CHLORAMPHENICOL ALONE VERSUSCHLORAMPHENICOL PLUS PENICILLIN FOR

BACTERIAL MENINGITIS IN CHILDREN

FRANK SHANN

Goroka Hospita~ Papua New Guinea

JANE BARKER

Kundiawa Hospital, Papua New Guinea

PETER POORELae Hbspita~ Papua New Guinea

Summary 367 children with cerebrospinal-fluidfindings suggestive of bacterial meningitis

were randomised to receive either chloramphenicol alone byintramuscular injection, or chloramphenicol plus penicillinby intravenous injection. Sequential analysis showed nodifference in mortality between the two treatments. 48 (26%)of the 183 children in the chloramphenicol alone group died,and 49 (27%) of the 184 children in the chloramphenicol pluspenicillin group died. In children with bacterial meningitischloramphenicol alone given by intramuscular injection is aseffective as chloramphenicol plus penicillin given

intravenously.Introduction

THE clinical significance of antagonism between penicillinand chloramphenicol has been debated since the in-vitrointeraction was first described in detail by J awetz et al in1951.1 If the two drugs are to be used together, two quiteseparate issues need to be considered: the effect of addingchloramphenicol if a sensitive organism is being treated withpenicillin, and whether anything is gained by addingpenicillin when a patient is receiving chloramphenicol.

In the treatment of infection with an organism,such asStreptococcus pneumoniae, that is sensitive to penicillin, whatis the effect in vivo of giving chloramphenicol in addition topenicillin? A study in dogs-with pneumococcal meningitis2and a randomised controlled trial in patients with meningitis3both suggested that penicillin (or ampicillin) alone wassubstantially better than penicillin plus chloramphenicol.Similarly, Leeper and Dowling4 demonstrated that

tetracycline antagonised the action of penicillin in adults withpneumococcal meningitis. On the other hand, Bodine et al5reported that chloramphenicol did not antagonise the actionof penicillin in pneumococcal meningitis in rabbits. BecauseHaemophilus influenzae is only moderately sensitive topenicillin, and S pneumoniaeoften has reduced sensitivity ropenicillin in Papua New Guinea,6,7 we decided that it wouldnot be ethical to compare benzylpenicillin alone withbenzylpenicillin plus chloramphenicol in children withmeningitis in Papua New Guinea.

When an organism is not very sensitive to penicillin, orwhen the organism or the antibiotic sensitivity is not known,and chloramphenicol is used, is there any benefit from givingpenicillin or ampicillin as well as chloramphenicol? Astudy in rabbits with H influenzae meningitis showed little.difference betWeen chloramphenicol alone, ampicillin alone,and both drugs together.5 There have been rio prospectiverandomised controlled trials in ~an designed to answer thisquestion~ In tWo retrospective studies of children with Hinfluenzae ~eningitis, sequelae were four times morecommon after treatment with ampicillin plus chlor-amphenicol than after chloramphenicol alone.8,9

Chloramphenicol sodium succinate is well absorbed afterintramuscular administration in children, and the injection isnot particularly painfu1,lo-l2 If meningitis is treated withchloramphenicol alone, it can reasopably be givenintramuscularly for 2 or 3 days, then orally when the childstarts to improve. This regimen is cheap and easy toadminister. On the other hand, if both chloramphenicol andpenicillin are given, the large number of injections requiredmeans that ah intravenous drip is usually used.Consequently, use of chloramphenicol plus penicillin is moreexpensive, makes greater demands on the time of staff, andcarries the risks of sepsis13 and overhydration 14,15 associated

with intravenous fluid therapy.

A multicentre prospective randomisedcontrolled trial hasbeen carried out in--children with cerebrospinal-fluid (CS~findings suggestive of bacterial meningitis to see whetherchloramphenicol plus penicillin (given intravenously) is anymore effective than chloramphenicol alone (given

intramuscularly).

8457 @ The Lancet Ltd, 1985

;:~682THELANCET, SEPTEMBER 28, 1985

,

"""", 0,..,""'""'

~~

~ "0.' .,""

"" ::;:::;:~'

~c.Chloramphenicol alone versus chloramphenicol plus penicillin

sequential analysis. ,

Repeated significance test plan, 2a <0.05, ]-{3 >0.95, 9, =0.70, n= 104.

preference had been decided, at a point indicating no

significant difference between the two treatment groups. 258(70%) of the 367 children in the trial were less than 12 monthsold, and 291 (79%) were less than 24 months old.

There was no significant difference (p>O .2, X2 test, I df)between the two treatment groups in the proportion ofchildren who we!'e male, comatose, 0!, aged <6 months, or inthe proportion who had bacteria present on CSF culture,neck stiffness, a bulging fontanelle, haemoglobin >9 .0 g/dl,white blood cell count (WCC) > 13 OOO/jJol, CSF WCC

<750/jJol, axillary tempera~re >38,0°C, or weight-for-age<80% of the Harvard median (the value of each of the

continuous variables chosen for comparison was midwaybetween the medians of the two treatment groups);Howeve!,at the time of admission to the trial, more children in the

chloramphenicol alone group had been ill for 5 days or more<x.2 5.747, I dt; p<0.02), whereas children in the

chloramphenicol plus penicillin group were somewhat morelikely not to be feedi~g <x.2 3.417,1 dt; 0.1 >p >0.05) and tohave had a convulsion <x.2 2.239, 1 dt; 0.2 >p >0. I ). The

proportions of children who died in the two treatment groupswere therefore compared by means of Cochran's testl1 to

,"adjust for the number who had been ill for 5 days or more,were not feeding, or had had ~ convulsion. Although the

unadjusted mortality rate was 26% (48/183) in thechloramphenicol alone group and 27% (49/184) in the

chloramphenicol plus penicillin group, the adjustedmortality rate was lower in the chloramphenicol plus

penicillin group by 2.1 %:t8 .5% (:t95% confidence limits).The outcome was poor (death, discharged with brain

damage, absconded still sick) in 38% (69/183) of the

chloramphenicol alone group and in 40% (74/184) of the

chloramphepicol plus penicillin group (table I); the adjusted

Methods

The study was carried out in three hospitals in Papua NewGuinea: Goroka Hospital (children admitted between May, 1979,

andJune, 1983), Kundiawa Hospital (August, 1979, to April, 1981)and Lae Hospital (October, 1979, to June, 1982). Children wereadmitted to the trial if bacteria grew from their CSF, or the CSFpolymorph count was >100 cellslJil, or the CSF polymorph countwas >20 cellslJilwith either CSF protein > 1.0 g/lor CSF glucose<2.2 mmol/l. The study was approved by the Papua New GuineaMedical Research Advisory Committee.

A table of random numbers was used to prepare sealed, numberedenvelopes. Mter a child had been admitted to the trial, the nextenvelope was opened to determine which treatment to be given. Theinvestigators were aware of the treatment being given to each child.Children randomised to receive chloramphenicol alone were given25 nig/kg chloramphenicol sodium succinate intramuscularly every6 h at first, then 25 mg/kg of chloramphenicol palmitate orally every6 h once clinical improvement had begun. Children who were too illto feed were given fluid by nasogastrictube. Children randomised toreceive chloramphenicol plus penicillin had an intravenous dripinserted. A solution of 4- 3% dextrose in 0. 18% sodium chloridewas infused at the following rates: body-weight 3-4 kg, 10 ini/h;5-9 ~g, 15 ml/h; 10-14 kg, 20 ml/h; 15-39 kg, 25 m1Ih;Chlor-amphenicol sodium succinate was given intravenously every6 h in a dose of 25 mg/kg, and benzylpenicillin was given

intravenously every 3 h in the following doses: body-weight 3-9 kg,500000 U; 10-19 kg, 1000000 U; 20-29 kg, 1500000 U; 30-39kg 2000 000 U. When the child began to recover, the treatment waschanged to 25 mg/kg of chloramphenicol palmitate qrally every 6 harid the above dose of benzylpenicillin intramuscularly every 6 h.Antibiotic therapy was given for a total of 14 days. All chidren weregiven quinin~ or amodiaquine. CSF was cultured on human bloodagar and chocolate agar.

Each child was paired to the next child in the other treatmentgroup at the same hospital. Treatment was said to have failed if thechild died. The preferences of untied pairs were analysed by meansof sequential analysis on a repeated significance test plan with2a <0,05, 1-/3 >0-95, el =0-70, and n=104.16 Thus, there was a95% chance of obtaining a significant result (with p<0-05) if theproportion of untied pairs yielding a prefel:ence for either treatmentwas 0.7 or more.

Results

Of the 367 children who were admitted to the trial, 215were in Goroka (where 3 children in the chloramphenicolalone group were unpaired), 49 in Kundiawa (5

chloramphenicol plus penicillin unpaired), and 103 inL~e (1chloramphenicol alone unpaired). Of the 179 pairs, 101 weretied, 39 favoured chloramphenicol alone, and 39 favoured

chloramphenicol plus penicillin (see figure). The border ofthe sequential analysis graph was reached after the 75th

TABLE I-OUTCOME, BY ANTIBIOTIC TREATMENT

Treatment groupDied

48 (26%)

49 (27%)

TABLE I!-PROPORTION OF CHILDREN IN WHOM TREATMI;N! FAILI;D,* BY RI;SULT OF CSF CULTURI;

Treatment groupTotal

! 69/183 (38%)

74/184 (~*Absconded still sick, discharged brain damaged, or died.

THELANCET,SEPTEMBER28,1985 683

difference was lower in the chloramphenicol alone group by0. 4%:t:9 .1% (:t:95% confidence limits). Table II shows theoutcome of treatment according to the results of bacterialculture of the CSF. An organism was cultured from the CSFof 181 children: 28 (31%) of the 91 treated withchloramphenicol alone died and 26 (29%) of the 90 treatedwith chloramphenicol plus penicillin died; the difference inmortality is 1.8%:t:13. 3% (:t:95% confidence limits), whichis not significant!

30 children were discharged with signs suggesting that theywere likely to have suffered permanent brain damage (table I).10 of these children ( 4 from the chloramphenicol alone group)were comatose or semicomatose and probably died at home,13 ~hi1dren (4 chloramphenicol alone) were conscious buthad spasticity or unequivocal evidence of impaired motor orsocial functioning, and the other 7 children had hemiparesis(3 cases), frequent convulsions (3 cases), or total deafness (Icase). The 48 children who were well at the time theya1:>sconded (table I) had all had at least 5 days of antibiotictreatment. Of the 16 children who were still sick when theyabsconded, 9 (5 chloramphenicol alone) had had less than 4days of antibiotic treatment, 6 (5 chloramphenicol alone)were semicomatose or had poorly controlled fits, and 1 childhad persistent fever and irritability after 10 days ofchloramphenicol alone.

if

I!

iiL

'Il

~

numbers of polymorphs, and increased protein or decreasedglucose in the CSF; these fmdings are not diagnostic ofbacterial meningitis. However, our study addresses the

question that is of importance to clinicians: in a child withCSF findings that suggest bacterial meningitis, should theinitial antibiotic therapy be chloramphenicol alone or

.chloramphenicol plus penicillin? Fu,thermore, in the 181children whose CSF grew an organism, the mortality in thetwo treatment groups differ~d by only 1.8%::t13 .30/1t(::t95%

confidence limits), which is not a statistically significantdifference.

The Combination of Chloramphenicol and Penicillin

The main reason for giving penicillin in addition to

chloramphenicol in bacterial meningitis is the hope that thepenicillin will cause more rapid killing of S pneumoniae andNeIsseria meningitl"dis. However, chloramphenicol itself is

usually bactericidal agairistboth these organisms (as well asH influenzaef9 and, when it is only bacteriostatic,

chloramphenicol may antagonise the bactericidal action ofpenicillin anyway.l After reviewing the literature onantibiotic combinations, Rahal and SimberkoW9 suggestedthat chloramphenicol plus penicillin is probably equivalentto chloramphenicol alone. Similarly, Lambert2O stated that

there is no reason to believe that the results of combination

therapy of bacterial meningitis (with penicillin,chloramphenicol, and a sulphonamide) are better than thoseobtained with chloramphenicol alone. One reason for givingampicillin in addition to chloramphenicol is the risk of H

influenzae being resistant to chloramphenicol. However,such organisms are very uncommon in most countries and,when they do occur, they are usually resistant to ampicillin aswell as chloramphenicol}1

The results of our trial suggest that chloramphenicol aloneis as effective as chloramphenicol plus penicillin in the

treatment ofbacterial meningitis in children. This conclusionis strengthened by the results of the comparison of

chloramphenicol alone with chloramphenicol plus penicillinin children with severe pneumonia,22 which was carried outin the same hospital wards at the same time as the meningitis

study.Intramuscular injections of chloramphenicol sodium

succinate are not very painful.10-12 2 or 3 days ofintramuscular administration of chloramphenicol, followedby oral administration, may be no more distressing tO a smallchild than a long struggle to insert an intravenous cannula(often on more than one occasion), followed by several days ofbeing tied down so that the drip does not come out.

Intramuscular administration of chloramphenicol alone hasseveral advantages over intravenous therapy with both

chloramphenicol and penicillin, particularly in developingcountries; it is much cheaper, it makes fewer demands on thetime of staff, and it does not carry the risks of sepsis and

overhydration associated with intravenous therapy.We thank Dr Stephanie Germer, Dr Simon Lansdown, Dr Victor

Linnemann, br Christopher Mackenzie, Dr Hilda Polume, Dr AlphonseRongap, Dr Benjamin Tahija, and Dr Likei Theo for their assistance.

Correspondence should be addressed to F. S., Intensive Care Unit, RoyalChildren's Hospital, Parkville, Victoria 3052, Austra\ia.

REFERENCESI Iawetz E, Gunnison IB, Speck RS, Coleman VR Studies on antibiotic synergism and

antagonism. Arch Inrern Med 1951; 87: 349-59.2. Wallace IF, Smith RH, Garcia M, Petersdorf RG. Studies on the pathogenesis of

meningitis. VI. Antagonism between penicillin and chloramphenicol in

experimental pneumococcal meningitis. J Lab Clin Med 1967; 70: 408-18.3 Wehrle PF,MathiesAW, leedomIM, IvlerD. Bacterialmeningitis,AnnNY AcadSci

1967; 145: 488-98.

References conrinued ar foor of nexr page

Discussion

In Papua New Guinea, standard treatment regimens havebeen developed to guide paramedical workers and doctorsjnthe management of the importarit diseases of children.lB Thestandard treatment for meningitis in children must take intoaccount the fact that none of the health centres and only someof the hospitals in Papua New Guinea have facilitie~ forperforming bacterial cultures.

Possible Sources of Error in This Trial

Since most of the children in this trial came from ruralviliages with poor communications, it was not possible tofollow up patients after they had left hospital, and only majorsequelae such as death or gross brain damage could bedetected. However, because the children presented late, theincidence of major sequelae was so high that any importantdifference in the effectiveness of the two antibiotic regimenswould probably have been detected. The investigators knewwhat treatment was being giverito each child in the study, butit is unlikely tha~ this caused bias, because the endpoint in thetrial was whether the patient lived or died. The results mayhave been affected by the high proportion of pneumococciwhich show reduced sensitivity to penicillin in Papua NewGuinea,6,7 because this might have obscured any benefit fromgiving penicillin in addition to chloramphenicol. However,of 49 strains of S pneumoniae isolated from CSF in Gorokaduring the trial, only 10 (20%) had a: penicillin MIC over0.1 /.lg!ml (6 strains 0.2 /.lg/nil, 2 strains 0.5 /.lg!ml, and 2strains 1 .0 /.lg!ml; Gratten M and Shann F, unpublisheddata). .

Random allocation resulted in close matching of the twotreatment groups fo'r all risk factors except length of illness,history of convulsions, and inability to feed. Cochran's testwas used to adjust for the effect of these variables.17 Theadjusted difference in the mortality rate was 7.1 %:!:8.5%(:t95% confidence limits), and the adjusted difference in therate of treatment failure (death, discharged' with braindamage, or absconded still sick) was 0.4%:!:9.1%. Thesedifferences are not statistically significant.

The criteria for inclusion in the trial included increased

fl~~c! rJNI :

Iil ! 1 j

684THEl.ANCET, SEPTEMBER28, 1985

CHLORAMPHENICOL ALONE VERSUSCHLORAMPHENICOL PLUS PENICILLIN FOR

SEVERE PNEUMONJA IN CHILDREN

FRANK S~NNGoroka Hospital, Papua New Guinea

JANE BARKERKundiawa Hospita~ Papua New Guinea

PETER PaDRELae Hospita~ Papua New Guinea

Summary 748 children with severe pneumonia in threehospitals in Papua New Guinea were

randomised to receive intramuscular injections of eitherchloramphenicol alone or chloramphenicol plus penicillin.Sequential ~nalysis showed no difference between the twotreatments. 48 (130/0) of the 377 children in thechloramphenicol alone group died, and 3 {0.8%) werec~a~ged to different treatment. 62 (17%) of the 371 childrenin the chloramphenicol-plus-penicillin group died, and 6( 1 .6%) were changed to different treatme!lt. The differencein failurc; rates (death or withdrawal for cha~ge of treatment)was 4.8%:!:5,2% (:!:95% confidence limits). In children withsevere pneum9nia, treatment with chloramphenicol alone isas effective as treatment with chloramphenicol plus penicil-lin.

Introduction

PNEUMONIA and diarrhoea are the commonest causes ofdeath in children.1 Pneumonia accounts for more than 40% ofdeat.hs in children at Goroka Hospital, in the highlands ofPapua New Guinea, and most cases are caused byHaemophilus injluenzae or Streptococcus pneumonl"ae.2Culture of lung. aspirate or lung biopsy is Jequired todetermine the aetiology of bacterial pneumonia in smallchildren.3 Since lung ~spira:tion is not usually done routinely,

F. SHANN AND OTHERS: REFERENCES-continued

4. Leeper MH, Dowling HF. Treatment of pneumococcic meningitis with penicillin

compare~ with penicillin plus aureomycin. Arch Intern Med 1951; 88: 489-94.5. BodineJ, Murray T,Sande MA. Combination therapy of eXperimental Haemophilus

influenzaeand Strepto-cus pneumoni;'e meningitis. Pediatr Res 1977; 25: 27 A.6 Gratten M, Naraqi S, Hansman D. High prevalence of penicillin-insensitive

pneumococci in Port Moresby, Papua New Guinea. Lancet 1980; ii. 192-95.7. Shann F, Gratten M, Germer S, Linnemann V, Hazlett D, Payne R. Aetiology of

pneumonia in children in Goroka Hospital, Papua New Guinea. Lancet 1984; ii:537-41.

8. Lindberg I, Rosenhall U, Nylen O,Ringner A. Long-term outcome or Haemophilus

influenzae meningiris related to antibiotic treatment. Ped!arrics 1977; 60: 1-6.9. Ware SI, McLauchlan. Haemophilus meningiris in Portsmouth. Lancet J978; ii

197-99.10. Shann F, Linnema~n V, Mackenzie A, Barker I, Polnear L, Crinis N. Absorption of

-chloramphenicol sodium succinate after intramuscular administration in childrenN EnglJ Med 1985; 313: 4JO-14.

II. Ross S, Puig IR, Zaremba EA. Chloramphenicol acid succinate (sodium salt): Some

preliminary clinical and laboratory observations in infants and children. In. We!chH, Marti-Ibanez F, eds. Antibiotics annual 1957-1958. New York: Medica!Encyclopedia, 1958. 837-41.

12. McCrumb FR, Snyder MI, Hic~en WI. The use ofchloramphenicolacid succinate inthe treatment of acute infections In. Welch H, Marti-Ibanez F, eds. Antibiotics

annua11957-58. New York. Medical Encyclopaedia, 1958: 837-41.!3 Stratton CW. Infection related to intravenous infusions. Heart Lung 1982; 11: 123-3514. Kaplan SL, Feigin RD The syndrome of inappropriate secretion ofantiduretic

hormone in children with bacterial meningitis. J Pediatr 1978; 92: 758-6115 Shann F, Germer S. Hyponatraemia associated with pneumonia or bacterial

meningitis Arch Dis Child (in press).16. Armitage P. Sequential medical trials Oxford: Blackwell, 1974: 5417. Armitage P. Statistical methods in medical research: Oxford. Blackwell, 197J. 369-72.}8. Papua New Guinea Paediatric Society. Standard treatment for common illnesses of

children in Papua New Guinea. Pon Moresby. Papua New Guinea Paediatric

Society,198419 Rahal II, Simberkoff MS Bactericidal and bacteriostatic action of chloramphenicol

against meningeal pathogens. AnlimicrobAg Chemorher 1979; 16: 13-18.20. Lamben HP. Use ofantibiotics. Meningitis. BrMedJ 1978; ii. 259-61.21 Anonymous Ampicillin and chloramphenicol resistance in systemic Haemophilus

i,iflunzae disease. MMWR 1984; 33: 35-3722. Shann F, Barker I, Poorc P. Chloramphenicol alone versus ~hloramphenicol plus

penicillin for severe pneumonia in children. Lancet J985.

children with severe pne~monia are often treated with both

penicillin (for S pneumoniae) and chloramphenicol (forH injluenzae).

When chloramphenicol is given because of the possibilityof' H injluenzae infection, what is the effeCt of adding

penicillin to cover S pneumoniae? A study in dogs with

pneumococcal meningitis suggested that chloramphenicolplus penicillin may be superior to chloramphenicol alone forsevere pneumococcal infeCtions.4 However, in two

retrospeCtive studies in children with H injluenzaemeningitis,5,6 sequelae were 4 times more common aft~rtreatment with chloramphenicol plus ampicillin than aft~r

chloramph~nicol alon~.A multicentre prospectiv~ randomisedcontrolled trial has

be~n carried out in children with sever~ pn~umonia in PapuaN~w Guinea to see whether chloramphenicol plus p~nicillinis any more effective than chloramphenicol alone.

Methods

Childr~n with s~ver~ pneumonia were studied at GorokaHospital, Kundiawa HospitaJ, and Lae Hospital between May,1979, and April, 1982. Children were admitt~dto the trial if theyhad cough and intercostal recession with any of the following:severe intercostal recession, pulse rate over 160/min with

hepatomegaly, irnibi)ity to feed, bronchial breathing, grunting,cyanosis, sev~re chest X-ray chang~s, or a total whit~ c~11 count(WCC) over 30 000 cellsl,ul, Children with meningitis wereexcluded. The study was approved by the Papua New Guin~aMedical Research Advisory Committee.

A table of random numbers was used to prepare sealed, numberedenvelopes. After each child had been admitted to the trial, the nextenvelope was opened to determine the treatment to be given. Theinvestigators were aware of the treatment being given to each child.Chloramphenicol sodium succinate was given intramuscularly in adose of 25 fig/kg every 6 h until the child began to improve, andthen 25 mg/kg of chloramphenicol palmitate was given orally every6h. Benzylpenicillin was given intramuscularly every 6 h in a doseof250 OOOu~its to children weighi~g3-9 kg, and 500000 units tochildren weighing 10-19 kg. All children were given quinine or

amodiaquine.Each child was paired with the next child in the other treatment

group at the same hospitaLTreatment was said to have failed if thechild died or if the antibiotic therapy was changed because ofdeterioration in the child's condition. The preferences of untiedpairs were analysed by sequential analysis on a repeated significancetest pJan with 2a <0.05, 1-{J >0.95, el =0.70, andn= 104! Thus,there was a 95% chance of obtaining a significant result (p<O .05)provided that the probability of a preference for 1 treatment groupover the other was 0; 70 or more.

Results

Of the 748 children admitted to the trial, 455 were in

Goz:oka (where I child in the chloramphenicol alone groupwas unpair~d), 174 wer~ in Kundiawa (2 chloramphenicolalone unpaired), and 119 were in Lae (3 chloramphenicolalone unpaired). Of the 371 pairs ofresults, 272 were tied, 58

favoured chloramphenicol alone, and 4 J favoured

chioramphenicol plus penicillin (see figure). 9 children (3chloramphenicol alone) had their treatment changed duringthe trial because their condition was deteriorating (see table);these children were counted as treatment failures in the

sequential analysis. The border of the sequential-analysisgraph was reached after the 87th p;eference had b~endecided, at a point ipdicating ho significant differencebetWeen the two tr~atment groups. 564 (77%) ofth~ 748childr~n in th~ trial were less than 12 months old, and 676

(92%) were less than 24 months old.

Randomisation resulted in good matching of the childrenin the, two treatment groups: there was no significant