0; 95% CI: 1.1-8.0). Conclusion: The identified risk factors are easily measureable without special equipment. They may be useful for health workers in the Solomon Islands, even in remote areas, to identify Plasmodium falciparum infected individuals at high risk for chloroquine resistance before a treatment decision is made.

* Correspondingauthor. TEL."+49-89-21803517; Fax." +49-89-336112. 0001-706X/96/$15.00 Copyright © 1996 Elsevier Science B.V. All rights reserved PH S0001-706X(96)00011-3

294

F.I. Hess et al./Acta Tropica 61 (1996) 293-306

Keywords: Plasmodiumfalciparum; Chloroquine resistance; Predictor 1. Introduction

Simple clinical criteria, such as fever, are important for diagnosing malaria in highly endemic areas, especially in the absence of microscopy. The use of antimalarial drugs is often based on clinical judgment alone. Apart from being potentially harmful, the liberal and inadequate use of antimalarial drugs may promote the development of resistant strains. However, clinical criteria that help to decide which drug is to be used are not readily available. Neither the presence of parasites nor symptoms during the first days after treatment unambiguously indicate treatment failure. Attempts have been made to find a simple case definition of treatment failure (Rieckmann, 1990). However, the WHO standard in vivo resistance test requires a follow-up of the treatment response for several days. It would be desirable to predict treatment failure at the same time when malaria is diagnosed. This was the aim of our study. Prediction of drug resistance could help avoid the use of ineffective drugs, optimize treatment strategies and minimize costs. While prognostic factors for poor outcome of malarial infections have been studied extensively (Krishna et al., 1994; McElroy et al., 1994), only a few studies examined risk factors for drug resistance. Predictors of drug resistance identified by other investigators (Fontanet and Walker, 1993) were: young age, high parasitemia, low hemoglobin level associated with a history of treatment failure, history of several recent malarial attacks, and diarrhea following treatment. However, only one-third of treatment failures occurred in patients with one or more of these risk factors. The remaining risk was attributed to the intrinsic drug resistance of the parasite; however, it could also be due to as yet unidentified host factors. The clinical response to infection is mainly determined by the immune status of the infected individual. Even infection with a resistant strain does not necessarily cause symptomatic disease. Decreased drug susceptibility of the parasite, low immunity and subtherapeutic drug levels in the patient are known causes of treatment failure in Plasmodium falciparum infections. Unfortunately, protective immunity against malaria is difficult to assess by means of laboratory tests. Determination of drug susceptibility of the parasite using in vitro techniques will not be very useful for a rapid clinical judgment under field conditions, since in vitro methods are time consuming and require special equipment. Furthermore, in vitro results are often poorly associated with the in vivo outcome. Finally, drug levels are not usually measured in the field. To keep the assessment of risk factors of drug resistance as simple as possible, it was decided to focus on host-related variables. Therefore, the relationship between treatment failure and easily obtainable clinical, behavioral and demographic information, and medical history was examined.

F1. Hess et al./Acta Tropica 61 (1996) 293-306

295

2. Material and methods

2.1. Study area and population Clinical observations suggested that resistance against chloroquine is occurring in the Solomon Islands. There, chloroquine is still the mainstay of malaria treatment. Therefore an area in North Guadalcanal, 10 km west to 15 km east of Honiara was chosen. Its estimated population is 40,000 and has a high incidence rate of P. falciparum malaria, which is the dominating malaria species in this area (Nukuro et al., 1989; Kere et al., 1993). The population of the study area is predominantly Melanesian, with an admixture of Poly- and Micronesian.

2.2. Spectrum of clinical malaria seen in the study area The status of malaria in the Solomon Islands is different from province to province, with highly malarious areas and areas of low incidence. The vast majority of malaria patients in the Solomon Islands suffer from uncomplicated disease, with a variety of symptoms, including fever, headache, malaise, cough, nausea, vomiting and diarrhea. However, in adults and even in children infections can be asymptomatic, usually depending on the parasitic load. Progression from mild to fatal malaria seems to be rare, since the majority of cases seen at clinics or hospitals in our study area, which is highly endemic, are generally young children in whom immunity has not yet developed and where self-medication has failed. Severe disease, presenting as cerebral malaria, hypoglycemia and anemia, is commonly seen in those children. Deaths occur in the very young, in some cases even after third line treatment with quinine intravenously. However, also one adult failed to respond to quinine (Leafasia, 1994).

2.3. Inclusion of study participants During October and December 1994, 188 school children, aged 5-14 years, at 6 randomly selected schools, were identified with uncomplicated P. falciparum monoinfection. Of these mainly asymptomatic children, 163 were included in the study.

2.4. Study procedure The study area was stratified into west, central and east Honiara. Six randomly selected schools (two in each stratum) were visited. Informed consent of the school principals was obtained prior to the visits. At each school, all children of randomly selected classes were screened for P. falciparum infection. A Giemsa stained thick smear of capillary blood from a finger prick was performed, and the spleen was examined. All children included in the study were weighed, and a questionnaire was administered.

296

F.L Hess et al./Acta Tropica61 (1996) 293-306

2.5. Treatment

Chloroquine, the drug of first choice in the study area, was given orally in standard dosage of 25 mg/kg over 3 consecutive days (10 mg; 10 mg; 5 mg) to all P. falciparum infected children. Inspection of the oral cavity was obligatory. Unswallowed and regurgitated tablets were immediately replaced with a new full dose. In the case of chloroquine resistance a single dose of pyrimethamine/sulfadoxine was given according to weight. 2.6. Follow-up

Follow-up visits were done on days 1, 2, and 7 (Rieckmann, 1990). Follow-up to the homes of participants was done if they were absent. A thick blood smear was taken and the participants were questioned for symptoms and side-effects. 2. 7. Questionnaire

The questionnaire was developed after interviews with local health staff and the field workers. Mainly dosed structured questions were asked face to face by local staff. The language was Pidgin English. Pilot testing of the questionnaire was conducted with a subset of 12 children covering the whole age range. The accuracy and consistency of children's answers were validated by using some questions repeatedly with different wording. Inconsistencies were clarified on the follow-up days. The questionnaire included demographic and socio-economic data, behavioral characteristics regarding protective measures, and the recent medical history. 2.8. Microscopy

The microscopic examination of thick smears, stained with Giemsa, were performed blind. Results were checked by a second microscopist. Parasitemia was calculated against 100-200 leukocytes, assuming an average leukocyte count of 8000/pl (Greenwood and Armstrong, 1991). If no parasites were seen on day 7, the microscopist tried to examine the blood film over an area covering 1000 leukocytes. 2.9. Outcome

A simplified in vivo test schema according to Rieckmann (1990), based on the WHO standard field test, was conducted to assess the response of patients to treatment with chloroquine. Parasite densities were determined on two occasions only - days 2 and 7 - after onset of treatment. Resistant cases of P. falciparum infection were defined as parasite density of more than 25% of the pre-treatment level on day 2, or presence of parasites at day 7 after standard treatment, irrespective of whether symptoms had cleared or not. Recurrence of parasites within 28 days after treatment (RI-late) was not considered. Controls were infected children with

F.I. Hess et al./Acta Tropica61 (1996) 293-306

297

a negative blood smear at day 7 and a parasite reduction under 25% pre-treatment level on day 2. 2.10. Laboratory investigations

Capillary blood (100-200/~I) was taken for hemoglobin and antibodies before treatment. An immunofluorescence test, using cultivated parasites as antigen to measure antibodies to the whole P. falciparum parasite (Voller and O'Neill, 1975), was performed from 50/tl of dried whole blood spots on Whatmann filter paper at the Department of Tropical Medicine, University of Munich. For drug levels, 50-100 #1 plasma (centrifugal force 2000 x g for 10 min within 2 h after sampling) was collected before treatment and on day 7. Drug levels were measured by bioassay at the Army Malaria Research Unit, Ingleburn, Australia (Kotecka and Rieckmann, 1993). Hemoglobin concentration was measured from 20/A whole blood added to 5 ml Van Kampen-Zijlstra's reagent (cyan-methemoglobin method) using an Erma hemoglobinometer. Spleen size was assessed according to Hackett's classification (Bruce-Chwatt, 1985) in upright position. 2.11. Nutritional status

The nutritional status was assessed by weight per age using growth tables for 2-18 years from the National Center for Health Statistics percentiles (Hamill et al., 1979). Being under the fifth percentile was the criterion for malnutrition. 2.12. Socio-eeonomic status

After discussions with community leaders and members of the field team, two classes of socio-economic status were defined. The higher class was represented by children whose father's profession was either businessman, plantation or transport owner, or civil servant. The lower class was represented by children of fishermen, laborers, farmers, or whose fathers were unemployed. 2.13. Statistical methods

In order to assess expected differences of exposure to each study factor and its relation to treatment failure, 2 x 2 table analysis was done for each variable one at a time. The odds ratio was determined, and statistical significance was established by means of the Mantel-Haenszel chi-square test. 95% confidence intervals of the odds ratios were calculated by means of the Cornfield test. For ordered nondichotomous and continuous variables which were categorized, trends were assessed using chi-square. In the logistic regression procedure age was analyzed as a continuous variable. Finally, all potentially important risk factors and interactions were

298

F.L Hess et al./Acta Tropica61 (1996) 293-306

assessed together by means of logistic regression to allow for adjustment for each other.

3. Results O f 188 P. falciparum infected children, 25 were excluded from the study. Thirteen had a history of chloroquine intake within the last 7 days, 4 children had serum samples which were too small to investigate drug levels, 3 were non-compliant during the questionnaire, 4 were lost to follow-up, and 1 vomited within 1 h of treatment. The excluded children did not differ significantly from those included in their baseline characteristics (i.e., sex, mean age, parasite density index, spleen rate and size, ethnicity, living area, socio-economic and nutritional status and symptoms). O f 163 included children, 125 responded well to chloroquine treatment. The remaining 38 children (23%) were chloroquine resistant at different levels: 3 R I I I , 17 RII, and 18 early R I resistance cases. In no case did the intensity of symptoms and the clinical condition during the follow-up period require immediate second line treatment before day 7. Baseline characteristics of the 163 participating P. falciparum infected children are summarized in Table 1. This table demonstrates that the 25 excluded children

Table 1 Baseline characteristics of the study population Characteristic

Included subjects (n= 163)

Excludedor lost subjects (n =25) a

Male Female Mean age Parasite density Mean Index Gametocytes present Spleen rate Average spleen size (Hackett) Symptoms at day 0 Ethnicity Melanesian Polynesian Micronesian Socio-economic status: low Living area West Honiara Central Honiara East Honiara Malnourished (low weight for age)

53% 47% 9.15

56% 44% 9.35

936//A 4.48 16% 74% 1.37 27%

868/ttl 4.26 18% 68% 1.27 24%

82% 11% 7% 41%

76% 8% 16% n.d.

46% 19% 35% 28%

54% 18% 28% 32%

aExcluded cases do not differ significantlyfrom included cases in these characteristics (used significance tests: t test for means of continuous variables, and X2 test for discrete variables).

El. Hess et al./Acta Tropica 61 (1996) 293-306

299

did not differ significantly in age and sex distribution, ethnicity, parasite densities, spleen rates, symptoms, and nutritional status.

3.1. Adverse effects Children were generally used to the bitter taste of the tablets. Only one child vomited within 1 h of treatment and was excluded from the study, but was subsequently observed until parasites cleared. Two of the 18 symptomatic cases at day 2 had not been symptomatic at day 0. These 2 cases had minor gastrointestinal symptoms, which could relate to chloroquine intake. Both children were symptomfree at day 7. On the other hand, only 4 of the 21 symptomatic cases at day 7 had been symptomatic throughout the study period. Another 4 became symptomatic at day 7, after a symptom-free interval, and 13 children reported symptoms for the first time at day 7. Adverse effects of the drug, although unlikely, could also not be ruled out here. However, in none of the cases were symptoms severe and we were not urged to take any further action.

4. Analysis 4.1. Demographic characteristics Cases and controls had similar sex and ethnic distributions (odds ratio (OR) for male vs. female: 1.0; 95% CI: 0.5-2.2; OR for Melanesian vs. other: 0.7; 95% CI: 0.3-1.8). Their socio-economic status (OR for low status: 1.4; 95% CI: 0.6-3.1) and residential histories, regarding rural-coastal distribution (OR for rural: 0.8; 95% CI: 0.4-1.9), were also similar. Compared with control subjects, cases were more likely to be found west of Honiara (OR: 3.2; 95% CI: 1.2-9.4) than central or eastward, but this was not significant in a logistic regression analysis which allowed for other prognostic variables. The age distribution of cases and controls was significantly different, with cases more likely to be younger (P=0.04; Student's t test, comparison of the means). An age younger than 7 years was significantly associated with treatment failure and remained significant after adjustment for confounders by logistic regression (Tables 2 and 3).

4.2. Clinical characteristics and medical history Compared with children who were symptom-free, the unadjusted risk of resistance was more than 3-fold for children that reported any symptoms (OR 3.4; 95% CI: 1.4-8.4). Symptoms which were generally of low or moderate intensity were, however, not significant when assessed together with other variables. An enlarged spleen was more often seen in individuals with chloroquine sensitive infections (OR: 0.5; 95% CI: 0.2-1.2). Although not significant in the 2 × 2 table

F.I. Hess et al./Acta Tropica 61 (1996) 293-306

300

analysis (Table 2), spleen size was included in the multivariate assessment, due to its relation to immunity and its importance for the clinical diagnosis of malaria in many areas. Interestingly, after multivariate adjustment, a normal spleen size became a significant predictor of resistance (Table 3). There was a higher proportion of treatment failure in children of lower weight in agreement with the findings regarding age (OR for < 2 0 kg: 3.2; 95% CI: 0.9-10.7). Malnutrition indicated by "low weight for age" was also significantly associated with a higher risk of chloroquine resistance (Tables 2 and 3). A history of more than two malaria episodes, recalled by 105 out of the 126 who answered this question, did not significantly reduce the risk of resistance (OR: 0.6; 95% CI: 0.2-2). A history of fever during the last 4 weeks was also not associated with resistance (OR: 2.0; 95% CI: 0.8-5.0). Drug intake during the last 4 weeks, excluding the last 7 days, increased the risk of treatment failure significantly (OR: 2.4; 95% CI: 1.1-5.6). However, it was difficult to determine whether antimalarials or other drugs had been taken, After allowing for the design effect and for confounders, drug intake was no longer significant.

Table 2 Mantel-Haenszel 2 x 2 table analysis of potential risk factors a of chloroquine resistance Variable

Cases resistant

Controls sensitive

Relative risk (odds ratio)

95% CI

P value

11 26

10 115

4.7 1.0

1.6-13.5

0.0007***

19 19

33 92

2.8

1.2- 6.3

0.006**

19 19

27 98

3.6 1.0

1.6- 8.4

0.0007***

10 28

16 109

2.4 1.0

0.9-6.5

0.05

Absent

Spleen size Normal Enlarged

13 24

28 97

1.9 1.0

0.8-4.5

0.1

Age < 7 years > = 7 years

Parasite density > 1000/#1 < = 1000/#1

Malnourished Well nourished Gametocytes Present

*Significant at P<0.05, **P<0.01, ***P<0.001. SE = standard error. aOnly those variables are shown which were significant in the subsequent logistic regression analysis after adjusting for each other. Variables analyzed but not significant in the logistic regression analysis included: sex, ethnicity, socio-economic status, rural/urban distribution, living area, presence of symptoms, number of past malaria episodes, history of fever, history of drug intake in last 4 weeks, self-medication, use of bednets, repellents or other protective measures, hemoglobin and antimalarial serostatus.

F.1. Hess et al./Acta Tropica 61 (1996) 293-306

301

Table 3 Logistic regression analysis of significant risk factors of chloroquine resistance Variable

95% CI exp (fl ± 1.96 x S.E.)

P value

0.67

0.04- 0.4 2.6 -25.0 0.6 - 0.9

0.001"** 0.001"** 0.01"*

0.7

0.3

0.2 5.0 1.3

0.09- 0.5 2.0 10.6 1.1 - 1.7

0.001"** 0.001"** 0.007***

Malnutrition

1.6

4.9

0.73

1.8 -13.2

0.0008***

Gametocytes present

1.1

3.0

0.42

1.1 - 8.0

0.04*

Spleen size normal

1.4

4.0

0.62

1.5 -10.8

0.007***

Age > =7 years <7 years Per year increase Parasite density < = 1000//~1 > 1000/#1 Per 500//~1increase

Parameter estimate fl~

Relativerisk (odds ratio) =exp (fl)

Stand. coeff.

-2.0

0.14 7.1 0.8

-0.3

-

1.5

ilia

*Significant at P<0.05, **P<0.01, ***P<0.001. SE = standard error. aEstimates are those obtained from the final model in a stepwise logistic regression analysis carried out by backwards elimination in SAS.

4.3. Behavioral characteristics Questions concerning the use o f protective measures were not always adequately answered, especially by y o u n g e r children. The achieved sample sizes in these categories, except that for self-medication, were too low to give e n o u g h power to reveal statistically significant effects. However, there was some indication that the practice o f self-medication was associated with a lower risk o f resistance (OR: 0.4; 95% CI: 0.2-1.0). Self-medication was assumed when children reported that they were sometimes given "bitter medicine" without having been to a clinic. 4.4. Laboratory data M e a n parasite densities o f resistant and sensitive children differed significantly (Student's t test with logarithmically transformed parasite densities, P = 0 . 0 0 0 1 ) . With increasing parasite density at diagnosis, the risk o f resistance increased significantly (Tables 1 and 2). Also the presence o f gametocytes, irrespective o f their concentration, indicated treatment failure (Tables 2 and 3). The hemoglobin concentration was equally distributed in cases and controls (Student's t test o f the means; P = 0 . 1 ; O R for H b < 10 g/dl: 3.7; 95% CI: 0.6-19.5). Also the a n t i b o d y serostatus against bloodstage parasites, measured by immunofluorescence a n t i b o d y assay, was

302

EL Hess et al./Acta Tropica 61 (1996) 293-306

not associated with treatment failure or success (OR for negative sero-status: 1.2; 95% CI: 0.4-2.9).

4. 5. Effect modification Effect modification, i.e., a difference in the effect of one factor according to the level of a second factor on the risk of treatment failure, was assessed by stratifying the subjects into groups which were biologically or socio-medically meaningful. With the Breslow-Day test for homogeneity of the odds ratios 3 significant interactions were identified: (1) Nutritional status with socio-economic status (chi-square: 4.35, 1 df, P=0.04). (2) Socio-economic status with symptoms (chi-square: 4.5, 1 df, P=0.034). (3) Hb level (categorized < / > 1 1 mg/dl) with spleen size (chi-square: 8.88, 1 df, P=0.003). Low socio-economic status in malnourished children was a risk factor in this study (odds ratio: 2.9; 95% CI: 1.0-8.0), whereas in well-nourished children a low socio-economic status was not a risk factor (odds ratio: 0.5; 95% CI: 0.2-1.9). The presence of symptoms in children of lower socio-economic status was not a risk factor (odds ratio: 1.3; 95% CI: 0.4-4.4), whereas in children of higher status symptoms were strongly associated with treatment failure (odds ratio: 7.4; 95% CI: 2.5-21.7). Finally, an enlarged spleen was only associated with a lower risk of resistance when the child had a normal hemoglobin level (Hb > 11 g/dl), but not in anemic children. None of the interactions was significant at a P level < 0.01 in the final multivariate assessment.

4.6. Logistic regression To adjust for all possible confounders and effect modifiers at the same time, logistic regression was performed as described in the Methods section. Table 3 shows the remaining significant variables. A young age and a high parasite density were significant risk factors, irrespective of being analyzed as categorical or continuous variables.

5. Discussion

We report here the results of a 7-day follow-up of 163 children treated with chloroquine for uncomplicated P. falciparum malaria. The risk of treatment failure was 23% (95% CI: 20-26). It should be emphasized that although the outcome of treatment was prospectively assessed, this study cannot establish causality. The study estimates the likelihood of an association between some variables and chloroquine resistance. Nonetheless, this study, along with another study (Fontanet and Walker, 1993) on predictors of treatment failure, provides sufficient evidence for the assertion that a young age and

F.1. Hess et al./Acta Tropica 61 (1996) 293-306

303

a high parasitemia are important risk factors for treatment failure. Taking into account that our study population were mainly asymptomatic children, whereas the study population in the above-mentioned study (Fontanet and Walker, 1993) consisted mainly of adult outpatients, these consistent findings are convincing. Reasons for this are clear, as immunity develops with increasing age (Nosten et al., 1991; Baird, 1995) and a high parasite density takes longer to be reduced. Our findings clearly indicate malnutrition as a predictor of treatment failure in the study area. In view of the controversial literature on nutritional status and malaria (Edington, 1967; Hendrickse et al., 1971; Edwards, 1987), this significant result is noteworthy. McGregor (1982) postulated a protective effect of severe malnutrition against malaria during childhood. Moderate malnutrition is highly prevalent in this region (Eason, 1986), but we did not see severely malnourished children. We found 28% of infected children and 50% of resistant cases to have a low weight for their age. Malabsorption of chloroquine could be excluded in all 38 resistant cases by investigation of post-treatment drug levels (ranging from 50 to 820 ng/ml). The protective effect of an enlarged spleen is possibly related to immunity. An enlarged spleen in this area is clearly the consequence of repeated exposure to malaria parasites. Anemia modified this effect in our study when analyzed without multivariate adjustment. In anemic children (Hb level < 11 g/dl), a large spleen was no longer protective. This may indicate more aggressive disease or the presence of a more virulent Plasmodium strain in these children. Finally, the presence of gametocytes indicated a higher risk of chloroquine resistance. This is interesting, since one would expect gametocytes in chronic infections to be in balance with the host. Other investigators (Jones et al., 1990) have reported that not only asexual stages, but also gametocytes, are better suppressed by immune individuals than by individuals with lower resistance to malarial infection. The high rate (73%) of asymptomatic P. falciparum infected children indicated some level of anti-disease immunity among the study population. Only in children of higher socio-economic status was the presence of symptoms strongly associated with resistance. Children in the lower socio-economic classes are possibly more likely to suffer from concomitant diseases responsible for symptoms. The study population were mostly semi-immune, asymptomatic children. Severely sick children are eventually more prone to treatment failure, but were not covered here. Also, late RI resistance cases could not be considered due to the short follow-up of 7 days, but Fontanet and Walker (1993) demonstrated that the rate of failure in children was highest in the first 9 days. Therefore, in order to detect the more pronounced resistance cases and for practical reasons, we preferred a short follow-up period. Treatment of asymptomatic parasite carriers is debatable. Nevertheless, we treated any identified parasitemic child. Studies have shown that intercurrent treatment does not interfere with the development of immunity in holo-endemic areas (Greenwood, 1984). Only infections with very low parasite densities, i.e., less than 100/#I were neglected. Even though pilot testing was conducted, some questions, particularly those

304

F.L Hess et al./Acta Tropica 61 (1996) 293-306

addressing protective behavior and medical history, could not be answered by a number of children. This could be improved in future studies by focussing on these issues only. The classification of socio-economic status based on the father's profession has been done in a pragmatic manner and may not reflect differences that would be found by a sociologist. However, this approach was chosen after having listened to the people from the islands and may therefore be appropriate for this purpose. We had a low proportion (5%) of resistant cases with none of the identified factors. However, it must be emphasized that several other human and parasitic factors, which were not considered, may still play an important role. Factors, such as genetic disorders or concomitant diseases interfering with immunity or drug absorption, could not be covered in this study.

6. Implications Based on the identified risk factors, an area-specific model of combined risk factors to predict chloroquine resistance was developed. The predictive value of a negative test result, i.e., chloroquine sensitivity, ranged from 86 to 97% depending on the sensitivity of the model. This would guarantee that the majority of patients who received chloroquine would be adequately treated when the decision is based on this prediction model. Nevertheless, in at least 3% of cases of chloroquine resistance, this predictive scheme would fail. As no information about the potential progression from mild to severe or fatal malaria was obtained in this study, an immediate application of this prediction model is therefore limited. On the other hand, these children would usually be given chloroquine as first line drug which would have resulted without risk factor evaluation in at least 23% treatment failures. Since this study covers uncomplicated cases only, the results of this prediction model cannot be applied to severely sick or hospitalized patients without additional studies of potential risk factors in such patients. Furthermore, due to the low predictive value of a positive test, ranging from 28 to 50%, the classification of a child as being resistant is not reliable, even if the sensitivity to detect drug resistance with this prediction model would be as high as 98%. This would result in overdiagnosis of resistance and overuse of second line drugs. The potential benefits of such prediction, resulting from an early warning signal on resistance, have to be calculated and weighed against potential disadvantages. They also have to be compared with conventional methods of resistance testing. The concept of prediction could result in a concentration of activities, such as monitoring of the treatment response, in individuals at higher risk of treatment failure than others. This could result in a higher effectiveness of clinical work. An immediate implication for malaria control in the study area is that gametocytocidal drugs will be useful to prevent the selection and spread of resistant strains, because in our study carriers of gametocytes were at particular risk to carry resistant strains. The improvement of the nutritional and socio-economic conditions, which

El. Hess et al./Acta Tropica 61 (1996) 293-306

305

is in itself a priority, might also have some beneficial effects on the clinical management of cases and the spread of drug resistance.

7. Conclusion It is concluded that the identification of further risk factors of resistance, including parasite, immunity and behavior-related factors, would make it possible to develop area-specific risk profiles. Such profiles which would also have to be developed for severely ill patients could be useful for clinicians and primary health care workers to improve the management of clinical cases and for policy makers to improve malaria control.

Acknowledgements We thank the Ministry of Health and Social Services of the Solomon Islands, in particular Ezekiel Nukuro, Jimmy Rodgers and Bernard Bakote (Solomon Islands Medical Training and Research Institute) for approval of the study and for their collaboration. We are also grateful to Dr. Kevin Palmer, WHO, for his support. Special thanks to Prof. Richard Taylor (Department of Public Health, University of Sydney) for many thoughtful comments and to Saefafia Kirimama, Uschi Breitreiner, Barbara Kotecka, and Shirley Bahr for technical assistance. Thanks also to Prof. St~irchler and Mrs. Mittelholzer for their cooperation and financial support. Folke I. Hess is a recipient of a German Academic Exchange Service (DAADSonderprogramm, Epidemiologie) scholarship. This study has also been supported by grants from Hoffmann La Roche, Basel, Switzerland, and by equipment from the Army Malaria Research Unit in Ingleburn, Australia. Informed consent was obtained from community leaders and school principals. The study was approved by the Health Research Committee of the Solomon Islands, Honiara. The decision about participation was taken at a communal rather than an individual level because of expected practical limitations (young children, absence of parents at schools, illiteracy, lack of awareness of the implications of participation, tribal obligations, etc.) for informed consent. However, the school principals and teachers were informed in advance and asked to discuss the issue with their community. Thus, permission to conduct the study was obtained through trusted and respected community leaders. This procedure fitted well in the cultural norm of the tribal society in the Solomons. Children and their parents were free to withdraw from the study at any time.

References Baird, J.K. (1995) Host age as a determinant of naturally acquired immunity to P. J'alciparum. Parasitol. Today 11, 105-111.

306

FI. Hess et al./Acta Tropica61 (1996) 293-306

Bruce-Chwatt, L.J. (1985) Essential Malariology, 2nd Edn. Alden Press, Oxford. Eason, R.J. (1986) Childhood malnutritionin the Western Province of the Solomon Islands. Trop. Doctor 16, 135-138. Edington, G.M. (1967) Pathology of malaria in West Africa. Br. Med. J. 1,715-718. Edwards, K. (1987) Childhood malaria: aspects of immunity affecting morbidity and mortality. Papua New Guinea Med. J. 30, 135-141. Fontanet, A.L. and Walker, A.M. (1993) Predictors of treatment failure in multiple drug resistant falciparum malaria: results from a 42-day follow-up of 224 patients in eastern Thailand. Am. J. Trop. Med. Hyg. 49, 465-472. Greenwood, B.M. (1984) The impact of malaria chemoprophylaxis on the immune status of Africans. Bull. WHO 62 (Suppl.), 69-75. Greenwood, B.M. and Armstrong, J.R.M. (1991) Comparison of two simple methods for determining malaria parasite density. Trans. Roy. Soc. Trop. Med. Hyg. 85, 186-188. Hamill, P.V., Drizd, T.A., Johnson, C.L., Reed, R.B., Roche, A.F, and Moore, W.M. (1979) Physical growth: National Center for Health Statistics percentiles. Am. J. Clin. Nutr. 32, 607-629. Hendrickse, R.G., Hasan, A.H., Olumide, L.O. and Akinkumni, A. (1971) Malaria in early childhood. Ann. Trop. Med. Parasitol. 65, 1-20. Jones, T.R., Baird, K., Basri, H., Danudirgo, P. and Emon. (1990) Prevalence of malaria in native and transmigrant populations. Trop.-Geogr.-Med. 43, 1-6. Kere, N.K., Parkinson, A.D. and Samrawickrema, W.A. (1993) The effect of permethrin impregnated bednets on the incidence of Plasmodiumfalciparum, in children of North Guadalcanal, Solomon Islands. Southeast Asian J. Trop. Med. Publ. Hlth 24, 130-137. Kotecka, B.M. and Rieckmann, K.H. (1993) Chloroquine bioassay using malaria microcultures. Am. J. Trop. Med. Hyg. 49, 460-464. Krishna, S., Waller, D.W., Ter Kuile, F., Kwiatkowski, D., Crawley, J., Craddock, C.F.C., Nosten, F., Chapman, D., Brewster, D., Holloway, P.A. and White, N.J. (1994) Lactic acidosis and hypoglycaemia in children with severe malaria: pathophysiological and prognostic significance.Trans. Roy. Soc. Trop. Med. Hyg. 88, 67-73. Leafasia, J. (1994) Workshop: Quinine Resistance in the Solomon Islands. Ministry of Health and Social Services, Honiara. McElroy, P.D., Beier, J.C., Oster, C.N., Beadle, C., Sherwood, J.A., Oloo, A.J. and Hoffmann, S.L. (1994) Predicting outcome in malaria: correlation between rate of exposure to infected mosquitoes and level of Plasmodiumfalciparum parasitemia. Am. J. Trop. Med. Hyg. 51,523-532. McGregor, I. (1982) Malaria: nutritional implications. Rev. Infect. Dis. 4, 798-804. Nosten, F., Ter Kuile, F., Chongsuphajaisiddhi, T., Luxemburger, C., Webster, H.K., Edstein, M., Phaipun, L., Thew, K.L. and White, N.J. (1991) Mefloquine resistant malaria on the Thai-Burmese border. Lancet 337, 1140-1143. Nukuro, E., Seijama, B., Villia, J. and Swillen, L. (1989) The status of malaria in Solomon Islands. In: The 10th South-West Pacific Malaria Meeting, Honiara, 1989. Rieckmann, K.H. (1990) Monitoring the response of malaria infections to treatment. Bull. WHO 68, 759-760. Voller, A. and O'Neill, P. (1975) Immunofluorescence method suitable for large scale application to malaria. Bull. WHO 45, 524-529.