Manuscript review #2, received along with rejection letter from Dr. Barnett Kramer, Feb. 11, 2003

Whilst this is a useful monograph and guide to the art of so-called cell culture drug resistance testing, in this manuscript, many questions are raised and almost none are satisfactorily answered in a way that delivers improved understanding in this field.

I would like to raise a few points here regarding the assertions made:

1. That there is substantial evidence that these assays truly provide art index of platinum drug sensitivity: The authors make this assertion repeatedly. However, although these assays are clearly a surrogate for this, it is unclear whether they relate to proliferative capacity/rate as opposed to drug resistance, The well differentiated tumors, with potentially longer cycling time, may be in fact not determinable in short term cultures due to this fact. This does not mean that patients might not benefit from platinum if they have well differentiated tumors, only that these assays do not usefully predict this benefit, and worse, that they do not closely track platinum resistance as surrogate endpoints, but represent the composite measurement of something else.

Furthermore the assertion that these are "the most validated assays for...the circumvention of platinum resistance" bears more scrutiny. Validation in this context implies clinical utility. In reality these assays have excellent quality control and are technically reproducible, however, it is quite unclear how valid they are since a randomized study using a highly evaluable assay has not been performed.

2. That these assays predict for patient survival in ovarian cancer. `Predict" implies that the assay represents an independent prognostic variable, otherwise it does not add much more than currently available factors, for instance what would be added to the knowledge that the ovarian cancer is poorly differentiated of having an assay performed? It is therefore not unreasonable to have expected a multivariate analysis to have been performed as a minimum in this study, and, if not, to have demonstrated some recognition of the fact that repeated t-testing will raise some "significant" p-values by chance. There is clear evidence that very extensive t-testing occurred in this study, and one should be cautious about p-values here.

3. The lack of multivariate analysis leads to the next point: The authors claim that this is a prospective study simply because the assay was performed prospectively and the result was therefore `locked' from the outset. This is to entirely miss the point about what a prospective study is. In a prospectivc study, they would have collected all the relevant information in a structured way so as to bc able to perform a multivariate analysis to support the assertion of the title of this manuscript. No comment is made about factors such as patient age, stage, surgical debulking, and histological type with regards to multivariate analysis, other than that the authors had no access to much of this information. Since this was not done, all manner of hidden bias will contaminate the conclusion, and indeed, for this reason, no real comment can be made. At best, this manuscript can only be regarded as an audit of practice, an even then, a poorly designed prospective audit

4. Where to go from here?

30 or more years of research in this field have not answered the question. If these assays work they should improve prognosis. Proof of this is the only thing worth publishing now. This needs to be a randomised study. Since claims are made that these assays allow one to devise new treatments, a regime should be identified from this vast retrospective dataset that can be suggested to be active when single agent platinum is apparently not. The untreated patients can be randomised to one of 3 arms. Standard treatment (platinum) in arm A, a single, standard utilitarian multidrug regimen in arm B targeted to the platinum resistant group, and designed to be as broad spectrum as possible for platinum ~resistant" patients according to the assay, and finally the utilisation of the assay in arm C to treat patients according to the assay result. If the assay works and is utilitarian, the median survival of the patients in arm C will be shown to improve, and the field can move on. However, unfortunately in the meantime, this manuscript adds no clarity to the field.