| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Am J Epidemiol 2004; 159:717-718.
Copyright © 2004 by the Johns
Hopkins Bloomberg School of Public Health
LETTERS TO THE EDITOR |
RE: "RISK ANALYSIS OF ASEPTIC MENINGITIS AFTER MEASLES-MUMPS-RUBELLA VACCINATION IN KOREAN CHILDREN BY USING A CASE-CROSSOVER DESIGN"
Department of Statistics, The Open University, Milton Keynes MK7 6AA, United Kingdom
Ki et al. (1) used a case-crossover design to evaluate the relative risk of aseptic meningitis after measles-mumps-rubella vaccination. Although their results are broadly in line with those of others on the same topic, the case-crossover design cannot generally be recommended for investigating associations between adverse events and childhood immunizations because it requires the probability of exposure (i.e., vaccination) to be constant over time (2). This requirement is certainly not met by childhood immunizations administered according to highly age-dependent schedules. Measles-mumps-rubella vaccines, for example, are typically given at 12–15 months of age.
Ki et al. (1) correctly state that, in the case-crossover method, each person serves as his or her own control, but they then assert that confounding due to vaccination age is thereby eliminated. This is not the case: If the underlying probability of vaccination varies with age, then exposure and control periods are not comparable, introducing what is in effect a control selection bias.
For example, suppose for simplicity that the hazard period includes the month preceding the event and that the control period includes the 6 months prior to the hazard period. Consider an aseptic meningitis case with onset at the end of month 14. The hazard period then includes month 14. The probability of measles-mumps-rubella vaccination in this period is high. The control period covers the period from age 7 months to age 13 months. The average monthly probability of measles-mumps-rubella vaccination during this period is likely to be lower. In contrast, an aseptic meningitis case occurring at the end of month 18 has a hazard period including month 18 and a control period including months 12–17. In this case, the chances are that the case will be vaccinated during the control period. These age effects can produce biases, which are apparent in table 2 of the paper by Ki et al. (1): The relative risks are highest for aseptic meningitis cases that occur at less than 18 months of age and much lower for cases that occur at later ages. Although it is quite possible that the biases roughly cancel out across cases in this particular study, this cannot be relied upon in general. For example, an adverse event unrelated to measles-mumps-rubella vaccination but occurring primarily before age 18 months would appear to be positively associated with measles-mumps-rubella vaccination. On the other hand, if the event occurred primarily after age 18 months, measles-mumps-rubella vaccination would appear protective.
The requirement of constant underlying exposure probability is fundamental to the case-crossover method. In fact, it has been shown that exposures in successive time periods must satisfy the even stronger requirement of exchangeability to avoid bias in logistic regression analyses of case-crossover designs (3). This mirrors the tacit assumption in matched case-control studies that the controls are exchangeable. The exchangeability requirement, a demanding one for case-crossover designs, stems from the method’s origins in the case-control paradigm, and it cannot be sidestepped by controlling for age at exposure. In contrast, the self-controlled case-series method (4) referred to by Ki et al. (1) does not suffer from this shortcoming. This design, superficially similar to the case-crossover design insofar as both are self-matched, derives from cohort (fixed exposure, random event) rather than case-control (fixed event, random exposure) logic and hence has quite different properties. Full adjustment for age and time effects is possible in self-controlled case-series designs, which have been applied to the study of adverse events after vaccination (5) and drug safety more generally (6).
REFERENCES
REFERENCES
- Ki M, Park T, Yi SG, et al. Risk analysis of aseptic meningitis after measles-mumps-rubella vaccination in Korean children by using a case-crossover design. Am J Epidemiol 2003;157:158–65.
[Abstract/Free Full Text] - Greenland S. Confounding and exposure trends in case-crossover and case-time-control designs. Epidemiology 1996;7:231–9.[ISI][Medline]
- Vines SK, Farrington CP. Within-subject exposure dependency in case-crossover studies. Stat Med 2001;20:3039–49.[CrossRef][ISI][Medline]
- Farrington CP. Relative incidence estimation from case series for vaccine safety evaluation. Biometrics 1995;51:228–35.[CrossRef][ISI][Medline]
- Murphy TV, Gargiullo PM, Massoudi MS, et al. Intussusception among infants given an oral rotavirus vaccine. N Engl J Med 2001;344:564–72.
[Abstract/Free Full Text] - Hubbard R, Farrington P, Smith C, et al. Exposure to tricyclic and selective serotonin reuptake inhibitor antidepressants and the risk of hip fracture. Am J Epidemiol 2003;158:77–84.
[Abstract/Free Full Text]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||