- Email: [email protected]
PHENOBARBITONE TO PREVENT PERIVENTRICULAR HAEMORRHAGE IN VERY-LOW-BIRTHWEIGHT BABIES
285 In well-controlled diabetes glycosylate haemoglobin is too insensitive to be used as a measure of diabetic control. The relation between mean glucose and birthweight centile suggests that, with the control achieved, we had, not the mimicking of a "prediabetic" state, as suggested by Knight et al, but the mimicking of gestational diabetes. Almost normal blood glucose levels are not good enough to prevent all diabetic complications of pregnancy. Even with CSII it will be difficult to reach a complete normoglycaemia. Department of Obstetrics and Clinical Endocrinology, University Hospital, Groningen, Netherlands
G. H. A. VISSER E. VAN BALLEGOOIE W. J. SLUITER
SIR,-We describe here the infant of a diabetic mother who was born with macrosomy and hypertrophic despite very tight control of diabetes mellitus. Hypertrophic cardiomyopathy is found in up to 50% of cases when systematic studies are done in babies born to diabetic course of the disease is usually benign. The mothers.2development of cardiac septal hypertrophy has always been control2-5 or with correlated with poor maternal hyperinsulinaemia in the newborn.6 This diabetic woman (modified White class B) had been managed by a subcutaneous insulin-infusion system (’Auto syringe’) from before conception,and the basal rate was increased during pregnancy from 15 to 30 distance amounts. Careful home monitoring with four daily ’Dextrostix’ blood glucose measuremerits was done and every week the patient was seen in
cardiomyopathy
echocardiographic The
diabetic
outpatient. Glycosylated haemoglobin levels were 12-13% before pump treatment and fell to 7-4-9-4% (with one exception, 10 - I%, at 6 month of pregnancy). Despite the attempt at very tight metabolic control the woman experienced three short spells of hyperglycaemia-one during the second week of pregnancy lasting 2 days; one in the twelth week of pregnancy lasting3 days; and a rise in blood glucose up to 250 mg/dl 24 h before delivery. Neither acidosis nor other metabolic hospital
as an
abnormalities were ever found. At 38 weeks a boy was delivered by elective caesarean section. He had a typical macrosomic appearance, weighed more than expected for gestational age (3’4kg), and had a systolic murmur grade li/vi with maximum intensity at the left sternal border. His blood glucose was 39 mg/dl. The boy was moved to the neonatal intensive care unit. His blood glucose fell to 15 mg/dl 2 h.after delivery and infusions with glucose concentrations up to 15% were required. Supplementary glucose infusions were necessary for the first three days of life. Transient hypocalcaemia (7-99 mg/dl) was noted 12 h after delivery with mild excitability, which reverted promptly on treatment with calcium gluconate. No dyspnoea, lethargy, convulsions, or signs of cardiac decompensation were seen. Typical 2 echocardiographic signs2 were present. No biochemical abnormalities were noted after the fourth day. The murmur disappeared progressively and the echocardiogram was almost normal after one week, except for some residual thickening of the interventricular septum. Thus, even under near optimal regulation of diabetes during pregnancy, the newborn baby may have the syndrome of hyperglycaemia, hypocalcaemia, macrosomy, and cardiomyopathy. Factors other than strict control of maternal blood glucose could be responsible for this syndrome. The fact that insulin administered by ,
1. Cowett 2. 3. 4.
5. 6. 7.
RM, Schwarts R. The infant of the diabetic mother. Pediatr Clin 1982; 29: 1213-31 Mace S, Hirschfeld SS, Riggs T, Fanaroff AA, Merkatz IR. Echocardiographic abnormalities in infants of diabetic mothers. J Pediatr 1979; 95: 1013-19. Halliday HL. Hypertrophic cardiomyopathy in infants of poorly-controlled diabetic mothers Arch Dis Childh 1981; 56: 258-63. Lusson JR, Gaulme J, Raynaud EJ, Cheynel J. Hypertrophie septale asymetrique du nouveau-né de mere diabetique. Arch Fr Pediatr 1982; 39: 433-36. Way GL, Wolfe RR, Eshagpour E, et al. The natural history of hypertrophic cardiomyopathy in infants of diabetic mothers J Pediatr 1979; 956: 1020-25. Breitweser JA, Meyer RA, Sperling MA, Tsang RC, Kaplan S. Cardiac septal hypertrophy in hyperinsulinemic infants. J Pediatr 1980; 963: 535-39. Mills JL, Baker L, Goldman AS. Malformations in infants of diabetic mothers occur before the seventh gestational week: Implications for treatment. Diabetes 1979; 28: 292-93.
subcutaneous infusion does not pass the portal system, as does endogeneously secreted hormone, is probably of prime importance. Divisions of Paediatric Cardiology and Paediatric Endocrinology,
Department of Paediatrics; Department of Endocrinology (Internal Medicine); and the Department of Obstetrics and
Gynaecology,
University Hospital, Ghent State University, 9000 Ghent, Belgium
HENRI A. VERHAAREN MARGARITA CRAEN PATRICK DE GOMME ROBERT RUBENS WALTER PAREWIJCK ROBERT DEROM
PHENOBARBITONE TO PREVENT
PERIVENTRICULAR HAEMORRHAGE IN VERY-LOW-BIRTHWEIGHT BABIES SIR,-The Hammersmith Hospital group has an international
reputation for their studies of periventricular haemorrhage (PVH) in very-low-birthweight infants. Their Nov 19 (p 1168) paper has prompted the following thoughts. "Proper methodology for randomised control trials requires careful planning before the trial begins. This planning must consider what the size of the sample,ought to be for an important clinical therapy effect to be detected if it exists".1 Dr Whitelaw and his associates would need a sample three times larger than the one they studied to appreciate a 50% reduction in the incidence of PVH from 36 - 6% to.18’ 3%. Appreciation of a less prominent effect of the phenobarbitone treatment requires an even larger sample. Thus, the results should be interpreted cautiously. "Concerns for covariates balance often arise first from the desire without bias the effect of treatment: that is, to avoid the slanting results in one way or the other. However, a second important consideration is that covariance adjustment can improve the precision of estimates of effects, even when balance is excellent".12 of the phenobarbitone group had a 1 min Apgar score of less than 3, compared with only 9 in the control group. Without log linear or logistic regression modelling to reduce bias and to increase precision,, it is unclear whether or not phenobarbitone did have some (although small) beneficial effect. That is, steps were not taken to consider whether or not the treated group was more ill and thereby more susceptible to PVH. If the groups are not comparable, then an alternative interpretation might to assess
be appropriate. Whitelaw et al did not ,
plan this clinical trial to determine if phenobarbitone treatment increased the risk of ventilatory dependency. In attaching a "p" value to an observed association they may lead some readers to think that this hypothesis was tested. An alternative interpretation of the data is that the greater need for
ventilator support in the phenobarbitone treated group was related the severity of pulmonary disease. The notion that the treated group was more ill is supported by the ventilator requirement and the Apgar data. The provocative hypothesis generated by Whitelaw and his colleagues deserves to be tested in a different sample. to
Departments of Neurology and Neuroepidemiology, Children’s Hospital, Boston, Massachusetts 02115, USA
KARL C. KUBAN ALAN LEVITON
***This letter has been shown to Dr Whitelaw and his colleagues, reply follows.—ED. L. l SIR,-We had randomised 60 infants because Donn et all in reduction demonstrated a significant periventricular haemorrhage (PVH) at the 1% level in their open trial of phenobarbitone in 60 infants. 36’ 6% of our placebo-treated infants developed PVH. Finding that 40% of the phenobarbitone-treated infants in the study had PVH did not suggest to us that increase in the trial population would be likely to show benefit from
whose
phenobarbitone. Because we had clear evidence of respiratory depression following phenobarbitone after 60 babies had been enrolled in our study, we felt that the trial should be terminated. We 1.
2.
Freiman JA, Chalmers TC, Smith H, Jr, Kuebler RR The importance ofbeta, the type II error and sample size in the design and interpretation of the randomized control trial: Survey of 71 "negative" trials. N Engl J Med 1978; 299: 690-94. Lavori PW, Louis TA, Bailar JC III, Polansky M. Statistics in practice: Designs for experiments: Parallel comparisons of treatment. N Engl J Med 1983,309: 1291 -99.
.
286 agree that well over 100 babies would be needed to evaluate the possibility of a preventiveI effect on PVH considerably smaller than that found by Donn et al. Dr Kuban and Dr Leviton go on to claim that we did not consider whether the phenobarbitone groups was more ill and therefore more susceptible to PVH. Although 12 of the phenobarbitone group had 1 min Apgar scores of less than 3 (as against 9 in the placebo group) 5 of the placebo group had an Apgar score of below 7 at 5 min (as against 3 in the phenobarbitone group). When the 1 min and 5 min Apgar scores are considered together there is no overall difference between the two groups, nor is there for gestational age, birthweight, sex, respiratory failure before the injection, pneumothorax, hypercapnia, acidosis, and vaginal delivery. Thus the randomization produced very even matching. We cannot conclude that the increase in ventilator dependency after phenobarbitone injection was explained by differences in Apgar scores or pulmonary disease. We affirm that 20 mg/kg of phenobarbitone may have an adverse effect on spontaneously breathing, very-low-birthweight infants, and we await with interest the results of further studies of phenobarbitone in ventilated
truly
National Perinatal Epidemiology Unit, Radcliffe Infirmary, Oxford OX2 6HE
infants. Department of
Paediatrics and Neonatal Medicine, Royal Postgraduate Medical School, Hammersmith Hospital, London W12 0HS
ANDREW WHITELAW LILLY DUBOWITZ MALCOLM LEVENE
SiR.—Dr Whitelaw and colleagues’ is the latest of four trials2-5
to
the role of phenobarbitone in preventing intraventricular and cases are being recruited for at least two other trials. In view of the small size of the published trials, it is not surprising that investigators have come to different conclusions about the value of phenobarbitone. There are two important consequences of small trials. Firstly, real and clinically significant differences between treatments may be incorrectly dismissed because they fail to achieve statistical significance. Secondly, if statistically significant differences are observed, these will tend to overestimate the true differences. Whitelaw et al make the important observation that in all four trials there have been fewer parenchymal haemorrhages among phenobarbitone-treated infants than among controls (table). The summary relative risk (0 - 36) of parenchymal haemorrhage in babies 6 given phenobarbitone, calculated by the Mantel-Haenszel method on data from all four trials, suggests that the incidence of this most serious form of periventricular haemorrhage may be substantially reduced by prophylactic phenobarbitone. As Whitelaw and his colleagues have observed, however, this difference does not achieve conventional levels of significance (p 0 - 07; 95% confidence limits relative risk x 0-12 -1 - 09). We hope that this analysis may serve as a reminder of the dangers of false inference from both non-randomised comparisonsand small randomised trials. Many trials in the perinatal period require sample sizes larger than any single unit can generate within a reasonable length of time. A recognition of this reality, combined with the impact of Whitelaw’s diplomacy, has resulted in more than assess
haemorrhage,
=
1. Donn
dozen neonatal units joining together to enter cases into the first collaborative trial of neonatal practice ever to have been mounted in Britain. The objective of this is to assess alternative ways of managing post-haemorrhagic ventricular dilatation-but there is no reason why it could not provide a framework for collaboration to address other clinically important questions. A high priority among these must surely be to assess whether phenobarbitone or other drugs9,10 can reduce the incidence of parenchymal haemorrhage in very low birthweight infants. a
S, Roloff D, Goldstein G. Prevention of intraventricular haemorrhage in preterm infants by phenobarbitone. Lancet 1981; ii: 215-17. 2. Goldstein G, Donn S, Roloff D. Further observations on the use of phenobarbital to prevent neonatal intracranial hemorrhage. In: 2nd Special Ross Conference in Perinatal Intracranial Hemorrhage (Washington, DC). Columbus, Ohio: Ross Laboratories, 1982: 810-15. 3. Cooke RWI, Morgan MEM, Massey RF. Phenobarbitone to prevent intraventricular haemorrhage. Lancet 1981; ii: 414-15. 4. Bedard M, Shankaran S, Slovis T, Pantoja A, Dayal B, Poland R. Decreased severity of periventricular haemorrhage following phenobarbital treatment. In: 2nd Special Ross Conference in Perinatal Intracranial Hemorrhage (Washington, DC). Columbus, Ohio: Ross Laboratories, 1982: 1051-64 5 Whitelaw A, Placzek M, Dubowitz L, Lary S, Levene M Phenobarbitone for prevention of periventricular haemorrhage in very low birth weight infants: A randomised double-blind trial. Lancet 1983; ii: 1168-70. 6. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959; 22: 710-48. 7. Miettinen O. Estimability and estimation in case-referent studies. Am J Epidemiol 1976; 103: 226-35. 8 Hope PL, Stewart AL, Thornburn RJ, Whitehead MD, Reynolds EOR, Lowe D. Failure of phenobarbitone to prevent intraventricular haemorrhage in small preterm infants. Lancet 1982; i. 444-45.
IAIN CHALMERS DIANA ELBOURNE ADRIAN GRANT
PANCURONIUM AND PNEUMOTHORAX and colleagues (Jan 7, p 1) report a highly in the incidence of pneumothorax in ventilated babies treated with pancuronium and in whom a specific respiratory pattern had been identified. Encouraged by a preliminary report of their work, but unable to do the respiratory function tests Greenough et al employed, we used early pancuronium paralysis selectively in ventilated infants of less than 32 weeks’ gestation who struggled against the ventilator or who required peak airway pressures of more than 20 cm water or ventilation rates above 30/min. On these criteria, 50% of our series of infants were paralysed, compared with 40% in the Cambridge study. We compared the results of the first 6 months (July to December, 1983) of this new "policy" with those of the two preceding 6 month periods when pancuronium was used only when other methods had failed to produce adequate gas exchange. Infants of 26 weeks’ gestation and less are considered separately from those of 27-32 weeks’ gestation (table). In the smallest infants no significant change was seen in frequency of pneumothorax, survival, or intraventricular haemorrhage (IVH) after the introduction of the new policy. In the larger infants the frequency of pneumothorax fell significantly, and there was a reduction (not significant) in mortality. Disappointingly, there was no change in the incidence of IVH. There seems to be an association between IVH and previous pneumothorax, and mechanisms linking the two directly have been suggested. 1-3 Prevention of pneumothorax would seem a practical way of reducing perinatal brain damage and death. Perhaps these studies are too small to show such an effect. IVH is multifactorial in origin, and pneumothorax may be the preceding factor in only 16% of cases in very-low-birthweight infants.4Alternatively, factors such as hypercapnia and acidosis rather than the pneumothorax itself may be more important in the with IVH than aetiology of IVH. Both are more strongly associated 4 pneumothorax when treated as single variables.4
SIR,-Dr Greenough significant reduction
9
Morgan MEI, Benson JWI, Cooke RWI. Ethamsylate reduces the incidence of periventricular haemorrhage in very low birth weight babies. Lancet 1981; ii 830-31. .
10. Chiswick ML, Johnson M, Woodhall C, Gowland M, Davies J, Toner N, Sims D. Protective effect of vitamin E (DL-alpha-tocopherol) against intraventricular haemorrhage in premature babies. Br Med J 1983; 287: 81-84. 1. Lipscomb AP, Thorburn RJ, Reynolds EOR, et al Pneumothorax and cerebral haemorrhage in preterm infants. Lancet 1981; i: 414-16. 2 Dykes FD, Layara A, Ahmann P, Blumenstein B, Schwartz J, Brann AW. Intraventricular hemorrhage: a prospective evaluation of etiopathogenesis. Pediatrics 1980; 66: 42-49 3. Hill A, Perlmann JM, Volpe JJ. Relationship of pneumothorax to occurrence of intraventricular hemorrhage in the premature newborn. Pediatrics 1982; 69: 144-49. 4. Cooke RWI, Morgan MEI. Precipitants of neonatal periventricular hemorrhage. Variability with postnatal age In: Proceedings of 2nd Perinatal Intracranial Hemorrhage Conference, 1982. Columbus, Ohio: Ross Laboratories (in press).
G. H. A. VISSER E. VAN BALLEGOOIE W. J. SLUITER
SIR,-We describe here the infant of a diabetic mother who was born with macrosomy and hypertrophic despite very tight control of diabetes mellitus. Hypertrophic cardiomyopathy is found in up to 50% of cases when systematic studies are done in babies born to diabetic course of the disease is usually benign. The mothers.2development of cardiac septal hypertrophy has always been control2-5 or with correlated with poor maternal hyperinsulinaemia in the newborn.6 This diabetic woman (modified White class B) had been managed by a subcutaneous insulin-infusion system (’Auto syringe’) from before conception,and the basal rate was increased during pregnancy from 15 to 30 distance amounts. Careful home monitoring with four daily ’Dextrostix’ blood glucose measuremerits was done and every week the patient was seen in
cardiomyopathy
echocardiographic The
diabetic
outpatient. Glycosylated haemoglobin levels were 12-13% before pump treatment and fell to 7-4-9-4% (with one exception, 10 - I%, at 6 month of pregnancy). Despite the attempt at very tight metabolic control the woman experienced three short spells of hyperglycaemia-one during the second week of pregnancy lasting 2 days; one in the twelth week of pregnancy lasting3 days; and a rise in blood glucose up to 250 mg/dl 24 h before delivery. Neither acidosis nor other metabolic hospital
as an
abnormalities were ever found. At 38 weeks a boy was delivered by elective caesarean section. He had a typical macrosomic appearance, weighed more than expected for gestational age (3’4kg), and had a systolic murmur grade li/vi with maximum intensity at the left sternal border. His blood glucose was 39 mg/dl. The boy was moved to the neonatal intensive care unit. His blood glucose fell to 15 mg/dl 2 h.after delivery and infusions with glucose concentrations up to 15% were required. Supplementary glucose infusions were necessary for the first three days of life. Transient hypocalcaemia (7-99 mg/dl) was noted 12 h after delivery with mild excitability, which reverted promptly on treatment with calcium gluconate. No dyspnoea, lethargy, convulsions, or signs of cardiac decompensation were seen. Typical 2 echocardiographic signs2 were present. No biochemical abnormalities were noted after the fourth day. The murmur disappeared progressively and the echocardiogram was almost normal after one week, except for some residual thickening of the interventricular septum. Thus, even under near optimal regulation of diabetes during pregnancy, the newborn baby may have the syndrome of hyperglycaemia, hypocalcaemia, macrosomy, and cardiomyopathy. Factors other than strict control of maternal blood glucose could be responsible for this syndrome. The fact that insulin administered by ,
1. Cowett 2. 3. 4.
5. 6. 7.
RM, Schwarts R. The infant of the diabetic mother. Pediatr Clin 1982; 29: 1213-31 Mace S, Hirschfeld SS, Riggs T, Fanaroff AA, Merkatz IR. Echocardiographic abnormalities in infants of diabetic mothers. J Pediatr 1979; 95: 1013-19. Halliday HL. Hypertrophic cardiomyopathy in infants of poorly-controlled diabetic mothers Arch Dis Childh 1981; 56: 258-63. Lusson JR, Gaulme J, Raynaud EJ, Cheynel J. Hypertrophie septale asymetrique du nouveau-né de mere diabetique. Arch Fr Pediatr 1982; 39: 433-36. Way GL, Wolfe RR, Eshagpour E, et al. The natural history of hypertrophic cardiomyopathy in infants of diabetic mothers J Pediatr 1979; 956: 1020-25. Breitweser JA, Meyer RA, Sperling MA, Tsang RC, Kaplan S. Cardiac septal hypertrophy in hyperinsulinemic infants. J Pediatr 1980; 963: 535-39. Mills JL, Baker L, Goldman AS. Malformations in infants of diabetic mothers occur before the seventh gestational week: Implications for treatment. Diabetes 1979; 28: 292-93.
subcutaneous infusion does not pass the portal system, as does endogeneously secreted hormone, is probably of prime importance. Divisions of Paediatric Cardiology and Paediatric Endocrinology,
Department of Paediatrics; Department of Endocrinology (Internal Medicine); and the Department of Obstetrics and
Gynaecology,
University Hospital, Ghent State University, 9000 Ghent, Belgium
HENRI A. VERHAAREN MARGARITA CRAEN PATRICK DE GOMME ROBERT RUBENS WALTER PAREWIJCK ROBERT DEROM
PHENOBARBITONE TO PREVENT
PERIVENTRICULAR HAEMORRHAGE IN VERY-LOW-BIRTHWEIGHT BABIES SIR,-The Hammersmith Hospital group has an international
reputation for their studies of periventricular haemorrhage (PVH) in very-low-birthweight infants. Their Nov 19 (p 1168) paper has prompted the following thoughts. "Proper methodology for randomised control trials requires careful planning before the trial begins. This planning must consider what the size of the sample,ought to be for an important clinical therapy effect to be detected if it exists".1 Dr Whitelaw and his associates would need a sample three times larger than the one they studied to appreciate a 50% reduction in the incidence of PVH from 36 - 6% to.18’ 3%. Appreciation of a less prominent effect of the phenobarbitone treatment requires an even larger sample. Thus, the results should be interpreted cautiously. "Concerns for covariates balance often arise first from the desire without bias the effect of treatment: that is, to avoid the slanting results in one way or the other. However, a second important consideration is that covariance adjustment can improve the precision of estimates of effects, even when balance is excellent".12 of the phenobarbitone group had a 1 min Apgar score of less than 3, compared with only 9 in the control group. Without log linear or logistic regression modelling to reduce bias and to increase precision,, it is unclear whether or not phenobarbitone did have some (although small) beneficial effect. That is, steps were not taken to consider whether or not the treated group was more ill and thereby more susceptible to PVH. If the groups are not comparable, then an alternative interpretation might to assess
be appropriate. Whitelaw et al did not ,
plan this clinical trial to determine if phenobarbitone treatment increased the risk of ventilatory dependency. In attaching a "p" value to an observed association they may lead some readers to think that this hypothesis was tested. An alternative interpretation of the data is that the greater need for
ventilator support in the phenobarbitone treated group was related the severity of pulmonary disease. The notion that the treated group was more ill is supported by the ventilator requirement and the Apgar data. The provocative hypothesis generated by Whitelaw and his colleagues deserves to be tested in a different sample. to
Departments of Neurology and Neuroepidemiology, Children’s Hospital, Boston, Massachusetts 02115, USA
KARL C. KUBAN ALAN LEVITON
***This letter has been shown to Dr Whitelaw and his colleagues, reply follows.—ED. L. l SIR,-We had randomised 60 infants because Donn et all in reduction demonstrated a significant periventricular haemorrhage (PVH) at the 1% level in their open trial of phenobarbitone in 60 infants. 36’ 6% of our placebo-treated infants developed PVH. Finding that 40% of the phenobarbitone-treated infants in the study had PVH did not suggest to us that increase in the trial population would be likely to show benefit from
whose
phenobarbitone. Because we had clear evidence of respiratory depression following phenobarbitone after 60 babies had been enrolled in our study, we felt that the trial should be terminated. We 1.
2.
Freiman JA, Chalmers TC, Smith H, Jr, Kuebler RR The importance ofbeta, the type II error and sample size in the design and interpretation of the randomized control trial: Survey of 71 "negative" trials. N Engl J Med 1978; 299: 690-94. Lavori PW, Louis TA, Bailar JC III, Polansky M. Statistics in practice: Designs for experiments: Parallel comparisons of treatment. N Engl J Med 1983,309: 1291 -99.
.
286 agree that well over 100 babies would be needed to evaluate the possibility of a preventiveI effect on PVH considerably smaller than that found by Donn et al. Dr Kuban and Dr Leviton go on to claim that we did not consider whether the phenobarbitone groups was more ill and therefore more susceptible to PVH. Although 12 of the phenobarbitone group had 1 min Apgar scores of less than 3 (as against 9 in the placebo group) 5 of the placebo group had an Apgar score of below 7 at 5 min (as against 3 in the phenobarbitone group). When the 1 min and 5 min Apgar scores are considered together there is no overall difference between the two groups, nor is there for gestational age, birthweight, sex, respiratory failure before the injection, pneumothorax, hypercapnia, acidosis, and vaginal delivery. Thus the randomization produced very even matching. We cannot conclude that the increase in ventilator dependency after phenobarbitone injection was explained by differences in Apgar scores or pulmonary disease. We affirm that 20 mg/kg of phenobarbitone may have an adverse effect on spontaneously breathing, very-low-birthweight infants, and we await with interest the results of further studies of phenobarbitone in ventilated
truly
National Perinatal Epidemiology Unit, Radcliffe Infirmary, Oxford OX2 6HE
infants. Department of
Paediatrics and Neonatal Medicine, Royal Postgraduate Medical School, Hammersmith Hospital, London W12 0HS
ANDREW WHITELAW LILLY DUBOWITZ MALCOLM LEVENE
SiR.—Dr Whitelaw and colleagues’ is the latest of four trials2-5
to
the role of phenobarbitone in preventing intraventricular and cases are being recruited for at least two other trials. In view of the small size of the published trials, it is not surprising that investigators have come to different conclusions about the value of phenobarbitone. There are two important consequences of small trials. Firstly, real and clinically significant differences between treatments may be incorrectly dismissed because they fail to achieve statistical significance. Secondly, if statistically significant differences are observed, these will tend to overestimate the true differences. Whitelaw et al make the important observation that in all four trials there have been fewer parenchymal haemorrhages among phenobarbitone-treated infants than among controls (table). The summary relative risk (0 - 36) of parenchymal haemorrhage in babies 6 given phenobarbitone, calculated by the Mantel-Haenszel method on data from all four trials, suggests that the incidence of this most serious form of periventricular haemorrhage may be substantially reduced by prophylactic phenobarbitone. As Whitelaw and his colleagues have observed, however, this difference does not achieve conventional levels of significance (p 0 - 07; 95% confidence limits relative risk x 0-12 -1 - 09). We hope that this analysis may serve as a reminder of the dangers of false inference from both non-randomised comparisonsand small randomised trials. Many trials in the perinatal period require sample sizes larger than any single unit can generate within a reasonable length of time. A recognition of this reality, combined with the impact of Whitelaw’s diplomacy, has resulted in more than assess
haemorrhage,
=
1. Donn
dozen neonatal units joining together to enter cases into the first collaborative trial of neonatal practice ever to have been mounted in Britain. The objective of this is to assess alternative ways of managing post-haemorrhagic ventricular dilatation-but there is no reason why it could not provide a framework for collaboration to address other clinically important questions. A high priority among these must surely be to assess whether phenobarbitone or other drugs9,10 can reduce the incidence of parenchymal haemorrhage in very low birthweight infants. a
S, Roloff D, Goldstein G. Prevention of intraventricular haemorrhage in preterm infants by phenobarbitone. Lancet 1981; ii: 215-17. 2. Goldstein G, Donn S, Roloff D. Further observations on the use of phenobarbital to prevent neonatal intracranial hemorrhage. In: 2nd Special Ross Conference in Perinatal Intracranial Hemorrhage (Washington, DC). Columbus, Ohio: Ross Laboratories, 1982: 810-15. 3. Cooke RWI, Morgan MEM, Massey RF. Phenobarbitone to prevent intraventricular haemorrhage. Lancet 1981; ii: 414-15. 4. Bedard M, Shankaran S, Slovis T, Pantoja A, Dayal B, Poland R. Decreased severity of periventricular haemorrhage following phenobarbital treatment. In: 2nd Special Ross Conference in Perinatal Intracranial Hemorrhage (Washington, DC). Columbus, Ohio: Ross Laboratories, 1982: 1051-64 5 Whitelaw A, Placzek M, Dubowitz L, Lary S, Levene M Phenobarbitone for prevention of periventricular haemorrhage in very low birth weight infants: A randomised double-blind trial. Lancet 1983; ii: 1168-70. 6. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959; 22: 710-48. 7. Miettinen O. Estimability and estimation in case-referent studies. Am J Epidemiol 1976; 103: 226-35. 8 Hope PL, Stewart AL, Thornburn RJ, Whitehead MD, Reynolds EOR, Lowe D. Failure of phenobarbitone to prevent intraventricular haemorrhage in small preterm infants. Lancet 1982; i. 444-45.
IAIN CHALMERS DIANA ELBOURNE ADRIAN GRANT
PANCURONIUM AND PNEUMOTHORAX and colleagues (Jan 7, p 1) report a highly in the incidence of pneumothorax in ventilated babies treated with pancuronium and in whom a specific respiratory pattern had been identified. Encouraged by a preliminary report of their work, but unable to do the respiratory function tests Greenough et al employed, we used early pancuronium paralysis selectively in ventilated infants of less than 32 weeks’ gestation who struggled against the ventilator or who required peak airway pressures of more than 20 cm water or ventilation rates above 30/min. On these criteria, 50% of our series of infants were paralysed, compared with 40% in the Cambridge study. We compared the results of the first 6 months (July to December, 1983) of this new "policy" with those of the two preceding 6 month periods when pancuronium was used only when other methods had failed to produce adequate gas exchange. Infants of 26 weeks’ gestation and less are considered separately from those of 27-32 weeks’ gestation (table). In the smallest infants no significant change was seen in frequency of pneumothorax, survival, or intraventricular haemorrhage (IVH) after the introduction of the new policy. In the larger infants the frequency of pneumothorax fell significantly, and there was a reduction (not significant) in mortality. Disappointingly, there was no change in the incidence of IVH. There seems to be an association between IVH and previous pneumothorax, and mechanisms linking the two directly have been suggested. 1-3 Prevention of pneumothorax would seem a practical way of reducing perinatal brain damage and death. Perhaps these studies are too small to show such an effect. IVH is multifactorial in origin, and pneumothorax may be the preceding factor in only 16% of cases in very-low-birthweight infants.4Alternatively, factors such as hypercapnia and acidosis rather than the pneumothorax itself may be more important in the with IVH than aetiology of IVH. Both are more strongly associated 4 pneumothorax when treated as single variables.4
SIR,-Dr Greenough significant reduction
9
Morgan MEI, Benson JWI, Cooke RWI. Ethamsylate reduces the incidence of periventricular haemorrhage in very low birth weight babies. Lancet 1981; ii 830-31. .
10. Chiswick ML, Johnson M, Woodhall C, Gowland M, Davies J, Toner N, Sims D. Protective effect of vitamin E (DL-alpha-tocopherol) against intraventricular haemorrhage in premature babies. Br Med J 1983; 287: 81-84. 1. Lipscomb AP, Thorburn RJ, Reynolds EOR, et al Pneumothorax and cerebral haemorrhage in preterm infants. Lancet 1981; i: 414-16. 2 Dykes FD, Layara A, Ahmann P, Blumenstein B, Schwartz J, Brann AW. Intraventricular hemorrhage: a prospective evaluation of etiopathogenesis. Pediatrics 1980; 66: 42-49 3. Hill A, Perlmann JM, Volpe JJ. Relationship of pneumothorax to occurrence of intraventricular hemorrhage in the premature newborn. Pediatrics 1982; 69: 144-49. 4. Cooke RWI, Morgan MEI. Precipitants of neonatal periventricular hemorrhage. Variability with postnatal age In: Proceedings of 2nd Perinatal Intracranial Hemorrhage Conference, 1982. Columbus, Ohio: Ross Laboratories (in press).