Childhood leukaemia and intramuscular vitamin K: findings from a case-control study

BMJ 1996; 313 doi: (Published 27 July 1996) Cite this as: BMJ 1996;313:204
  1. Pat Ansell, senior research nursea,
  2. Diana Bull, research officera,
  3. Eve Roman, senior scientista
  1. a Cancer Epidemiology Unit, Imperial Cancer Research Fund, Radcliffe Infirmary, Oxford OX2 6HE
  1. Correspondence to: Dr E Roman, Leukaemia Research Fund, Clinical Epidemiology Unit, Leeds LS2 9NG.
  • Accepted 12 July 1996

Exposures before birth and in early life have long been thought to be important determinants of cancer in children. Anxiety about the neonatal administration of vitamin K was raised in 1992, when Golding et al linked intramuscular, but not oral, vitamin K with childhood malignancy.1 Of particular concern was the 2.65-fold increased risk of leukaemia.1 Much debate ensued, and, although Golding's findings have not been confirmed elsewhere,2 3 4 much public anxiety remains. We present the findings relating to vitamin K from a case-control study designed to investigate associations between leukaemia and prenatal and neonatal exposures.

Subjects, methods, and results

Cases comprise children (0-14 years) diagnosed with leukaemia whose mothers' obstetric notes are stored at the John Radcliffe Hospital, Oxford (born 1951 or later), the Rosie Maternity Hospital, Cambridge (born 1956 or later), and the Royal Berkshire Hospital, Reading (born 1969 or later). Of the eligible cases identified by the Childhood Cancer Research Group,5 records covering delivery were found for 90% and obstetric notes for 85%. Two controls per case (matched on sex, hospital, and year and month of birth) were randomly selected from registers of all births held at the relevant hospitals. Multiple pregnancies, children with chromosomal anomalies, and babies who died before discharge were omitted. Delivery, obstetric, and neonatal records were abstracted by midwives using specially designed structured forms. Results are given for acute lymphoblastic leukaemia alone and all leukaemias combined. Findings are presented in two ways: firstly, by what was written in the notes (deriving route from hospital practice when it was not recorded); and, secondly, by what could be imputed from hospital policy. To ensure comparability with the results of others,1 2 3 4 we present odds ratios for intramuscular vitamin K versus all other possibilities combined (none/oral/no record of vitamin K). To take account of potential confounding, we adjusted odds ratios for delivery type and admission to a special care nursery, and these are listed alongside their unadjusted counterparts.

The adjusted odds ratio among those whose medical notes indicated they had received intramuscular vitamin K was 1.0 (95% confidence interval 0.5 to 1.9, based on 45 informative sets) for acute lymphoblastic leukaemia and 1.2 (0.7 to 2.3, 52 informative sets) for all leukaemias (table 1). Imputation resulted in lower risk estimates, but as the calculation was based on fewer informative case-control sets the confidence intervals are wider.

Table 1

Vitamin K status of cases and controls assessed according to what was recorded in hospital notes and what was imputed from hospital policy

View this table:


The retrospective assessment of whether a baby received vitamin K, and by what route, is not straightforward. Information about vitamin K may be found in several places: the mother's obstetric notes, the delivery register, neonatal notes, and nursing cardex. Additionally, hospital policy on vitamin K varies from routine (oral or intramuscular) to a more selective regimen of intramuscular vitamin K for high risk neonates. When vitamin K is given, details are generally recorded at the time of administration, although route is not always stated. Because vitamin K administration cannot always be confirmed from the records available some researchers have imputed information from hospital policy.1 2 3 4 Imputation can be problematic: historical records are rarely available, and current staff are sometimes ill informed about past policies; and on an individual basis it is an act of faith to assume that where no written record is found hospital policy prevailed. For this reason we presented our data in two ways. With either method, our findings do not support the suggestion that the risk of childhood leukaemia is increased in neonates who receive intramuscular vitamin K.

We thank Judith Black, Susie Boon, and Pat Townshend for data collection; the hospital staff who helped trace the records; and Gerald Draper, Richard Doll, and Charles Stiller for comments on a previous draft.


  • Funding Imperial Cancer Research Fund.

  • Conflict of interest None.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
View Abstract