|
|
| |||||||||||||||||||||||||||||||
Background Two recent studies have found that infants who received intramuscular vitamin K were at twice the expected risk for cancer during childhood. Since nearly all newborns in the United States receive this drug, the public health implications of this association, if confirmed, would be substantial.
Methods We examined the relation between vitamin K and cancer in a nested case-control study that used data from the Collaborative Perinatal Project, a multicenter, prospective study of pregnancy, delivery, and childhood. Among 54,795 children born from 1959 through 1966, 48 cases of cancer were diagnosed after the first day of life and before the eighth birthday. Each case child was matched with five randomly selected controls whose last study visit occurred at or after the age when the case child's cancer was diagnosed. Exposure to vitamin K was determined from study forms and medical records.
Results Vitamin K had been administered to 68 percent of the 44 case children and 71 percent of the 226 controls for whom data were available (matched odds ratio, 0.84; 95 percent confidence interval, 0.41 to 1.71). The odds ratio was 0.47 (95 percent confidence interval, 0.14 to 1.55) for leukemia and 1.08 (95 percent confidence interval, 0.45 to 2.61) for other cancers. Sequential adjustment for potential confounding factors did not change the results substantially.
Conclusions We found no association between exposure to vitamin K and an increased risk of any childhood cancer or of all childhood cancers combined, although a slightly increased risk could not be ruled out. The benefits of neonatal vitamin K prophylaxis against hemorrhagic disease have been well described. Unless other evidence supporting an association between vitamin K and cancer appears, there is no reason to abandon the routine administration of vitamin K to newborns.
Methods
The subjects of this study were members of the Collaborative Perinatal Project (CPP) cohort. The CPP was a prospective, multicenter study designed to evaluate risk factors for the occurrence of various neurodevelopmental disorders of childhood. The study methods have previously been described in detail4,5,6. Briefly, pregnant women were enrolled from 1959 through 1966 at 12 sites in the United States. Their children were followed through the age of seven years (approximately 80 percent of the cohort)5 or, at some centers, eight years (approximately 36 percent of the cohort)6 and examined according to a standard protocol. Each child's parent or guardian was interviewed at regular intervals during the first two years of the child's life and annually thereafter. Whenever study personnel learned (usually through an interview with the parent) that a child had died or had been hospitalized, the relevant medical records and, when appropriate, autopsy records were obtained.
In this nested case-control study, potential cases of cancer were ascertained in several ways. First, study identification numbers of children with cancer were obtained from the authors of a previous investigation of cancer and x-ray exposure in this cohort7. Second, the records of all children without life-threatening anomalies who weighed at least 1500 g at birth and who died after the first week of life were reviewed. Third, diagnostic summary forms were completed for all children at one and seven years of age. If a neoplasm was reported, the record was reviewed to obtain a specific diagnosis.
The study records of each child with a potential diagnosis of cancer were reviewed by two investigators who were board-certified pediatricians and who were blinded to the child's vitamin K exposure status. Definite cases were required to have a histologically proved diagnosis of cancer noted on a medical-record summary, a clinical course (including treatment) consistent with the diagnosis, or both. Stillborn infants with cancer (n = 3) were excluded from the analysis, as were neonates whose cancer was diagnosed or strongly suspected during the first day of life (n = 7, including 3 with neuroblastoma in the late second trimester), because vitamin K could not have been a factor in these cases. None of the remaining case children were recognized to have a syndrome associated with an increased risk of childhood cancer. For each case, five controls were selected from among all the children without cancer whose last study visit took place at or after the age (in months) at which the disease was diagnosed. The controls were selected with a computerized random-number generator.
Vitamin K was given in the delivery room or the nursery. The CPP employed labor-and-delivery-room observers (who were not involved in the clinical care of the mother or child), who were instructed to record data related to all events during and after delivery, including the name, dose, and route of administration of all drugs given to the mother and newborn. Exposure in the delivery room was determined on the basis of a single question about drugs administered to the child. The CPP nursery-history form contained a specific question about the receipt of vitamin K, including the dose, route of administration, and brand. In addition, labor-room records were reviewed for all neonates who did not have a notation in the delivery-room or nursery report indicating that they had received vitamin K. In 14 subjects, prophylaxis was accomplished by administering vitamin K to the mother during labor (5 mg intramuscularly).
Exposure to vitamin K was determined independently from the delivery-room and nursery forms by two investigators, who were unaware of the children's case-control status. For 43 children, exposure could not be ascertained from the study records. For 25 of these, personnel at the hospitals where the children were born were contacted, and the original medical records were reviewed (without knowledge of case or control status) to obtain information on exposure to vitamin K. For 18 children, it remained undetermined whether they had been exposed to vitamin K.
A matched analysis was performed. Unadjusted odds ratios were derived from a conditional logistic-regression model8 that included only a term for exposure to vitamin K; 95 percent confidence intervals were derived from the standard error of the regression coefficient. Adjusted odds ratios were derived from conditional logistic-regression models that included factors putatively associated with childhood cancer (race, sex, birth weight, maternal age, exposure to x-rays during pregnancy, and breast-feeding)9,10. Because of the relatively small number of cases of cancer, adjustment was made for only one confounding factor at a time.
Results
The cohort contained 54,795 live-born children, 48 of whom developed cancer after the first day of life. After loss to follow-up was accounted for by life-table methods,11 this corresponded to a cumulative incidence of 1.1 cases of all cancers and 0.4 cases of leukemia per 1000 children followed to 90 months of age. Leukemia was the most common cancer (n = 16), followed by brain tumors (8), neuroblastoma and Wilms' tumor (6 each), lymphomas (4 non-Hodgkin's, 1 Hodgkin's), retinoblastoma (3), rhabdomyosarcoma (2), fibrosarcoma (1), and hepatoblastoma (1). Each case child was matched to 5 controls, resulting in a study population of 48 cases and 240 controls. Characteristics of the cases and controls are presented in Table 1. In this cohort, use of vitamin K increased over time (56 percent of newborns received vitamin K in 1959 through 1960, 66 percent in 1961 through 1962, 77 percent in 1963 through 1964, and 86 percent in 1965 through 1966).
|
|
Several brands of vitamin K used in the CPP are no longer produced or no longer recommended for neonatal prophylaxis. In the United States, the only brands of vitamin K currently approved for this use are Aquamephyton and Konakion. Infants who received these brands had an odds ratio for cancer of 0.57 (95 percent confidence interval, 0.24 to 1.36), which provides further evidence of the safety of the drugs used currently. The odds ratio for cancer among children who received brands of vitamin K containing phenol was 0.69 (95 percent confidence interval, 0.28 to 1.74).
In 4 children with cancer and 14 controls, data on exposure to vitamin K could not be ascertained. Because at a particular time and in a particular hospital, either all children or no children generally received vitamin K, we could deduce whether vitamin K was probably administered by checking the records of other children born at the same time in the same hospital. When this was done, one child with cancer and eight controls were assumed to have received vitamin K; the odds ratio for cancer among exposed as compared with unexposed children in this analysis was 0.82 (95 percent confidence interval, 0.43 to 1.56). Under the most extreme assumption (that all case children and no controls for whom information or exposure was unavailable received vitamin K), the odds ratio would be 1.19 (95 percent confidence interval, 0.60 to 2.35), and cancer would still not be significantly more common among children who received vitamin K.
Discussion
We found no increased risk of either cancer in general or leukemia in particular among children who received vitamin K during the perinatal period; this finding is in accordance with the observation that the incidence of childhood leukemia in the United States did not increase from 1948 (when few or no newborns received vitamin K) to 1980 (when nearly every newborn received the drug)13. The upper confidence limits of our findings exclude the odds ratios reported for all cancers (2.6 and 1.97) and for leukemia (2.65) among children exposed to vitamin K in the two reports by Golding et al.,1,2 although our confidence intervals and theirs overlap. In one of the reports,2 orally administered vitamin K was found not to result in an increased risk of cancer. In the CPP sample, only one child received vitamin K orally; therefore, we cannot address this issue. However, our results provide evidence of the safety of vitamin K administered intramuscularly.
Our study has several strengths. The data were collected prospectively, and the sample was not likely to have been biased in any way with respect to cancer; the records of vitamin K use could not possibly have been influenced by the outcome. Because the records were part of a research study, they are more complete than routinely collected medical data; individual data on exposure were obtained for 94 percent of the subjects. On the basis of national data on incidence according to age, race, and sex,14 the probability of having cancer by the age of 7 1/2 years in a cohort with the racial makeup of the CPP sample is 1.1 per 1000; for leukemia, the probability is 0.35 per 1000. These rates are nearly identical to the observed incidence rates of 1.1 and 0.4 per 1000, respectively, suggesting that few cases, if any, were missed. In addition, controls were matched to case children for length of follow-up, making differences in the probability of detecting cancer unlikely.
Controls were not matched to case children according to study center or date of birth, because at any given center and time, generally either all children or no children received vitamin K. Not matching for these factors might theoretically cause a problem, however. The presence of a carcinogen in one hospital or at one time could artificially strengthen the association between vitamin K and cancer if vitamin K was being given or artificially weaken the association if vitamin K was not being given. Fortunately, there is little evidence of temporal, local, or hospital-based cancer "epidemics" in the 12 CPP centers. In the absence of evidence that the study site itself was a risk factor for childhood cancer, matching for site would have constituted overmatching, which could have resulted in a serious reduction in the statistical power of our study to detect an association between vitamin K and cancer.
The two brands of vitamin K currently approved for neonatal prophylaxis (Aquamephyton and Konakion) were not found to be associated with an increased risk of cancer; the upper 95 percent confidence limit of 1.36 is statistically compatible with only a small increase in risk. The possibility that phenol, the preservative in Konakion, is carcinogenic has been raised15; our results do not support this hypothesis. The vehicles for Konakion differ slightly between the United States and United Kingdom preparations16. None of the brands used in the CPP contained the vehicle used in the United Kingdom, and we therefore have no data on the safety of that particular preparation.
In conclusion, we found no evidence that perinatal exposure to vitamin K is associated with an increased risk of leukemia or of cancer in general during childhood. The number of cases was small, however, and slight increases in the overall risk of cancer, particularly of specific types of cancer, cannot be ruled out on the basis of our data. Because of the well-documented benefits of vitamin K prophylaxis, we support the recommendation of the American Academy of Pediatrics that newborns receive intramuscular vitamin K17 unless other, more compelling data are collected that demonstrate an increased risk of childhood cancer.
We are indebted to Drs. Linda van Marter (Harvard University), Joan Regan (Columbia University), James Levin (University of Minnesota), Cynthia Morris (University of Oregon), and Stuart Weiner (Pennsylvania Hospital) for reviewing the medical records of the children whose vitamin K exposure could not be determined from the CPP study forms; to Dr. C.S. Chung (University of Hawaii) for providing us with the study identification numbers of the children with cancer; and to Ms. Patricia Moyer for technical assistance.
Source Information
From the Division of Epidemiology, Statistics, and Prevention Research, National Institute of Child Health and Human Development, Bethesda, Md. (M.A.K., J.S.R., J.L.M.), and the Center for the Future of Children, David and Lucile Packard Foundation, Los Altos, Calif. (P.H.S.).
Address reprint requests to Dr. Klebanoff at the Division of Epidemiology, Statistics, and Prevention Research, National Institute of Child Health and Human Development, National Institutes of Health, 6100 Bldg., Rm. 7B03, Bethesda, MD 20892.
References
| |||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
