Post-publication Peer Reviews to:
|
|
|
|
|||
|
Facundo Garcia Bournissen, Pediatrics Resident Hospital de Niņos "R. Gutierrez", Buenos Aires, Argentina
Send letter to journal:
facugb{at}hotmail.com Facundo Garcia Bournissen
|
Dear Sir, I find the paper by van Driel et al. misleading in the abstract and some parts of the results when they state that "an IQ score below 80 was found in 11 children in the exposed compared to 3 children in the non- exposed cohort". The fact that this difference is non-statistically significant is not stated anywhere in the abstract, and the low power of this kind of sub-group analysis is not discussed in their paper. Following their methods, one could state that coumarin exposed boys have a higher probability of an IQ over 140 than non-exposed boys [p=0.091 chi- square test] (information extracted from figure 1, page 126). van Driel et al. provide very useful information on children that were exposed in-utero to coumarin, but it is not that they have a higher risk for an IQ under 80. Rather, van Driel et al. show that these children DO NOT have an increased risk for a lower inteligence, as no difference was found between mean IQ of the exposed vs. controls. Yours, Facundo Garcia Bournissen. Hospital de Niņos "R. Gutierrez". Buenos Aires, Argentina |
|||
|
|
|||
|
dieneke van driel, pediatrics resident Beatrix Children's Hospital, Groningen, The Netherlands
Send letter to journal:
dienekevandriel{at}hotmail.com dieneke van driel
|
Dear sir, We disagree with the criticism of the author of this letter, since we put the odds ratio of 'intelligence quotient below 80' with it's confidence interval in the text and abstract of our paper. We found 11 children with an IQ below 80 in the exposed cohort of 291 children compared to 3 in the cohort of 253 nonexposed children. The odds ratio was 3.2, with a confidence interval of 0.8 to 11.6. We did not explicitely state that this is not significant, as we assumed all readers would realize this because the confidence interval includes one. We feel a three -fold increased ratio is remarkable and should be kept in mind when weighing the risks of mother and fetus in anticoagulant therapy during pregnancy. The odds ratio was based on the complete study cohort of 291 exposed and 253 nonexposed children, not on a subgroup analysis. Although large numbers give more accurate effect measures, we felt no need to discuss a low power of our analysis of 544 children, the largest study on this subject so far. The author suggests an evaluation of an IQ score above average. When analyzing the occurence of a high IQ score, we prefer to use comparable methods to our analysis of low IQ (below two standard deviations of the mean IQ score of our control group) above a random cutoff point (e.g. IQ score above 140). According to our methods, an IQ score above two standard deviations of the mean IQ score of our nonexposed control group, i.e. above 130, can be defined as 'above average'. In the exposed cohort, we found 6 children compared to 11 in the nonexposed cohort; odds ratio is 0.5 (confidence interval 0.2 to 1.3). NB figure 1 on page 126 displays the distribution of IQ score of the exposed and nonexposed cohort and not the IQ score of exposed and nonexposed boys. Sincerely, P.J.J. Sauer, D. van Driel. |
|||