retinopathy of prematurity.org :
a documentation of patient-harming frauds in medical research
The study-authors' own words
show their gross ethics violations
highlightings and comments added for emphasis
Go straight to the highlights on this page:
Letter dated January 17, 1995, from James D. Reynolds, M.D., at the Children's Hospital of Buffalo, Pediatric Ophthalmology,
to Don Everett, M.A.
We were pleased with the thorough way the study section reviewed our LIGHT-ROP grant. They identified the strengths and weaknesses of our proposal, praising our design, manual of procedures, and personnel.
We also appreciate their concerns which fall into three areas. The first issue arises from the recent light reduction study reported by Seiberth. (1) Although this study did not find any effect of light reduction on the incidence of ROP, there are several design features of the Seiberth study that are not clear in the manuscript. The section on sample size refers to a sample size of 62 with alpha = 0.10 and power of .80 "for interim analysis". A sample size of 62 in each group is actually the fixed sample size required to detect the reduction in rates they specify (from 40% to 20%). This does not allow for interim analysis. The remainder of the sample size discussion deals with another fixed sample size calculation of 473 in each group; hence, it is not clear whether the authors intended to do a study of size 62 or 473 in each group or if interim testing was appropriately dealt with to allow consideration of early termination of the study. If the procedure were followed as they describe then the study would be stopped prematurely. The study was discontinued with what may have been about 13% of the data that was originallv intended for the study because the early interim analysis "showed a tendency for increased ROP incidence in the treatment group". Seiberth may be incorrectly describing their sample size considerations or there may be an inappropriate use of statistical methodology.
In the section of Fundus Examination the schedule of examinations is not clearly specified by a protocol. It appears that the infants may have had masked examinations at 5 and 8 weeks and at 3 and 6 months, but the wording in this section is not clear. It leaves the impression that there were other examinations that were not masked, and only these examinations were masked. If so, were the results of non-masked examinations used in reporting the results on the incidence of ROP? Did interim examinations still allow one to accomplish the maskings? And were the two groups examined for ROP with equal vigor? The description does not allow one to answer these and related questions. The incidence of ROP and difference in ROP between treatment groups could be influenced by these protocol issues.
Of the original cohort of infants randomized who survived, there were 20% lost to follow-up. This is a substantial number for a small study for which there are questions regarding .important aspects of the design of the study.
By contrast our proposal will follow a specified protocol of examinations and we will enroll almost 50% more patients. This will increase the power of the study and also provide a 30% reduction in the confidence interval for the treatment effect. Our study would enroll many more at risk babies. Approximately one- third of Seiberth*s patients were over 1250 grams at birth. Babies this large have a very low event rate and are at very low risk of adverse effects of ROP. Hence, the incidence of ROP and the incidence of more severe ROP will be much higher in our cohort. With a larger sample size of higher risk infants, our study will be much more powerful and provide more precise estimates than Seiberth's study.
Additionally, the figures presented for the ambient light in the Seiberth study indicate that both treatment groups were exposed to lower light levels than in our study. We will produce a bigger difference in light exposure for infants who are occluded compared to those who are not occluded because of the relatively bright nurseries in our study together with a longer period of patching.
The study section review expressed concern for the "lack of a compelling scientific rationale". This is indeed a catch-22. However, the indirect evidence alluded to in our proposal and the possibility of finding a relatively simple means for preventing the development of ROP justify our feasibility trial. If successful, light reduction by preventing the development of ROP could have a major impact on the need for intervention in severe ROP.
Finally, there is the dilemma of the full study vs. our proposed feasibility. It is hard to justify a major expenditure without a firmer scientific rationale. We believe our feasibility trial is a good compromise. We avoid an expensive large trial while retaining the excellent design of the large trial. We will have enough patients to conceivably answer important questions raised by Glass' study definitively and the feasibility trial may supersede the need for a large trail. However, if further research is indicated the mechanism will be in place to respond rapidly to the needs of a larger study.
In summary, we have responded to past critiques and the current review. The design we propose is an excellent compromise of expense vs. statistical power. Our design is far superior to Seiberth, et al and the scientific rationale remains strong enough to support this feasibility study. The question of whether or not light reduction can prevent the development of ROP is important both scientifically and clinically. It should be answered with a properly designed clinical trail.
James D. Reynolds, M.D.
cc: Robert Hardy, Ph.D. Rand Spencer, M.D. Alistair Fielder, FRCS
(1) Seiberth, V. et al: Am.J.Ophthalmol. 118: 492-495, 1994.
Contact us at retinopathyofprematurity.org