Neonatal vitamin K prophylaxis: the Gordian knot still awaits untyingBMJ 1998; 316 doi: https://doi.org/10.1136/bmj.316.7126.161 (Published 17 January 1998) Cite this as: BMJ 1998;316:161
A small risk of leukaemia still has not been excluded, but safe and effective prophylaxis seems possible
- Rüdiger von Kries, Professor of paediatricsa
- aInstitute for Social Paediatrics and Adolescent Medicine, Ludwig-Maximilians-Universität, 81377 Munich, Germany
Since 1990 the issue has remained unresolved of whether prophylactic vitamin K, given to newborn infants to prevent bleeding, causes childhood cancer. Four more studies appear in this week's issue. Do they finally conclude the vitamin K and cancer story?
When I went into paediatrics some 20 years ago vitamin K prophylaxis was not an issue. In Germany at least almost all newborn infants who required medical care and many who had had instrumental deliveries were given 1 mg of vitamin K intramuscularly, whereas the practice in healthy infants varied. Altogether only about half of all live births received intramuscular vitamin K, but vitamin K deficiency bleeding did not seem to be major problem until an apparently new condition-late vitamin K deficiency bleeding-emerged in Europe in the early 1980s.1
About half the cases with late vitamin K deficiency bleeding presented with intracranial haemorrhage. The condition was confined mainly to breast fed infants, had a peak incidence at weeks 2-6, and was often caused by underlying cholestatic disease, diagnosed only after the bleeding occurred. In the absence of prophylaxis the incidence of late vitamin K deficiency bleeding in Europe was estimated to be 40-100 per million live births.2 Although the annual number of cases was low, the condition became a major issue in paediatrics, partly because of the severity of the bleeding-which led to death or severe handicap in about a third of cases-but mainly because the bleeding was thought to be preventable by neonatal vitamin K prophylaxis.1 This prophylaxis had originally been introduced to prevent classic haemorrhagic disease of the newborn in the first week of life.
The causes of vitamin K deficiency in the first week of life (mainly low intake of human milk with low vitamin K content) and of late vitamin K deficiency bleeding (mainly malabsorption and occasionally increased requirement for vitamin K) are different,3 and a single oral dose of vitamin K, although sufficient to prevent classic haemorrhagic disease, is not enough to prevent late vitamin K deficiency bleeding.2 3
Somewhat surprisingly a single intramuscular dose of 1 mg vitamin K administered at birth was found to protect against late vitamin K deficiency bleeding, whereas the same or even higher single doses given orally proved much less effective.2 Intramuscular vitamin K for all newborn infants was therefore recommended in 1991.3 4
After the identification of this effective prophylaxis, the problem seemed solved until 1990, when the first report of a potential association between vitamin K prophylaxis and childhood cancer appeared.5 This unexpected finding could not be ignored, as plasma concentrations of vitamin K after intramuscular administration of 1 mg exceed endogenous levels by a factor of up to 10 000.6 We needed rapidly to clarify whether such high vitamin K uptakes might be harmful.
Golding's subsequent case-control study in Avon unfortunately confirmed the presumed risk for all cancers, particularly leukaemia.7 A series of ecological8 and analytical epidemiological9 10 11 studies followed which failed to confirm the increased risk, and in 1996 in a BMJ editorial Zipursky suggested it was time to abandon worries about a potential cancer risk from intramuscular vitamin K.12 At that time the four studies on vitamin K and cancer published in this issue of the BMJ were still in progress. Do they now resolve the issue?
Two of the studies are reassuring. The strengths of the population based case-control study by McKinney et al are the representativeness for the population studied and the standardised and refined protocol for determining exposure to vitamin K (p 173).13 They found that odds ratios for central nervous system and other solid tumours were 1.0 or below and those for leukaemia, including acute lymphoblastic leukaemia in children aged 1-6, never exceeded 1.3 with 95% confidence intervals including 1. Similarly reassuring are the results of the ecological study by Passmore et al (p 184).14 From these data a cancer or leukaemia risk of the size suggested by Golding's data seems unlikely, but a smaller, 20% to 30%, increase in risk cannot be excluded with this type of study.
The results from Parker et al are, however, worrying (p 189).15 Although a risk of solid tumours was almost certainly excluded, a “significant” twofold risk for acute lymphoblastic leukaemia among 1-6 year olds was reported. Several possible biases exist in the design of this study-for example, only about half the eligible cases were included; those who ascertained exposure were not blind to the status of cases and controls; and no adjustment for multiple testing was made in the analyses. We made a similar observation on the risk of acute lymphoblastic leukaemia in the same age group,11 but we observed the association only when the cases were compared with local controls (predominantly from the same hospital) and not with controls randomly selected from the whole area. We considered this a chance finding related to multiple testing.
Even more worrying are the results of the case-control study by Passmore et al (p 178).16 The most remarkable finding in this study is a borderline significant risk for all cancers (odds ratio 1.44; 95% confidence interval 1.00 to 2.08), mainly due to an increased risk of acute lymphoblastic leukaemias. This risk could partly be accounted for by abnormal delivery. What was noteworthy, however, was that the risk of leukaemia was associated with abnormal delivery in children born in hospitals with policies to give intramuscular vitamin K selectively (based on specific clinical criteria) but not in those born in hospitals with non-selective policies-that is, vitamin K was given to all newborn infants or to none.
It is disappointing that these four new studies do not resolve the vitamin K-cancer question. A risk of solid tumours can almost definitely be ruled out, but a small risk of leukaemias cannot be excluded. Epidemiology has been blamed for producing all kinds of spurious results, but how can epidemiology prove that these results are wrong? Retrospective ascertainment of exposure to vitamin K at birth often requires imputations that leave room for bias. Subtle differences in study design and analysis seem to account for some of the differences between these studies. Does the choice of controls (from the same hospital or not) matter? To what extent do differences in the assessment of vitamin K exposure matter? Does grouping of the analysis according to whether the children had been born in hospitals with selective or non-selective prophylactic policies matter? How much adjustment for confounding makes sense and is necessary?
Analyses of pooled records from the European case-control studies are underway and might help to answer these questions and allow the calculation of risk estimates with 95% confidence intervals narrow enough to rule out even a small risk for leukaemias. While epidemiologists are analysing the pooled data, parents and doctors have to decide what to do. Almost all cases of late vitamin K deficiency bleeding are preventable with intramuscular vitamin K prophylaxis, with a potential risk for some forms of leukaemia that seems more hypothetical than real. Repeated oral doses may be almost as effective if given to breastfed infants weekly (1 mg)17 or daily (25 µg).18 The efficacy of three oral 2 mg doses with the old polyethoxylated castor oil preparation (Konakion) or the new mixed micellar preparation (Konakion MM) remains to be established.