Objective New strategies are required to rapidly identify novel cytostatic agents before embarking on large randomized trials. This study investigates whether a change in rate of rise (slope) of serum CA125 from before to after starting a novel agent could be used to identify cytostatic agents. Tamoxifen was used to validate this hypothesis.

Methods Asymptomatic patients with relapsed ovarian cancer who had responded to chemotherapy were enrolled and had CA125 measurements taken every 4 weeks, then more frequently when rising. Once levels reached 4 times the upper limit of normal or nadir, they started continuous tamoxifen 20 mg daily, as well as fortnightly CA125 measurements until symptomatic progression. Because of the potentially nonlinear relationship of CA125 over time, it was felt that to enable normal approximations to be utilized a natural logarithmic standard transformation [ln(CA125)] was the most suitable to improve linearity above the common logarithmic transformation to base 10.

Results From 235 recruited patients, 81 started tamoxifen and had at least 4 CA125 measurements taken before and 4 CA125 measurements taken after starting tamoxifen, respectively. The mean regression slopes from using at least 4 1n(CA125) measurements immediately before and after starting tamoxifen were 0·0149 and 0·0093 [ln(CA125)/d], respectively. This difference is statistically significant, P = 0·001. Therefore, in a future trial with a novel agent, at least as effective as tamoxifen, using this effect size, the number of evaluable patients needed, at significance level of 5% and power of 80%, is 56.

Conclusions Further validation of this methodology is required, but there is potential to use comparison of mean regression slopes of ln(CA125) as an interim analysis measure of efficacy for novel cytostatic agents in relapsed ovarian cancer.