A review of randomized controlled trials of routine antimalarial drug prophylaxis during pregnancy in endemic malarious areas.

A review of randomized controlled trials of routine antimalarial drug prophylaxis during pregnancy in endemic malarious areas P. Garner1 & B. Brabin2 Current global recommendations for routine malaria chemoprophylaxis in pregnant women living in endemic malarious areas are not clear. To assist in policy formulation, the evidence from randomized controlled trials was reviewed. The literature was extensively searched, and studies identified were systematically analysed in relation to outcomes in the mother and the baby. Routine chemoprophylaxis appears to have an effect on antenatal morbid episodes and packed cell volume. There is a trend towards higher birth-weight values in chemoprophylaxis groups, which reached statistical significance in some studies. Evidence of an effect on gestation was only examined in one study. The effects on perinatal and neonatal mortality have only been examined in a few studies, with small sample sizes. The analysis questions whether routine malaria chemoprophylaxis is the best use of scarce resources in developing countries, and suggests that chemoprophylaxis might be targeted at anaemic women and primigravidae. Large controlled trials, with treatment available to placebo groups, are required to test whether routine chemoprophylaxis has advantages over early, effective treatment of clinical malaria. Introduction Although malaria in pregnancy has been extensively researched, debate conceming antimalarial chemo- prophylaxis during pregnancy continues. First and second pregnancies are associated with a higher pre- valence of Plasmodium falciparum parasitaemia in the first half of pregnancy in women living in ende- mic malarious areas (1). Malaria may contribute to antenatal anaemia (2), and slowing of fetal growth, especially in primigravidae (3). Clinical episodes in late pregnancy may cause preterm delivery in non- immune women, although in semi-immune women a few studies suggest that growth retardation occurs more frequently than preterm delivery (3-5). Elimi- nation of malaria, for example, by residual spraying with insecticide, has been associated with an increa- se in mean birth weight (6). The presence of malaria parasites in the placenta is associated with low birth weight (7, 8); and matemal malaria infection, defi- ned as the presence of parasites in the placenta or I Lecturer in Epidemiology and Health Service Management, Health Policy Unit, Department of Public Health and Policy, Lon- don School of Hygiene and Tropical Medicine, Keppel Street, London WClE 7HT, England. Requests for reprints should be sent to this author. 2 Senior Lecturer in Tropical Paediatrics, Liverpool School of Tropical Medicine, Liverpool, England. Reprint No. 5456 peripheral blood during labour, may be associated with a higher perinatal mortality (9). In the past, WHO recommended routine malaria drug prophylaxis throughout pregnancy for women living in endemic malarious areas (10). Although routine chemoprophylaxis is still recommended, it is not always easy to ensure good compliance, even with targeted health promotion activities. Good com- pliance has been achieved in the context of special research studies (11, 12); other studies have demon- strated that good compliance may be difficult to achieve (13). The potential efficacy of chemoprophylaxis with chloroquine is likely to be reduced with the emer- gence of chloroquine-resistant Plasmodium falcipar- um. Daily proguanil is an altemative, and good com- pliance has been reported in studies in Tanzania (Mutabingwa, personal, communication, 1992) and Nigeria (Fleming, personal communication, 1990). However, there is a question over whether a daily schedule would be widely acceptable in other set- tings. Pyrimethamine has few side-effects, but its use is limited by parasite drug resistance. The safety of other drugs, such as Maloprim (dapsone and pyri- methamine) or mefloquine, is less well established; and they are more expensive. While the benefits of chemoprophylaxis in non- immune women are clear (14), the possible benefits in indigenous, semi-immune women are not so obvi- ous. Currently, WHO recommends that "in settings Bulletin of the World Health Organization, 1994, 72 (1): 89-99 © World Health Organization 1994 89 P. Garner & B. Brabin where placental parasitaemia is associated with low birth weight, and an effective antimalaria drug can be provided on a regular basis, chemoprophylaxis may be considered" (15). The ambiguity of this statement reflects the current dilemma of policy- makers balancing efficacy of a potential chemopro- phylactic agent against its potential adverse effects and the problems in delivering it. Some of the effi- cacy of chemoprophylaxis is based on an effect on birth weight: whether this necessarily affects perina- tal or infant mortality has recently been questioned (16). Prevention of preterm delivery, on the other hand, could have an important impact on neonatal mortality.a Before recommending chemoprophylaxis, with whichever drug regimen, to all pregnant women liv- ing in a particular endemic malarious area, clear evi- dence is required of its efficacy. To assist in this assessment, we have conducted a systematic review of the existing evidence derived from randomized controlled trials of chemoprophylaxis. Materials and methods Inclusion criteria. The criteria for inclusion in this review were trials during pregnancy in which an attempt had apparently been made to conduct a ran- domized comparison either between a policy of rou- tine antimalarial chemoprophylaxis and a policy of treatment of symptomatic malaria; or between alter- native antimalarial chemoprophylaxis regimens. Trial identification. Trials were identified through existing literature reviews and a database of perinatal trials. Reference was made to Kramer (3), who reviewed the determinants of birth weight, searching the literature from 1970 to 1984; Brabin (2), who reviewed the risks and severity of malaria in preg- nant women, and documented recent and current field studies; and the Oxford Database of Perinatal Trials, and the Cochrane Pregnancy and Childbirth Group, which maintains a register of published, unpublished, ongoing and planned trials concemed with perinatal health (17). A MEDLINE search from 1966 to 1992 was conducted, searching titles and abstracts for "ran- dom", "malaria", and "pregnant or pregnancy". This process was repeated using EMBASE (1974 to 1992), a general medical database with a high coverage of European joumals. a Slutsker L et al. Neonatal mortality associated with low birth- weight in Malawi. 38th Annual Meeting of the American Society of Tropical Medicine and Hygiene, Hawaii, December 1989. All the journals publishing relevant trials identi- fied using the above methods were then systemati- cally hand-searched for the period 1960-92, except those which had already been systematically search- ed for by staff compiling the Oxford Database of Perinatal Trials. The final trials list was sent to the Special Pro- gramme for Research and Training in Tropical Dis- eases of the World Health Organization to check that all trials had been identified. In addition, citations to the selected trials were identified by the Science Citation Index database (SciSearch) from 1981 to 1992 as a final check. Methods and results from each study were systematically scrutinized. Any loss be- tween the initial cohort of women and those included in the final analysis of outcomes was examined. Where information appeared to have been collected but was not fully reported, the authors were approached for further details where possible. Description of trials. Nine published studies were identified (Table 1) (11, 18-25). Advance informa- tion from one further study was kindly provided by the investigators (26). Additional data from complet- ed studies were kindly provided by Dr A. Green- wood (Gambia) and Dr T.K. Mutabingwa (Tanzania). Setting. All but one of the studies were conducted in Africa (Table 1). Factors indicative of host immu- nity, including the level of endemicity (with respect to parasite prevalence and seasonal variation) were only sketchily reported on. Exposure risk as meas- ured by the entomological inoculation rate was not given in any of the studies. Two of the studies were conducted in urban populations, where endemicity is likely to be low. The studies occurred over some 30 years, during which time resistance of the parasite to a variety of antimalarial drugs developed. A qualitative descrip- tion of suspected and confirmed parasite drug resis- tance in study areas was commented on in all papers published after 1988 (Table 1). Characteristics of participants. All studies were of antenatal attenders. Thus the efficacy of chemo- prophylaxis is only under test in this group by this delivery mechanism. In only one study (11) was the whole pregnant population examined, so that the out- come of pregnant women not attending clinics was available. Methodological characteristics. Random allocation to experimental and control groups was described in six studies. Women were the unit of randomization in five studies, and compound inhabited in the sixth (Table 2). Two studies used alternate allocation; one used a quasi-random design dependent on which 90 WHO Bulletin OMS. Vol 72 1994 Antimalarial drug prophylaxis during pregnancy Cu-i .Ct + I + + + + -oE CD 0 CD( co CO 0 + E 0 0 + + ++ fl 0 0( COL +. + + + + + + + + + .*.-CoC *C >~~~~~~~. ~~ 00 00 0 o~~~~~~0 D c0ZZ ZZ Z Z0 0 0 0 -0 0D -CQ C.) ~~~r C ~ C0C ..o 3.og .0 E .20 >Co) >,Co- CoO CD O CO ~~~~~a)C 0C Co CCo OCco Cs C co- coC oC 2 C53 0aC) 0 0..- COCO~~L- co -~E ECoC 0) a). ED ~~~ C~~( CoO 0C0C 0 C oECc0 ci 0 CO CIO0 CoCO CCoCCU -CO () C >Coci E >, Co mIv D (/)Wu I 0 0)~~~~~~~~~~~~~~~~~~~~c 0 C (U .z 0 0 0 C0. CO C E C C(0 V ~~~~~~~~~~~ C co 0 N 0) C ~~~ _ 0. CO~ 0. (D cx.coCOC Co COL ~~~0 -~~~~c CoC0CO:3 j S r 0CO Cj - Ci * (D0) O CD O 0 0 3C I- 0 E0 co C 0 -No V ~~~~~50 z z (zm - - .2 co~~~~~~~~~~~~~~~c 0O iz 0) C Eo ~~~~~~~~~o vY f 0. CO co 0)0) 00~~~ 0 '-C0)0 ME C>. . * C0 '- 0 U 0 CEY) CO09.-- C_C 0 -CO 0) 0 z 0CO - CID LL (Z 0 D co F-C) .C 0 CD m WHO Bulletin OMS. Vol 72 1994 91 P. Garner & B. Brabin clinic women were referred to. Loss or exclusions from the initial cohort that were entered into the trials varied from 60% to 3.5% (Table 2). Three studies appeared to be double-blind (19, 21, 26). In the study by Greenwood et al. (11), the study super- visors were aware of the status (whether control or placebo) and village dispensers may have deduced this from the different colour of placebo and active drugs. Only two studies (11, 18) reported on the duration of chemoprophylaxis; in particular, the pro- portion of women taking chemoprophylaxis for short periods was not usually noted, and this is an impor- tant factor when considering potential drug efficacy. Comparison of interventions. Nine studies com- pared malaria chemoprophylaxis (either chloroquine, pyrimethamine, proguanil, Maloprim or mefloquine) with a placebo prophylactic drug or nothing. The other study compared a regimen of proguanil with chloroquine. The dose of antimalarial drugs used in chemo- prophylaxis may be insufficient to clear a current parasitaemia, and a therapeutic dose of antimalarial drugs at the start of chemoprophylaxis is often rec- ommended. However, this regimen was only fol- lowed in three studies (Table 2). Only two papers describe access to treatment in the event of a malar- ial illness (23, 26), although it seems likely that in all the other studies subjects had access to treatment for illness. Oral iron and folic acid supplements were used in some studies (Table 2). Various combinations were used in both antimalarial and comparison groups. Hamilton et al. (20) compared chloroquine prophylaxis plus ferrous sulfate with ferrous sulfate alone and ferrous sulfate plus folic acid. Two studies of primigravid Hausa women in Nigeria (19, 21) used similar study designs, comparing placebo with a single dose of chloroquine followed by proguanil prophylaxis. Each experimental group was divided into four subgroups: proguanil alone, proguanil plus iron, proguanil plus folic acid, and proguanil plus folic acid and iron. Outcomes measured. The outcomes of each study were examined first with respect to the mother, which were classified as parasitaemia prevalence, prevalence of anaemia, incidence of morbidity, and mode of delivery. Second, the outcomes with respect to the baby were examined, which were placental infection, birth weight, gestation, and perinatal, neo- natal or infant mortality. Statistical methods. Weighted averages were calcu- lated when necessary. In the two studies where stan- dard deviations for birth weight were not provided (22, 23), an estimate of 500 g was used. Odds ratios with 95% confidence limits were calculated between antimalarial and control groups for each study for each outcome using standard methods for dichoto- mous and continuous data (27). Calculations were checked using meta-analysis software. Results Maternal effects Parasitaemia. All studies that examined antenatal parasitaemia showed a significant effect of chemo- prophylaxis in reducing the parasitaemia prevalence. This effect was observed even in multigravidae, where it was examined in two studies. Haemoglobin/packed cell volume. Four studies examined haemoglobin or packed cell volume values during late pregnancy (Table 3). Two studies showed a positive effect with all parities combined. Green- wood et al. (11) showed a positive effect in primi- gravidae, but not multigravidae. One study, not included in the Table, examined haemoglobin values in primigravidae postnatally; no statistically signifi- cant difference in anaemia prevalence was found (20). Morbidity. Only two studies examined clinical mor- bidity systematically. Morley et al. (23) found fewer complaints of fever in the experimental group, and Mutabingwab found fewer clinical episodes of fever in both primigravidae and multigravidae. Mode of delivery. Obstetric outcome was examined in four studies (Table 3). Interventions possibly indi- cating cephalopelvic disproportion (Caesarian sec- tion, forceps delivery, and craniotomy) were exam- ined by treatment and control groups. In three studies, chemoprophylaxis was not statistically sig- nificantly related to the incidence of interventions. In the study by Harrison et al. (21), drug prophylaxis with proguanil was associated with a lower risk of obstetric intervention. However, some women in the proguanil group also received iron and/or folic acid, and in these women growth (height) also increased during pregnancy. The nutritional supplementation (not the chemoprophylaxis) appeared to assist mater- nal growth in pregnancy, and to reduce the risk of abdominal delivery. There was no difference in obstetric intervention between placebo alone (4/14) and proguanil alone (2/23), but the numbers were small. b Mutabingwa TK. Studies on malaria chemosuppression during pregnancy in Tanzania (PhD thesis). Amsterdam, University of Amsterdam, 1993. 92 WHO Bulletin OMS. Vol 721994 Antimalarial drug prophylaxis during pregnancy L) co CD CD C0 co cv,_ o N- N1- cN ctv c LO (0 LO CM *oc N _ C\(°CMco 0 C(0 co DN1 (0 Y CO) 0)a) a -O 0a0 a) N (0 0 c () - (0 Z _ _-co Z CXO o 0 - U) a) U 1 ta) 0. C ): c 0 0WC 07 a U) :3 - C -~~0 : -z 75 00a ) z o o,, 2 o N-Cv, 0 r-LO t 0 LO CV) (D 0) LA) LO _ a) c 0 z *0 C co a) 0 0 V LL. 0 CM C) LO 0Y) (0 N 0) CIv, Cf) N D co CM aw 0 Ea) V0 Ca .-..O .-CcC : _ Co< C% CT .2a) C2L0.S.. 0 3 'a . Ca a) - E a) -O 00o L c o o L ._- CO U ._- C Ca C_E ~~< 8 -o E* E v C D Ca C CIS 0' 0~ Ca co * 0 o -0 *- c a) E0 a' 0) 0) EUc) E. oD (D o: a)a.0E a)3 .COE 2 2.7>~a.- 3.Ca)a)a) al)a a0 o, 0 0 0 *0. --0 a) a ,.- '0 '0ca C cu M -C - 0.0~~ 0U)g 0o.a) C.CCE 0.C 0u Ca0 n0 a) c, ca Ea 0 a ir 0 OD@ - C CU C Ucc 0 C<ct 0 10.0 CO t CIO Aor-CDsuc c O , cn _ a=,c .C r<3 *=Lo <: Iua:-. C >1 0 -' E *E 0 Ca co 2 I: 2 C 0 U1) .I Ic a)a -o ._0~ 0) 0ao- .0Ca-co C -c E _0 air Co 0 aC 0)a)8 0.- cn cn m a) Ca ( In 0) C Ch E iL 0 z -o C 0 ._ E 0 .- .10 C CU 0 0a)C -- c 0 cc :3 CD C a E 0 0) a) -0 0 0 a) a) c-I a) a)a) ui co U 0 w) c o a) Z >- 0. >- EC~~~ CE . c co 0e co (, (D= 0) 0 C CD8 En .Ca la m °<0 I cn 0 C rt (D 0 In a) c 0)c a C c a) a)a)-C 0) C TD Cla)a) U)CU, .E . 2 0Da) C-0 ~CM ICO) :l coC a1) c .) U) co 0 = -0O C Un L- a) c a) Ca a1) C CIO 0) c c C a) CaOCr - U)Ca 0 : z 0) Z WHO Bulletin OMS. Vol 72 1994 rN 0Cj) -@ C n O. cc (D 0 0. 0 c a) - U- *0 'a :0 0a0 0- CA 0 U) E0c 0 oCL Ul) 0. 0 0) RCA c ._ c0. C. Ca a) E a) cU .0 C6 E a)0 I -i E c r_ = a) .2 0 - Z Q L :3 0. C 0 N U) E )Eo0 0 Ca Ca: 0 *C O3 0a)0- C >1 a) cn'a M'O COC cn co0 :3 *0 la 0'a) E 0a- .c la la .3. cq g- Cvi a) 0) a)C) a1) Ca U1) a/) U) a) C a1) Ca C U) a) :3 0)i- QO 93 P. Garner & B. Brabin Table 3: Effect of prophylaxis on the mother Antimalarial Comparison Odds ratio, prophylaxis group: group: or difference Effect and study occurrences/total, occurrences/ in the means (first author) Factor measured Groups or mean (n) total, or mean (n) (95% Cl) Maternal parasitaemia: Fleming Parasite rates 28 weeks gestation 3/137 9/36 0.03 (0.01, 0.13)8 36 weeks gestation 21106 5/22 0.02 (0.00, 0.14)8 Greenwood Parasite prevalence Primigravidae 4/21 5/13 0.38 (0.08, 1.78) (last trimester) Multigravidae 9/120 21/103 0.33 (0.15, 0.72)8 Nahlen Parasite prevalence after Primigravidae 6/23 6/22 0.94 (0.25, 3.48) clearance Mutabingwa Parasite prevalence Primigravidae 23/30 37/39 0.21 (0.05, 0.84)a Multigravidae 62/86 61/68 0.34 (0.15, 0.74)8 Nosten Women infected at least All 5/153 37/148 0.16 (0.09, 0.32)a once Packed cell volume (PCV): Hamilton PCV at last antenatal All 37.0 (751) 36.6 (1095) +0.36 (0.01, 0.71)a clinic visit Martin PCV at 40 weeks gestation All 34.5 (40) 32.1 (37) +2.4 (0.16, 4.64)a Greenwood PCV last trimester Primigravidae 30.1 (21) 26.6 (11) +3.5 (0.70, 6.3)a Multigravidae 30.7 (126) 30.4 (118) +0.3 (-0.71, 1.31) Nosten PCV at term Primigravidae 34.4 (43) 32.0 (43) +2.4 (0.98, 3.82)a Multigravidae 31.5 (128) 32.3 (125) -0.8 (-1.70, 0.1) Antenatal morbidity: Morley Fever incidence during 21/119 45/108 0.31 (0.18, 0.56)a pregnancy Mutabingwa Clinical malaria Primigravidae 20/30 36/39 0.19 (0.06, 0.64)a Multigravidae 52/86 60/68 0.25 (0.12, 0.51)a Obstetric outcomes: Hamilton LSCSb All parities 60/404 70/614 1.36 (0.93, 1.98) (OK = 62%)C Harrison LSCS for dysproportion, Primigravidae 8/77 5/14 0.13 (0.02, 0.65)a forceps, craniotomy (OK = 40%) Nosten Complicated labour All 9/153 9/148 0.97 (0.37, 2.5) a Significant difference at the 5% level. b LSCS = Caesarian section. c OK = percentage of cohort where obstetric Outcome Known. Greenwood et al. (11) was the only study to report on maternal mortality, with 1/518 deaths in the experimental and 3/531 deaths in the control group, and a further nine women died from the group not reporting for chemoprophylaxis. In all the other studies, pregnancy outcome data were largely depen- dent on hospital delivery for recording outcome. Considering the large number of women lost or excluded from the final analyses (Table 1), any re- sults of maternal mortality are likely to be unreliable. Fetal/neonatal effects Placental parasitization. Two studies examined pla- cental parasitaemia, and showed that chemoprophy- laxis significantly reduced this. The study by Cot et al. (18) demonstrated an effect even in an area with presumed chloroquine resistance (Table 4). Birth weight. Seven studies reported on mean birth weight values (Table 4). In the six studies that included multigravidae, five reported on the overall effect. Although four showed a trend towards higher mean birth weight, this was statistically significant in only one study. One study (20) reached a statistically significant difference when only regular attenders at the clinic were examined, thus only testing the inter- vention efficacy in a particular subgroup of women. Five studies reported on birth weight in primi- gravidae (where an effect would be more readily WHO Bulletin OMS. Vol 72 199494 Antimalarial drug prophylaxis during pregnancy Table 4: Effect of prophylaxis on the fetus Experimental group: occurrences/total, or mean ± SD (n) Control group: occurrences/total, or mean ± SD (n) Odds ratio, or differ- ence in the means (95% Cl) Placental parasitaemia: Parasites present Parasites present Birth weight Birth weight Birth weight Birth weight Birth weight Birth weight Birth weight Dubowitz assessment; less than 37 weeks All All 1/115 19/444 Primigravidae Parity 1-3 Parity 4+ All parity groups Primigravidae All Regular attenders All Primigravidae Primigravidae Parity 1-4 Parity 5+ Primigravidae All All All 18/105 83/437 2770 (27) 2912 (93) 3060 (76) 2954 (196) 2935 ± 480 (114) 3020 ± 597 (464) 3044 ± 520 (60) 3369 (40) 2855 2872 ± 330 (67) 3101 ± 345 (173) 3132 ± 350 (131) NA (106) 2937 ± 452 (594) 2877 + 433 (159) 4/102 2579 (28) 2844 (90) 2821 (78) 2797 (196) 2895 ± 504 (167) 3008 ± 605 (685) 2848 ± 660 (114) 3137 (37) 2723 2726 ± 465 (50) 3074 + 370 (194) 3042 ± 380 (115) NA (113) 2932 ± 467 (554) 2957 + 496 (144) 8/97 0.13 (0.05, 0.33)8 0.24 (0.16, 0.36)8 + 190 (-80, 460) + 68 (-77, 214) + 248 (89, 407)a + 157 (58, 256)a + 40 (-77, 159) + 23 (-62, 108) + 277 (43, 511)a + 232 (-5.31, 460) + 132 (NS)d + 146 (-5.2, 297)' + 31 (-46, 108) + 93(1,185)a + 82 (NS)d + 5.6 (-47, 58) - 80 (-182, 23) 0.47 (0.15, 1.50) Morley Flemingf Greenwood Nosten Infant mortality: Nosten Stillbirth/neonatal death All Perinatal death Stillbirth and perinatal death Stillbirths Primigravidae Primigravidae Multigravidae All All 14/210 11/128 6/74 22/427 11/159 25/144 13/209 5/32 10/65 27/446 4/152 24/144 a Indicates a significant difference at the 5% level. bConfidence limits estimated by us using birth weight standard deviation to be 500 g. cLarge number lost to follow-up for birth weight. d NA = data not available; NS = difference tested by the researchers, and reported as not significant. e Authors corrected for age and season, giving different values: +159 (95% Cl: 8, 310). fIncomplete follow-up. demonstrated given their higher susceptibility to malaria). All showed a trend towards increased birth weight with chemoprophylaxis. Simple analysis of the data using 95% CI of the difference in the means did not show a significant effect in any of these stud- ies. However, in the study by Greenwood et al. (11), an analysis by the authors adjusting birth weight for age and season demonstrated a significant difference between the experimental and control groups. Stratification by parity in two studies failed to detect any effect on birth weight in women of mid- parity. In high parity women, the prophylaxis group had a significantly higher birth weight in both of the studies reporting on this subgroup. Two studies examined the prevalence of low birth weight in primigravidae. In the study by Green- wood et al. (11), there were significantly fewer low- birth-weight infants in the experimental group (pro- phylaxis group, 4/67; control group, 11/50; OR=0.24 (95% CI: 0.08, 0.72)). In the study by Cot et al. (18) the difference was not statistically significant (OR=0.88 (95% CI: 0.64, 1.21)). WHO Bulletin OMS. Vol 72 1994 Effect and study (first author) Factor measured Groups Morley Cot Birth weight: Morleyb Hamilton Martinb Flemingc Greenwood Cot Nosten Gestation: Nosten Perinatal/neonatal mortality: 1.07 (0.49, 2.35) 0.45 (0.12, 1.66) 0.49 (0.17, 1.39) 0.84 (0.47, 1.50) 2.54 (0.90, 7.15) 1.05 (0.57, 1.94) 95 P. Garner & B. Brabin Only one study reports on the prevalence of high birth weight: this showed that two births greater than 4 kg in the prophylaxis group and none in the place- bo group (2/144 (mefloquine), 0/143 (placebo) (26)). In the study by Greenwood et al. (11), the authors kindly provided unpublished data from the study showing that there was no statistically significant difference in the proportion of babies weighing more than 3.5 kg (Maloprim group, 60/430; placebo group, 50/145; OR=1.18 (95% CI: 0.79-1.77)) (A. Green- wood, personal communication, 1990). Thus the evi- dence for an effect on birth weight shows there is a tendency for a small positive effect in experimental groups. The lack of statistical significance may reflect a small effect overall, or an effect which only operates in certain subgroups of women. Gestation. Only one study examined the duration of gestation, despite the fact that preterm delivery is highly predictive of neonatal mortality. The trend was towards fewer preterm infants in the treatment group, but the difference was not significant. The paucity of studies examining gestation probably reflects the difficulty in measuring gestation accu- rately (16). Perinatal/neonatal mortality. Four studies systemati- cally recorded stillbirth and perinatal, neonatal or infant mortality (Table 4). No study detected an effect on mortality. In two studies the trend was towards a protective effect in the chemoprophylaxis groups. The small size of all studies means that none had sufficient power to detect a possible effect. Discussion This review presents evidence that chloroquine, pro- guanil and mefloquine prophylaxis is associated with a reduction in antenatal illness episodes. There is evidence from the controlled studies reviewed here and other work that chemoprophylaxis results in higher mean haemoglobin levels in preg- nant women, that the higher prevalence of parasitae- mia in the first half of pregnancy in primigravidae is associated with an increased risk of anaemia, and that moderate and severe anaemia in pregnant women is associated with poor matemal and fetal outcomes (2). While it is clear that starting chemo- prophylaxis early in the first trimester of pregnancy in primigravidae is likely to be the most effective in preventing anaemia, precisely how early this needs to be for maximal efficacy is not known. In addition, the efficacy of chemoprophylaxis on anaemia in multigravidae has not been demonstrated, although anaemia remains prevalent in these women. The presence of malaria parasites in the placenta is reduced by chemoprophylaxis, even when chloro- quine is used in the presence of chloroquine-resistant malaria, as shown by Cot et al. (18). Although pla- cental parasitization is associated with low mean birth weight in observational studies (7), it cannot be assumed that an intervention that reduces the preva- lence of parasites in the placenta will automatically benefit the baby. An intervention that reduces the prev- alence of matemal anaemia may be more beneficial. This review shows a positive effect of chemo- prophylaxis on mean birth weight in primigravidae and in grand multigravidae, but this is not a consis- tent finding. The effect of malaria on the human host varies in different parts of the world owing to differ- ing biological, immunological and epidemiological circumstances affecting both the host and the para- site: this may account for the variation in the results obtained. When an effect on birth weight has been demonstrated, it is not clear if this is due to an effect on fetal growth or through an effect on the preterm delivery rate. Observational studies suggest that low birth weight in malarious areas is primarily associat- ed with small infants or intrauterine growth retarda- tion, not preterm delivery (4, 5). Preterm births have a higher mortality than term births of a similar weight (28), and evidence of a reduction in the preterm rate would be strong indirect evidence of a protective effect of chemoprophylaxis on mortality. In the absence of gestational age data, future studies should at the very least report on very low birth weight prevalence (less than 1500g) as it is likely to be strongly associated with preterm birth in areas where the average birth weight tends to be low. No study was of sufficient size to demonstrate a difference in perinatal mortality between experimen- tal and control groups. While low birth weight is sta- tistically associated with high perinatal mortality, and matemal malaria is associated with perinatal mortality (9), whether interventions that increase mean birth weight result in a concomitant reduction in mortality is open to question. Indeed, it can be hypothesized that increasing birth weight could result in higher, not lower, perinatal mortality (16). If infants bom in malarious areas are normally pro- portioned, with head circumference and length val- ues appropriate to their weight, it can be hypothe- sized that chemoprophylaxis may simply reduce the malarial effect and produce infants of a larger size, including a larger head. Although the overall effect on mean birth weight (where demonstrated) tends to be small, if the effect was occurring in a particular subgroup of women it is possible the effect could predispose those women to an increased risk of obstructed labour. Although in both the studies of Nosten and Greenwood there was a greater propor- tion of high birth weight babies in the chemoprophy- laxis groups, the differences were not statistically WHO Bulletin OMS. Vol 72 199496 Antimalarial drug prophylaxis during pregnancy significant. To explore this further, we examined some observational data from McGregor et al. (7). The prevalence of high birth weight (.4000 g) was 4.5% (231/5076) in those women with placentae negative for malaria parasites, compared with 2.7% (32/1183) in the positive group. This gives a relative risk of high birth weight with a parasite-negative placenta, compared with a positive placenta, of 1.7 (95% CI: 1.2, 2.4). However, in primigravidae, who may be at higher risk of cephalopelvic disproportion, no difference in prevalence of high birth weight infants between placental negative and positive groups was demonstrated. Although the data suggest that the absence of malarial infection in the placenta is associated with an increased prevalence of high birth weight infants, whether this is actually the result of the absence of malarial infection or some other confounding factor is not known. Out of the four studies where obstetric outcome was examined, malarial prophylaxis was not asso- ciated with a worse obstetric outcome. Two studies were only of primigravidae, and one of these demon- strated a remarkable positive effect of iron or folic acid supplementation, but not malarial prophylaxis, on adolescent growth in pregnancy (21), and a reduc- tion in the use of operative delivery in the women taking folic acid. However, the malarial prophylaxis was started relatively late in pregnancy (in the sec- ond trimester), after the peak of parasitaemia and occurrence of anaemia, so that haematinic drugs may be effective but antimalarial drugs not. On the other hand, the finding raises questions about the effects of malaria chemoprophylaxis in teenage mothers in the absence of iron or folic acid supplements, as birth weight may be increased but height and pelvic capacity may remain static, possibly resulting in a higher incidence of cephalopelvic disproportion. In women who have not received supplementation, and in whom the epiphyses have fused, malaria chemo- prophylaxis leading to a larger baby may increase the risk of cephalopelvic disproportion. Although this remains as conjecture in the absence of data, it does emphasize the need to measure obstetric out- come in future clinical trials of prophylactic regi- mens. In the meantime, it highlights the need to maintain folic acid supplementation for all women receiving chemoprophylaxis. Future work There is a need for large, randomized controlled studies to examine malaria chemoprophylaxis inter- ventions in relation to matemal health, perinatal and neonatal mortality, obstetric outcome, and the rela- tionship with matemal growth. These studies should be placebo-controlled and double-blind, with careful attention to randomization, commencing early in pregnancy and preceded by a treatment of malaria infection. The endemicity of malaria in the study site needs describing in detail, in terms of both parasite prevalence and the morbidity it causes, as the effects of malaria on the human host vary considerably at different levels of endemicity and host immunity. Limiting the study to primigravidae only, in whom malaria infection is more prevalent, may be pragma- tic for study design but will mean that questions about the effect in multigravidae remain unanswered. The use of a placebo control group requires that all participants have ready access to antimalaria treatment when they develop symptoms. In theory, intermittent presumptive treatment (if correctly timed) may achieve equal efficacy to continuous chemoprophylaxis, so the study would demonstrate whether chemoprophylaxis had any additional advantage to simply treating malarious illness in pregnant women. To answer questions about the effectiveness of routine prophylaxis in stable endemic areas, research needs to carried out in Africa because of its high level of cephalopelvic disproportion (29). Groups should receive iron and folic acid supplements. Out- comes to be measured must include matemal illness (malarial and non-malarial), obstetric outcome, infant gestation, and perinatal and neonatal mortality. The impact of routine prophylaxis in areas of epi- demic, unstable malaria may be different: studies in parts of Asia, such as those being carried out by Nosten et al., will be important in answering ques- tions conceming the effectiveness of routine prophy- laxis in such areas. At an intemational level, we need to clarify cur- rent WHO guidelines for women living in endemic malarious areas as regards (i) targeted chemoprophy- laxis of anaemic women or primigravidae, and (ii) appropriate, early presumptive treatment of illness episodes in all pregnant women. Acknowledgements This review has been carried out as part of the Cochrane .Collaboration, which was established to prepare and main- tain systematic reviews of randomized controlled trials across all areas of health care. We are very grateful to Dr lain Chalmers (Cochrane Centre, Oxford) for his sup- port, encouragement and advice at all times. We also thank all the scientists, particularly Frangois Nosten, Alice Greenwood and T.K. Mutabingwa for providing infor- mation and additional data; D. Morley, V.J. Hartfield, M. Kramer and J. Cattani for their advice; and J. Eyers for assisting with the computer database searches. Paul Garner is currently supported by the Overseas Develop- ment Administration (UK). WHO Bulletin OMS. Vol 72 1994 97 P. Gamer & B. Brabin R6sum6 Essais controles randomis6s sur la prophylaxie antipaludique de routine pendant la grossesse dans les r6gions d'end6mie palustre Les recommandations mondiales actuelles en matibre de chimioprophylaxie antipaludique de routine chez la femme enceinte dans les regions d'end6mie palustre ne sont pas clairement defi- nies. L'efficacite potentielle de la lutte antipalu- dique par chimioprophylaxie pendant la grossesse est limit6e du fait de l'emergence de parasites resistants. L'innocuite des nouveaux medicaments est parfois moins bien etablie que celle des m6di- caments classiques et ils peuvent etre plus coO- teux. De plus, I'administration de la chimioprophy- laxie pendant la grossesse exige d'importantes ressources en matibre de services de sant6, et des efforts de promotion de la sant6. Nous avons examine les donnees relatives aux essais contr6les randomises sur 1'efficacit6 de la chimioprophylaxie antipaludique pendant la gros- sesse chez des femmes vivant dans des zones d'endemie. Toutes ces donnees ont et6 recher- ch6es dans la litterature. Les etudes ont ete sys- tematiquement regroupees et analys6es du point de vue des resultats de la prophylaxie chez la mbre et l'enfant. La chimioprophylaxie de routine semble avoir un effet sur les episodes morbides ant6natals et sur l'hematocrite. On observe une tendance a un poids de naissance plus 6lev6 dans les groupes traites, de fagon significative dans certains cas. Une seule 6tude a tenu compte de l'age gesta- tionnel. Les effets sur la mortalite p6rinatale et neonatale ont ete peu etudi6s sauf dans quelques 6tudes portant sur de faibles effectifs. Cette analyse conduit a se demander si la chimioprophylaxie antipaludique de routine est le meilleur usage que l'on puisse faire de ressources limitees dans les pays en d6veloppement, et sug- gbre que la chimioprophylaxie pourrait etre axde sur les femmes an6miques et les primigestes. De vastes 6tudes contr6lees contre placebo offrant un traitement satisfaisant aux groupes placebo sont necessaires pour etablir si la chimioprophy- laxie de routine presente des avantages sur le traitement precoce efficace du paludisme clinique. References 1. Brabin BJ. An analysis of malaria in pregnancy in Africa. Bulletin of the World Health Organization, 1983, 61: 1005-1016. 2. Brabin B. The risks and severity of malaria in preg- nant women. Geneva, WHO (Applied Field Research in Malaria reports No. 1), 1991. 3. Kramer MS. Determinants of low birth weight: methodological assessment and meta-analysis. Bul- letin of the World Health Organization, 1987, 65: 663-737. 4. Brabin BJ et al. Consequences of maternal anae- mia on outcome of pregnancy in a malaria endemic area in Papua New Guinea. Annals of tropical medi- cine and parasitology, 1990, 84: 11-24. 5. Garner PA et al. Birthweight and gestation of vil- lage deliveries in Papua New Guinea. Journal of tropical paediatrics (in press). 6. MacGregor JD, Avery JG. Malaria transmission and fetal growth. British medical journal, 1974, 3: 427-482. 7. McGregor IA, Wilson ME, Billewicz WZ. Malaria infection of the placenta in the Gambia, West Africa: its incidence and relation to stillbirth, birthweight and placental weight. Transactions of the Royal Society of Tropical Medicine and Hygiene, 1983, 77: 232-244. 8. Steketee RW. Recent findings in perinatal malaria. Bulletin of the International Paediatric Association, 1989, 10: 418-433. 9. Nyirjesy P et al. Malaria during pregnancy: neona- tal morbidity and mortality and the efficacy of chloro- quine prophylaxis. Clinical infectious diseases, 1993, 16: 127-132. 10. Advances in malaria chemotherapy. Report of a WHO Scientific Group. Geneva, World Health Organization, 1984 (WHO Technical Report Series, No. 71 1). 11. Greenwood BM et al. The effects of malaria chemo- prophylaxis given by traditional birth attendants on the course and outcome of pregnancy. Transactions of the Royal Society of Tropical Medicine and Hygiene, 1989, 83: 589-594. 12. Brabin BJ et al. Failure of chloroquine prophylaxis for falciparum malaria in pregnant women in Madang, Papua New Guinea. Annals of tropical medicine and parasitology, 1990, 84: 1-9. 13. Kaseje DCO, Sempebwa EKN, Spencer HC. Malar- ia chemoprophylaxis to pregnant women provided by community health workers in Saradidi, Kenya. I. Reasons for non-acceptance. Annals of tropical medicine and parasitology, 1987, 81(S1): 77-82. 14. Nathwani D et al. Plasmodium falciparum malaria in pregnancy: a review. British journal of obstetrics and gynaecology, 1992, 99: 118-121. 15. Practical chemotherapy of malaria. Report of a WHO Scientific Group. Geneva, World Health Organization, 1990, (WHO Technical Report Series, No. 805): 136. 16. Garner P, Kramer MS, Chalmers I. Might efforts to increase birthweight in undernourished women do more harm than good? Lancet, 1992, 340: 1021- 1023. 17. Chalmers I et al. The Oxford database of perinatal trials: developing a register of published reports of controlled trials. Controlled clinical trials, 1986, 7: 306-324. 18. Cot M et al. Effect of chloroquine chemoprophylaxis during pregnancy on birthweight: results of a ran- 98 WHO Bulletin OMS. Vol 721994 Antimalarial drug prophylaxis during pregnancy domized trial. American journal of tropical medicine and hygiene, 1992, 46: 21-27. 19. Fleming KA et al. The prevention of anaemia in pregnancy in primigravidae in the guinea savanna of Nigeria. Annals of tropical medicine and parasitol- ogy, 1986, 80: 211-233. 20. Hamilton PJS et al. The role of malaria, folic acid deficiency and haemaglobin AS in pregnancy at Mulago Hospital. Transactions of the Royal Society of Tropical Medicine and Hygiene, 1972, 66: 594-602. 21. Harrison KA et al. Growth during pregnancy in Nigerian teenage primigravidae. British journal of obstetrics and gynaecology, 1985, supp. 5: 32-39. 22. Martin GE, Nkwate CC. Administration de la dose unique mensuelle de 600 mg de chloroquine base dans le contr6le du paludisme chez les femmes enceintes. CEAC bulletin, 1982, 53: 41-47. 23. Morley D, Woodland M, Cuthbertson WF. Con- trolled trial of pyrimethamine in pregnant women in an African village. British medical journal, 1964, 1: 667-668. 24. Mutabingwa TK, Malle LN, Mtui SN. Chloroquine therapy still useful in the management of malaria during pregnancy in Muheza, Tanzania. Tropical and geographical medicine, 1991, 43: 131-135. 25. Nahlen BL et al. Lack of efficacy of pyrimethamine prophylaxis in pregnant Nigerian women. Lancet, 1982, 2: 830-833. 26. Nosten F et al. Mefloquine prevents malaria during pregnancy: a double-blind placebo controlled trial. Journal of infectious diseases (in press). 27. Bracken MB. Statistical methods for analysis of effects of treatment in overviews of randomized trials. In: Sinclair JC, Bracken MB, Siverman WA, eds. Effective care of the newborn infant, Oxford, Oxford University Press, 1992: 13-18. 28. Victora CG et al. Birthweight and infant mortality: a longitudinal study of 5914 Brazilian children. Interna- tional journal of epidemiology, 1987, 16: 239-245. 29. Harrison KA. Mode of delivery with notes on rup- ture of the gravid uterus and vesicovaginal fistula. British journal of obstetrics and gynaecology, 1985, suppl. 5: 61-71. WHO Bulletin OMS. Vol 72 1994