Jump to comment:

• Evidence seriously flawed due to survival-related biases and confounding
  Igor Karp
  Published on: 13 April 2016

• Published on: 13 April 2016

  Evidence seriously flawed due to survival-related biases and confounding

  • Igor Karp, Assistant Professor
  • Other Contributors:
    • Alexandra Manganas

  We read with great interest a recent article by D'hooghe et al. (1) on long-term effects of childbirth on MS. Unfortunately, the study appears to have been subject to important biases (in addition to those discussed by the authors themselves, including, notably, confounding).

  The first bias - namely, the left-truncation bias (2) - has to do with the fact that to be included in the analyses, patients had to be "...

  Show More

  We read with great interest a recent article by D'hooghe et al. (1) on long-term effects of childbirth on MS. Unfortunately, the study appears to have been subject to important biases (in addition to those discussed by the authors themselves, including, notably, confounding).

  The first bias - namely, the left-truncation bias (2) - has to do with the fact that to be included in the analyses, patients had to be "under regular care" at the referral centre in 2005-2007. Thus, patients who had the onset of MS in 1975, for example, would only be included in the study sample if they survived at least until 2005. Because MS survival is correlated with the rate of disease progression, those who were included represent a subpopulation with a slower rate of progression than among all MS patients (if they were followed from the onset of MS). As a result of the left truncation, the estimated Kaplan-Meier "survival" curves are upwardly biased.

  The second bias - namely, the immortal-time bias (3,4) - has to do with the fact that the person-time contributed by patients who had no children and those who had children after the onset of MS was handled incorrectly. Specifically, to be included in the "children only after onset of MS" group, a woman with no children before the onset of MS had to reach the time of the post-MS-onset childbirth without reaching EDSS of 6. During that time period, there were no events of reaching EDSS of 6 by definition - otherwise, the woman would have been included in the "no children" group. Thus, instead of being considered to have been contributed by women who had children after the onset of MS, this "immortal" time should have been considered to be contributed by women who had had no children up to that point. As a result of the immortal-time bias, the median time to EDSS of 6 was underestimated in the reference group (no children) but overestimated in the group with children only after the onset of MS. Through a similar mechanism, the median time to EDSS of 6 was underestimated in the group with children only before onset of MS but overestimated in the group with children before and after onset of MS.

  Overall, the evidence produced in the study by D'hooghe et al. (1) is, thus, seriously flawed and does not allow for any meaningful inference about the effect of childbirth on MS progression.

  References:

  1. D'hooghe MB, Nagels J, Uitdehaag BM. Long-term effects of childbirth in MS. J Neurol Neurosurg Psychiatry 2010; 81:38-41.

  2. Gail MH, Benichou J. Encyclopedia of epidemiologic methods. West Sussex, UK. Wiley 2000.

  3. Levesque LE, Hanley JA, Kezouh A, Suissa S. Problem of immortal time bias in cohort studies: example using statins for preventing progression of diabetes. BMJ 2010; 340:907-911.

  4. Renoux C, Suissa S. Immortal time bias in the study of effectiveness of interferon-bias in multiple sclerosis. Ann Neurol 2008 64:109-10.

  Conflict of Interest:

  None declared

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top