that cross sexual transmission increases severitv 1 Aaby P, Bukh J, I isse IM, Smits AJ. Cross-sex transmission of infection andincreased mortalitv due to measles. Rev Infeci Dis 1986:8:138-43.(unpublished observations) Hence, the effect may be---(unpublished observations)Hence,theeffectmay 2 Pison G, Langaney A. Ihe lesel and age pattern of mortality in Bandafassi

related to viral infections. We previously suggested (Eastern Senegal): restults from a small-scale and intensise multi-round

that virus may take sex determinants from host cells, survey. Populailon Studies 1985;39:387-405.3 Pison G. i)Dnamtque d'une population tradltiuonelle: les Peul Bande (Sencgalcausing those produced in male and female hosts to be Oriental). Paris: Presse Unisersitaires de Franice, 1982.different.' This could facilitate infection of cells in 4 Pison G, Bonneutl N. Increased risk of measles mortality for children withindividuals of the sex. Alternatively, virus

siblings among the Fula Bande Senegal. Rev Infiect DDis 1988;10:468-70.lindividuals Of the opposite sex. AlternatMvely, virus s NlcCullagh P, Nelder JA. Generalized linear models. New York: Chapman and

from someone of the same sex may be controlled more Hall, 1983.*1y by the immune system. From this perspectie it *6 McGregor IA. Measles and child mortality in the Gambia. West Afir M1edeasily by the immune system. From ths perspective it 196413:251-7.

is important to test whether the association between 7 Aabv P. Malnourished or overinfected. An analysis of the determinants ofcross sexual transmission and increased severity of acute measles mortality. Dan Med Bull 1989;36:93-113.

8 Aaby P, Leeuwenburg J. IPatterns of transmission and severity of measlesdisease is limited to viral infections or whether it is a infection. A reanalysis of data from the Machakos area, Kenya. ] InJfect Dismore general phenomenon of infections transmitted 1990;161:171-4.

9 Garenne M, Aaby P. Pattern of exposure and measles mortality in Senegal.directly between humans. Animal studies or laboratory _7 Itfect Dis 1990;161:1088-94.experiments also seem warranted to examine the role of 10 Aaby, P, Molbak K. Siblings of opposite sex as a risk factor for child mortality.cross sexual transmission in severity of infection. 11BM7 1990;301:143-5.c1 Babbott FL, Gordon JE. lodern measles. AmJMfedSci 1954;228:334-61.

12 Bhuiya A, Woltyniak B, D'Souza S, Nahar L, Shaikh K. Measles case fatalityThis study was supported by the Museum National among under-fives: a multivariate analysis of risk factors in a rural area of

d'Histoire Naturelle, the Institut National d'Etudes Demo- Bangladesh. Soc Scz Med 1987;24:439-43.13 Aaby P, Bukh J, Lisse IM, Smits AJ. Risk factors in subacute sclerosinggraphiques, the Centre National de la Recherche Scientifique panencephalitis (SSPE): age-and sex-dependent host reactions or intensisC(UA 49), the Institut National de la Sante et de la Recherche exposure. Rev Infiect Dis 1984;6:239-50.Medicale, and the Office de la Recherche Scientifique et 14 Fargues P, Nassour 0. Douze ans de mortaliti urbaine au Sahel. Paris: Presses

Universitaires de France, 1988.Technique Outre-Mer. PA received support from the Danish 15 Monastiri H. Quelques donnees statestiques relatives a la mortalite parCouncils for Development Research, Medical Research, and rougeole dans la Commune de runis. La 7unisie Medical 1961;39:179-87.Social Science Research. We thank the Ministere du Plan et 16 Aabv P. Severitv of measles and cross-sex transmission of infection inde la Cooperation and the Ministere de la Sante, Senegal, for 1 Cpenhagen, 1915-1925. In] Epideiol 199120:504-7.17 Aaby P, Leeuwenburg J. Sex and patterns of transmission of measles infection.their agreement, interest, and help in our work; Sophie A reanalysis of data from the Machakos area, Kenya. Ann Trop Paediatr inAuger, Josette Benaben, Francoise Branson, Sara Camara, press).Mamadou-Yero Diallo, Catherine Enel, Daniele Fouchier, 18 Aaby P, Lamb WH. Sex and transmission of measles in a (iambian village.Mussa Kebe, Kili Keita, Andre Langaney, Maria Ramirez, J InJeci 1991;22:287-92.and Lampa Sadiaho, who participated or helped in collectingand coding the data. Accepted 13 November 1991)

V "'

Impact of rice based oral rehydration solution on stool output andduration of diarrhoea: meta-analysis of 13 clinical trials//

Sheila M&ore, Olivier1Iontaine, Nathaniel Peerce

Abstract cholera diarrhoea and should be more preciselyObjective-To define the benefit of rice oral defined before its practical value can be judged.

rehydration salts solution in relation to the glucosebased World Health Organisation oral rehydrationsalts solution for treating and preventing dehydration Introductionin patients with severe dehydrating diarrhoea. Oral rehydration therapy with the glucose andDesign-Meta-analysis using data from 13 avail- electrolyte solution recommended by the World Health

able randomised trials that compared these two Organisation and Unicef is the preferred method forformulations. treating children with dehydration due to diarrhoea,Subjects-The studies compared 1367 patients provided that they are able to drink and do not have

with cholera, severe cholera-like diarrhoea, or acute signs of shock.' Although the solution is both safenon-cholera diarrhoea. 668 received the standard and effective (D Mahalanabis, unpublished WHOWHO solution and 699 the rice based solution. document), it has important limitations: it neitherIntervention-Each trial report was reviewed to reduces the rate of stool loss nor shortens the duration

determine patient eligibility, the number of patients of illness.>5 Mothers often do not understand the(\ledical Research Council>) who were randomised and the number of these relation between diarrhoea and dehydration, and theirBiostatistics Unit, excluded from analysis, details of the randomisation primary concern, shared by many health workers, is toCambridgeh procedure, and the precise timing of the outcome see the diarrhoea stop. This probably accounts for

Gore, P'HD, medical measurements. the continuing widespread use of ineffective "anti-statistician Main outcome measures-Stool output during the diarrhoeal" drugs and antibiotics to treat diarrhoea

World Health first 24 hours; weighted estimates of the difference in instead of, or in addition to, oral rehydration saltsOrganisation, Diarrhoeal mean stool output between treatments. solution (WHO diarrhoeal diseases control pro-Disease Control Results-The rice solution significantly reduced gramme, seventh programme report, 1988-89, 1990).Programme, 1211 Geneva the rate of stool output during the first 24 hours by If a packaged oral rehydration salts formulation27, Switzerland 36% (95% confidence interval 28 to 44%) in adults could be developed that not only had the positiveO Fontaine, MD, medical with cholera and by 32% (19 to 45%) in children with features of the WHO formulation, including low costofficer cholera. The rate of stool loss in infants and children and safety and stability during prolonged storage, butN F Pierce, MD, medical with acute non-cholera diarrhoea was reduced by also substantially reduced the duration of diarrhoea orofficer only18% (6 to 30%). the rate of stool loss, it would have considerable

Coresondncto: Dr Conclusions-The benefit ofrice oral rehydration advantages. In particular, it could be promoted asCorso ndence. salts solution for patients with cholera is sufficiently having a real antidiarrhoeal effect. This should improve

great to warrant its use in such patients. The benefit its acceptance and use by both health workers andBMJ 1992;304:287-91 iS considerably smaller for children with acute, non- mothers, especially if its benefits were sufficiehtly great

BMJ VOLUME 304 1 FEBRUARY 1992 287

to be evident to them. It might also result in less use ofineffective drugs and antibiotics. Such changes wouldrepresent a major advance in efforts to control morbid-ity and mortality associated with diarrhoea througheffective case management.

Several clinical trials have shown that an oralrehydration salts solution containing cooked ricepowder (50-80 g/l) in place of the usual glucose (20 g/l)substantially reduces the rate of stool loss due to acutediarrhoea.6'0 Other studies, however, have reported nosignificant benefit."''4 The subjects in these studiesvaried considerably and included infants, children,and adults both with diarrhoea associated with choleraand with acute diarrhoea not associated with cholera.Moreover, in some studies the number of patientsevaluated was probably insufficient to support firmconclusions. To define more precisely the true benefitof rice oral rehydration salts solution in relation to theWHO oral rehydration salts solution and to determinewhether this benefit is related to the patient's age or theaetiology of diarrhoea we performed a meta-analysis byusing data from all available randomised clinical trialsthat compared these two formulations.

MethodsSELECTION OF TRIALS

Studies included in this overview were identified bya computer aided search of the published work, byreviewing the references cited in relevant reports, andby inquiring about completed but unpublished studiesfrom our colleagues. Ten published reports6'5 andthree unpublished ones (A M Moechtar, E Guiraldes,and N H Alam, personal communications) wereidentified and are reviewed in this analysis. On thebasis of their design or method of analysis these 13studies yielded 17 comparisons between patient groupstreated with rice oral rehydration salts or the WHOoral rehydration salts solution. Table I gives theprincipal features of each comparison. In all cases thestudies were randomised trials that compared standardWHO oral rehydration salts solution with an experi-mental oral rehydration salts solution in which glucose(20 g/1) was replaced by 50-80 g/l of rice powder,the electrolyte concentrations remaining unchanged.In early studies (A N Alam, personal communi-cation)8' 1 8-15 the rice powder was cooked immediatelybefore use and salts were added after the rice solutionhad cooled. In the most recent6' 12 (Moechtar et al,Guiraldes et at) a commercially produced, precookedrice powder was prepackaged with oral rehydrationsalts, in sachets to make up one litre. This was

TABLE I -Characteristics ofrandomised trials of rice oral rehydration solution

No randomised toCholera (proportion Amount of rice in WHO/rice solution

Comparison Age Dehydration proved on culture) solution (g/l) (No excluded)

Patients with cholera or cholera-like illnessMoechtar

et al (1)* > 12 years Severe Yes 50 83/81 (0/0)Moechtar

et al (2)* > 12 years Severe No 50 12/14 (0/0)Alam et al (I)t Adults Moderate to severe Yes 50 47/46(?)Alam et al (2)t Adults Moderate to severe Yes 50 42/47 (?)Molla etal (1))* >10 years Moderate to severe Yes (65%) 80 74/85 (2/0)Molla et al (2)'* ?-<10 years Moderate to severe Yes (75%) 80 105/84 (4/0)Molla et al (3)" 1-5 years Moderate to severe Yes (55%) 50 42/37 (?)Molla (4)7 2-5 years Moderate to severe Yes (80%) 80 25/27 (0/0)Alam (3)9* 1-8 years Moderate to severe Yes (100%) 50 19/20 (2/2)Alam (4)9* 1-8 years Moderate to severe No 50 7/6 (0/0)Patra et at6 3 months-5 years Moderate to severe Yes (30%) 50 26/26 (212)

Patients without choleraGuiraldes et al 4-24 months Moderate No 50 49/51(1/2)Kenya etal'4 4-59 months Moderate to severe No 60 50/51 (1/1)Dutta et alt 4 months-4 years Moderate No 50 33/37 (0/0)Bhan et al' 3 months-5 years Moderate No 50 33/31 (0/0)El Mougi et al" 4-18 months Moderate No 50 30/30 (4/5)Mohan et alt' 3 months-3 years Mild to severe No 50 24/26 (1/3)

*Single studies in which results were stratified for analysis.tClinical trial with a factorial design (4 cell trial).

prepared for use in the same way as standard WHOoral rehydration salts, as precooked rice dissolvesrapidly in cold water.To our knowledge no other randomised trials of

these formulations ofrice oral rehydration salts solutionhave been completed, although several are underway.We have not included three trials of solutions thatcontained only 30 g/l of rice powder (O Fontaine,personal communication)'2 '6 and one of an oral rehy-dration salts solution containing rice powder andglycine. 7Each trial report was reviewed independently by a

statistician (SMG) and a clinician (OF) to determinepatient eligibility according to stated selection criteriafor age and dehydration status; the number of patientswho were randomised and the number of these sub-sequently excluded from analysis; details of therandomisation procedure; -and the precise timing of theoutcome measurements, such as stool output andintake of oral rehydration salts solution.

META-ANALYSIS

Each of the 17 comparisons yielded an estimate ofthe true difference in mean stool output betweenpatients treated with the two different salts solutions;and each difference in means (Di for comparison i) isqualified by a variance:

var (Di)=(sei)lThe larger the variance, the less precise is the

observed difference as an estimate of the true differencein mean stool output between treatment groups. Itfollows that the amount of information conveyed by asingle comparison about this true difference is inverselyproportional to the variance of the estimated differencefor that comparison. If the inverse of the variance istaken as a measure of the information contained incomparison i, then the weight (wi) that should beaccorded to comparison i, among all comparisons in itsset, may be taken to be the information in comparison idivided by the total information (that is, the sumover all comparisons i= 1, 2,. . . n of the inverse of thevariances):

wi =[(l/sei)']/-i_,I(1/sei)lClearly, these weights add up to one. A pooled or

weighted estimate of the difference in mean stooloutput between patients treated with rice oral rehy-dration salts solution or WHO oral rehydration saltssolution is obtained by summing the differencesin individual comparisons, each multiplied by itscorresponding weight, so that if a single comparisonaccounts for 10% of the information then 10% ofits estimated difference counts towards the pooledestimate. With the foregoing choice of weights, thevariance of the pooled difference has a particularlysimple form. Just as we defined information as theinverse of variance, so the variance of the pooledestimate is the inverse of the total information:var (pooled difference)= var vi-Iwi Di= I/i= I (1/sei)2A 95% confidence interval for the pooled estimate

runs from two standard errors below the pooledestimate to two standard errors above, where thestandard error is the square root of the variance of thepooled estimate.

ResultsEVALUATION OF TRIALS

The review identified problems in both the designand the analysis of some trials. These are summarisedbelow.

RandomisationThe randomisation of patients should have occurred

BMJ VOLUME 304 1 FEBRUARY 1992288

TABLE II-Stool output in first 24 hours in adults with diarrhoea associated with cholera or cholera-like illness

Treatment with WHO solutionAlean reduction (variance)

MNtean (SD) stool MNlean stool output/ in stool output duriiigNo analysed taking output (g or ml/kg Mlean stool mean intake of treatment with rice

Comparison WHO/rice solution in first 24 h) output/SD solution solution (g or ml/kg) Study weight

Moechtaretal(1)6-24hours 83/81 133(92) 1-4 0-63 39(148) 0-45Moechtar et al (2) 6-24 hours 12/14 106 (55) 1 9 0-63 48 (407) 0-16Alametal(1) 47/46 391 (157) 2 5 075 164(829) 0-08Alametal(2) 42/47 366(174) 2-1 0-81 144(1068) 0-06Mollaetal(1)' 72/85 159(109) 1 5 0-64 44(266) 0-25

TABLE III-Stool output in first 24 hours in children with diarrhoea associated with cholera or cholera-like illness

Treatment with WHO solution-Mean reduction (variance)

Mean (SD) stool Mean stool output/ in stool output duringNo analysed taking output (g or mi/kg Mean stool mean intake of treatment with rice

Comparison WHO/rice solution in first 24 h) output/SD solution solution (g or ml/kg) Study weight

Mollaetal(2)' 101/84 204(140) 1-5 0-54 49(362) 0-42Mollaetal(3)"' 42/37 343(151) 2-3 1-49 181 (628) ZeroMollaetal(4)7 25/27 210(158) 1-3 069 105(1206) 0-12Alametal(3)9* 19/20 290(190) 1-5 0-98 160(2500) 0-06Alametal(4)9* 7/6 90(75) 1-2 0-57 -40(700) 0-21Patraetal' 24/24 166(114) 1-5 0-62 69(794) 0-19

*Data not reported quantitatively; values in this table are approximated from graphic presentation.

TABLE IV-Stool output in first 24 hours in children with diarrhoea not associated with cholera

Treatment with WHO solution

Mean (SD) stoolNo analysed taking output (g or ml/kg

Mean reduction (variance)Mean stool output/ in stool output during

Mean stool mean intake of treatment with riceComparison WHO/rice solution in irst 24 h) output/SD solution solution (g or ml/kg) Study weight

Guiraldes etal 48/49 126(64) 2-0 0-30 14(219) 0-21Kenya et alt 49/50 103 (31) 3-3 0-48 3(35) ZeroDuttaetal'2 33/37 103(55) 1-9 0-60 17(145) 0-32Bhan et al" 33/31 77(58) 1-3 0-49 10(161) 0-29El Mougi et al" 26/25 245 (129)* 1-9 0-72 82 (1115) 0-04Mohan et all' 23/23 110(69) 1-6 22(340) 0-14

*The reported SD (25 -3) was very low in relation to the large observed difference in stool output; we assumed that the reported SD values were actually SEsand revised the SDs accordingly.

immediately before treatment with oral rehydrationsalts solution began-that is, after the completion ofany intravenous treatment for severe dehydration.However, in no trial was it stated when patients withsevere dehydration were randomised and outcomemeasurements initiated-that is, before or after initialintravenous rehydration. Thus it was unclear whetherthe first 24 hour measurement of stool output beganwhen intravenous rehydration was started or when oralrehydration salts solution was first given, as shouldhave been the case.

In one study patients were randomised irrespectiveof age, but were stratified into arbitrary age groupsduring analysis.' Ideally, such stratification shouldhave been part of the randomisation plan. Stratificationduring analysis was also done in two other studies(Moechtar et al) , but this was based on aetiology and sowas unavoidable.

Exclusion from analysisPragmatic analysis according to intention to treat

requires that all randomised patients continue to bemonitored and that their data be included in theanalysis. Nevertheless, in seven trials (Guiraldes etal)6 I9 13-1S 1-15% of randomised patients were excludedfrom the analysis (table I), either because they wereconsidered to be "treatment failures" (usually becauseadditional intravenous treatment was required) orbecause they had been randomised in error. Intwo trials that used a permuted block'° or factorialdesign (Alam et al) it seems that some patients wererandomised but not reported on, as the numbersspecified in the different treatment groups differedappreciably. The reasons for these differences werenot stated.

Analysis and internal consistency ofoutcome dataWhereas all studies reported stool output and oral

rehydration salts solution intake during the first 24hours, few reported total stool output until diarrhoeastopped, and only seven studies reported the durationof diarrhoea. Our analysis therefore focused largely onstool output during the first 24 hours. The followingresults for the first 24 hours are reported: mean(standard deviation) stool output (in g or ml/kg bodyweight) for patients randomised to WHO oral rehy-dration salts solution; the ratio ofmean stool output toits standard deviation; the ratio ofmean stool output tomean intake of WHO oral rehydration salts solution;and the mean reduction in stool output (in g or ml/kg)for patients given rice oral rehydration salts solutioncompared with those given WHO oral rehydrationsalts solution, and the variance of that value.

Tables II, III, and IV show the mean (SD) stooloutput (in g or ml/kg) during the first 24 hours forpatients in each study who were randomised to receiveWHO oral rehydration salts solution. Whether thedata were for adults with cholera or with cholera-likediarrhoea (severe dehydrating diarrhoea, clinicallyresembling that associated with cholera but fromwhich Vibrio cholerae 01 was not isolated) (table II),children with cholera or cholera-like diarrhoea (tableIII), or children with only acute non-cholera diarrhoea(table IV), the ratios of mean to standard deviation forstool output were roughly constant, averaging 1 6 andranging (with one exception) from 1-2 to 2 5. Thisregularity indicates the need for logarithmic trans-formation; however, no study reported logarithmicallytransformed data or performed calculations on thatscale. This finding also provides a criterion for judgingthe internal consistency of key outcome data. By this

BMJ VOLUME 304 1 FEBRUARY 1992

_^

289

criterion one trial seems to be atypical with a ratio of3.3,14 twice the mean value reported in other studies.Even more extreme was the ratio of 10 from the datareported in another study.'5 We suspected confusionbetween standard error and standard deviation in thisstudy, and therefore table IV shows what we believe tobe the correct standard deviation for this trial, a valuesimilar to those in the other studies.

Tables II, III, and IV also show a second measure bywhich to assess the internal consistency of trial data-namely, the ratio of mean stool output to mean intakeof WHO oral rehydration solution. Mean stooloutput averages about two thirds of mean oral rehy-dration salts solution intake. By this criterion one trialseems to be atypical,'0 the mean stool output beingalmost 50% greater than the mean intake of oralrehydration salts solution (table III). In the analysesthat follow, two studies'0 4 have been excluded (that is,zero weighted) for the reasons mentioned above.

SUBSTANTIVE RESULTS

The results of the analysis of stool output and intakeof oral rehydration salts solutions during the first 24hours have been grouped into three sets according tothe patient's age and aetiology of diarrhoea, as shownin tables II, III, and IV. For the duration of diarrhoeaall data have been combined, but comparisons inpatients with suspected cholera who received tetra-cycline before diarrhoea stopped have been zeroweighted (table V).

Adults with cholera or cholera-like diarrhoea-Table IIshows the weights assigned to each of the five com-parisons in this set. By using these weights theestimated mean stool output for patients given WHOoral rehydration salts solution was 170 ml/kg. Forpatients given rice oral rehydration salts solution thiswas reduced by a mean of 58 ml/kg (36%, 95%confidence interval 28 to 44%).

Children with cholera or cholera-like diarrhoea-TableIII shows the weights assigned to each of the fivecomparisons in this set. One study'5 was zero weightedfor reasons described above. With those weights theestimated mean stool output for patients given theWHO oral solution was 178 ml/kg. For patients giventhe rice solution this was reduced by a mean of48 ml/kg(32%, 19% to 45%).

Children with non-cholera diarrhoea -In this set of sixcomparisons, one study'4 was zero weighted for reasonsdescribed above (table IV). By using the weightscalculated for the other five comparisons the estimatedmean stool output for patients given the WHO solutionwas 107 ml/kg. For patients given the rice solution thiswas reduced by a mean of 18 ml/kg (18%; 6% to 30%).It is noteworthy that the estimated mean percentagereduction in stool output associated with the ricesolution in the zero weighted study is outside the 95%confidence interval derived from the other five studies.

Overall reduction in stool output-The figure presentsthe percentage reduction in mean stool output (with

TABLE v-Duration ofdiarrhoea

Treatment with WHO solution Mean reductionNo analysed (variance) duration of

taking WHO/rice Mean (SD) No of Mean diarrhoea with riceComparison solution patients duration/SD solution Study weight

.Adults with choleraMoechtaretal(1) 83/81 39(11) 3-5 2(3) ZeroMoechtaret al (2) 12/14 36(7) 5-1 7(13) ZeroAlametal(1) 47/46 86(22) 3-9 9(26) 0-23Alam et al (2) 42/47 85 (20) 4-2 4 (22) 0-27

Children with choleraAlamet al (3)9 24/24 90(43) 2-1 12(109) 005Patraet al 24/24 43 (22) 2-0 13 (36) 0-17

Children without choleraKenya et al' 49/50 46 (9) 5-1 4 (4) ZeroDuttaet al'2 33/37 79(37) 2-1 10(67) 009ElMougiet al'6 26/25 34(12) 2-8 6(31) 0-19

.C:t3cocacia

Moechtar et al (1 ) -Moechtar et al (2) -

CO Alam et al (1)-3' Alam et al (2)-tcUMolla etai(1) -

-c)

%_ Molla et al (2) -

1.o Molla et al (4)-o Alam et al (3) -

Alam et al (3) -co Patra et al -ai)0

Guiraldes et al-In D Dutta et al -

, Bhan et al-L° El Mougi et al -

c Mohan etal -0

Patients withcholera diarrhoea -

Patients with -non-cholera diarrhoea

"

l F

-8 -6 -4 -2 0 0 6

Percentage reduction in stool outputMean percentage reduction in 24 hour stool output in individual studiesof adults and children with cholera or cholera-like diarrhoea andchildren with non-cholera diarrhoea given rice oral rehydration saltssolution. Pooled (weighted) estimates ofpercentage reduction in meanstool output (95% confidence interval) for each group of studies areshown in shaded box

95% confidence intervals) for patients treated with therice solution in each of the comparisons considered inthis overview, as well as the pooled (weighted) estimatesof the percentage reduction in mean stool output forpatients with cholera (adults and children) and withoutcholera (details of these calculations are not shown).The effect of the rice solution on stool output wassignificantly less in children with non-cholera diarrhoeathan in children and adults with cholera or cholera-likediarrhoea (95% confidence interval 3% to 31% for thedifference in percentage reduction in stool outputin patients with cholera or cholera-like diarrhoea vpatients with non-cholera diarrhoea).

Duration of diarrhoea-Data from six comparisons,including both adults and children with cholera (whohad not received tetracycline before diarrhoea stopped)and acute non-cholera diarrhoea were considered forthis analysis (table V). The estimated mean duration inpatients given the WHO oral rehydration solution was68 hours. For those given the rice solution the durationwas reduced by a mean of eight hours (12%; 5% to19%). The 95% confidence interval excludes zero,indicating a modest but significant reduction in theduration of diarrhoea.

DiscussionIrrespective of their age, patients with cholera who

were given rice oral rehydration salts solution hadsubstantiall lower rates of stool loss than those whowere givenWHO oral rehydration salts solution. Stoolvolume was reduced by a mean of 48-58 ml/kg duringthe first 24 hours of treatment, which was 32-36% lessoutput than for patients given theWHO solution. Thispresumably reflects the fact that a greater amount ofglucose (and amino acids) is released when rice powderis fully digested than is present in the WHO solution.Assuming that glucose facilitated absorption of sodiumproceeds on an equimolar basis, 50-80 g/l of ricepowder would release sufficient glucose and aminoacids to promote the absorption of all the sodium (andwater) in the rehydration solution and, in addition,reabsorption of at least part of the sodium (and water)secreted into the bowel as part of the diarrhoealprocess, thus diminishing stool output.'8 In contrast,the WHO solution contains only enough glucose (20 g/1) to promote the absorption of the sodium and water in

BMJ VOLUME 304 1 FEBRUARY 1992

I:tI

Icc

ca

290

the solution, thus leaving the rate of stool loss essenti- one litre packet about three times the cost of theally uafce."~The lower osmolarity of the rice standard packet ofWHO oral rehydration salt packet.solution (about 200 mmol/I v about 310 mmolI1) would On the other hand, if the effect on total stool output isalso enhance the intestinal absorption of water, but not appreciably greater, owing to a concurrent shorteningof sodium.2 of illness, a change in oral rehydration salts formulation

In contrast with stool output in cholera patients, that might be justified. This requires further study. In thein children with acute non-cholera diarrhoea was meantime the current data show that rice oral rehy-reduced by a mean of only 18 ml/kg during the first 24 dration salts solution has enough advantage over thehours of treatment with rice solution-that is, by 18% WHO oral rehydration salts solution to justify its use incompared with that in patients given the WHO patients with cholera, where this is convenient.solution. The significantly smaller benefit of the ricesolution for these patients apparently reflects a partial

failure f theprcess decribed aove. ThelikelyI Avery ME, Synder JD. Oral therapv for acute diarrhoea. The underused

simple solution. N hngl_]Med 1990;323:891-4.explanations are that, at least in some patients, rice 2 Pierce NF, Barnwell JC, MNitra RC, Caranasos GI, Keimowitz RI, Mondal A,starch and protein were not fully digested, thus et al. Effect of- ititragastric glucose-electrolyte infusion upon water and

electrolyte balance in Asiatic cholera. Gastroenterolo6' 1968;55:333-43.reducing the amount of glucose and amino acids 3 Hirschhorn N, KinzieJL, ISaclhar DB, Northrup RS, Taylor JO, Ahmad SZ, e

available to promote sodium absorption; or released al. Decrease in net stool output in cholera during intestinal perfusion withglucose containing solutions. N Engli] Med 1968;279:176-80.

glucose was not fully absorbed; or both. Failure to 4 Mahalanabis D, Sack RB, Jacobs B, Mondal A, Trhomas J. Use otf an oral-

digest rice powder fully could reflect reduced amylase glucose-electrolyte solution in the treatment of pediatric cholera: a controlledor disacchardase activites related t young age,

study. Journal of 7ropical Pediatrics antd Envi'ronmental Child Ilealthor disaccharidas activities relaed to young age, 1974;20:82-7.malnutrition, or mucosal damage by the infective 5 Sack DA, Chowdhury AMIAK, Eusof A, Ali MA,AMersonAMH, Islam 5, et al.

aet21.21 uoa a aecudas as lcs Oral hydration in rotavimus diarrhoea: a double-blind comparison of sucrosewith glucose-electrolyte solution. Lancet 1978;ii:280-3.

malabsorption, which could exacerbate the rate of stool 6 Patra FC, Mahalanabis ID, Jalan KN, Sen A, Banerjee P. Is oral rice electrolyrteloss owing to the osmotic activity of unabsorbed solution sttperior to glucose electrolyte solution in infantile diarrhoea? .Archt

Dis Child 1982;57:910-2.glucose in the bowel lumen. If this occurred the 7 Miolla AM, Sarker A, Molla A, KhatoonAM, Greenough WB III. Rice-basedadverse effect would be greater for rice oral rehydration oral rehydration therapy in acute diarrhoea: a superior therapy and a

salts solution owing to the greater amount of glucose medium for calorie supplementation. In: Eeckels RE, Ransome-Kuti 0,Kroonenberg CC, eds. Chid health in the tropi'cs: sixth Nutri'cia-Cow and Gate

released when rice starch is fully hydrolysed. This sy,mposium, Leuwen, 18-21 October 1983. Dordrecht, Trhe Netherlands:meta-analysis affords no insight into which of these 8 Nijhoff, 1985:65-70.

MAolla AM1, Ahmed SM, Greenough WIB III. Rice-based oral rehtydrationmechanisms explains the reduced effect of rice oral solution decreases the stool volumne in acute diarrhoea. Bull WHO)rehydration salts on stool loss in patients with acute 1985;63:751-6.

non-cholera diarrhoea. ~~~~~~9Alam AN, Sarker SA, Mlolla AMt, Rahaman MM, Greenough WIB Ill.non-choleradiarrhoea. ~~~~~~~~~~~Hydrolysed wheat-based oral rehydration solution for acute diarrhoea. ArchThe meta-analysis shows that treatment with rice Dis Child 1987;62:440-4.

solution reduces both the rate of stool loss and, to a 10 Molla AM, Nath SK, Miolla A, Khatun M. Food based oral rehvdraItion saltsolution for acutet childhood diarrhoea. Lancet 1989;ii:429-3 1.

lesser extent, the duration of diarrhoea, as compared 11I Bhan MK, Ghai OP., Khoshoo V, Vasudev AS, Bhatnagar NK, Arora R, et al.

with treatmetwith the W O solution. oth of these Efficacy of mung bean and pop rice based oral rehydration soslutions inwith treatentwith he WHO soltion. Bot of thesecomparison with the standard glucose electrolyte solution. ] Pediatrvariables independently affect the total output of Gastrosenterol Nutr 1987;6:392-9.

diarrhoeal stool during the illness. Thus when both are 12 Dutta P, Dutta 1), Bhattacharva SK, Sinha AK, Mondal BC, Pal SC.Comparative efficacy of three different oral rehydration solutions for the

reduced the percentage reduction in total stool output treatment of dehvdrating diarrhoea in children. Indian]edRs1887would be greater than the percentage reduction in 229-33.

13 Mohan Mi, Antonyr TJ, Malik 5, Mathur Mi. Rice p(owder oral rehvdr'ationeither of the contributingmesr en .Thsug stmeasurements. This suggests solutiott as an alternative to glucose electrolyte solution. Indian]J Med Res

that total output of diarrhoeal stool would be the most 1988;87:234-9.important clinical outcome measure when different 14 Kenya PP, Odongo HWI, Oundo G, Wlaswa K, Muttunga J, Molla AM1, et al.

Cereal based oral rehydration solutions. Arch Dis Child 1989;64: 1032-5.oral rehydration salts formulations are compared. 15 El Mougi Mi, Hegazi E, Gailal 0, Ell Akkad N, IEI-Abhar A, Nour N, et al.

Unfortunately,this value was reported for only one Controlled clinical trial on the efficacy of rice powder-based oral rehydrationUnfortunately,this valuewas reported for only onesolution on the otttcome of acute diarrhoea in infants.]J Pedaiar Gastroenterol

study reviewed here6 and for another that compared Nutr 1988;7:572-6.a sorghum based oral rehydration salts solution 16 Mohan Mi, Sethi JS, Daral TS, Sharma M, Bhargava SK, Sachdev HPS.

24 ~~~~~~~~~~~~Controlled clinical trial of rice powder and glucose rehydration solutions aswith the WHO solution.4 In both studies the per- oral therapy for acute dehydrating diarrhoea in infants.]J Pediatr Gastro-

centage reduction was greater in total stool output enterol Nutr 1986;5:423-7.than ineitherthe rae of sool los or dration of 17 Patra FC, Miahalanabis D, Jalan KN, Maitra TK, Sen A, Banerjee P1. A

controlled clinical trial of rice and glvcine-containing oral rehvdrationdiarrhoea. solution for acute diarrhoea in children.] Diarrhoeal Dis Res 1986;4: 16-9.

The last point bears directly on whether the rice 18 Patra FC, Miahalanabis D, Jalan KN, Sen A, Banerjee P. In search of a supersolution: controlled clinical trial of a glycine-glucose oral rehydration

solution (or any other cereal based oral rehydration solution in infantile diarrhoea. Acta Paediatr Scand 1984;73: 18-2 1.

salts solution) would have sufficient advantage over the 19 Hirschihorn N. TFhe treatment of acute diarrhoea in children: an historical andphysiological perspective. Am] Clin Nutr 1980;33:637-63.

WHOslutio to rplaceit in routine use at health 20 Wapnir RA, Zdanowicz M, Teichberg 5, Lifuhitz F. Alanine stimulation offacilities, especially for treating children with acute water and sodium absorption in a model of secretory diarrhoea.]J Pediatr

non-cholera diarrhoea, who represent the overwhelm- 21 GastroenterolNutr 1990;1O:213-2 1.21Lebenthal E, Lev R, Lee PC. In: Lebenthal E, ed. Textbooh ofgastroenterology

ing majority of cases. The average 18% reduction in and nutrition in infancyl. New York: Raven Pren6, 1981:149-65.initial rate of stool loss, if applied to total stool output, 22 Lebenthal E, Heitlinger L, Lee PC, Nord KS, Hodge C, Brooks SIP, Geosrge

D. Corn svrup sugars: in vitro and in vivo digestibility and clinical toleranceis unlikely to justify the major effort and expense in acute diarrhoea of infancy.]J Pediatr 1983;i103:29-34.required to change over from glucose to precooked rice 23 Vesikari T, Isolauri E. Glycine supplemented oral rehydration solutions for

diarrhoea. Arch Di's Child 1986;61:372-6.in th orarehyratin sats fomulaion, specallyin 24 Lepage P, Hitimana DG, Vand Goethem C, Ntahorutaba M, Nsengumuremvi

developing countries. A crude estimation showed that F. Food based oral rehydration salt solution for acute childhood diarrhwathe current cost of commercially prepared oral rehy- Lancet 1989;ii:868.dration salts based on precooked rice would be for each (Accepted 19 November 1991)