Trial Math: Meeting the Primary Endpoint, Pt. 2

We firmly believe that Phase 3 testing should not be started unless you can adequately predict the outcome. It's critical to understand what the drug is doing, which patients are most likely to benefit, what other options those patients have, and which endpoints would be most convincing for government agencies to approved the labelled indication for the drug. -- Provectus' Eric Wachter, June 19th conference call

In my blog post Trial Math: Meeting the Primary Endpoint, Pt. 1 I noted DTIC's PFS is likely to be around 1.5 to 2 months for the stage of patient being recruited for the trial's comparator arm. The performance of systemic chemotherapies DTIC and TMZ are well documented, and generally yield a normal distribution of events (e.g., Middleton et al.). I also noted that in Provectus' metastatic melanoma Phase 2 trial the median PFS for Stage III patients (similar to those who would be enrolled in the Phase 3 trial) was at least 9.7 months (median PFS for Stage III subjects was not reached during the 12-month study interval), and the mean PFS of the all disease treated subgroup of the Phase 2 trial was 9.8 months. So, 9.7-9.8 months for PV-10 PFS > 1.5-2 months for DTIC/TMZ PFS.

That is all well and good, but for Provectus' upcoming Phase 3 trial to meet its primary endpoint of progress free survival ("PFS"), the confidence interval ("CI") for response in the PV-10 (test) arm cannot overlap the CI for response of the DTIC/TMZ (comparator) arm. The trial's power is 90%, and statistics are two-sided with an alpha of 0.05, which means there is a 5% chance the true response is above or below a 90% CI.

Information about DTIC/TMZ PFS figures abound for applicable patient populations, together with 95% CIs. These intervals can be converted or adjusted to present 90% CIs (applicable to the Phase 3 trial design).

Provectus provided durable objective response  ("DOR") data for the all disease treated subgroup from its Phase 2 trial on its ASCO 2014 poster.

Click to enlarge

DOR, or durable response rate, is complete response ("CR") plus partial response ("PR") that is durable, and is measured in months. PFS is essentially CR + PR + SD (stable disease), in that the disease does not progress or become "PD" (progressive disease). I say SD because shrinkage is neither sufficient to qualify for PR nor a sufficient increase to qualify for PD. As such, PFS ultimately will be a greater figure than DOR. I utilized the CI for PV-10 DOR as a guide to estimate the CI for PV-10 PFS. Comparison of 95% and 90% CIs for DTIC and PV-10 are below.

Click to enlarge.

Neither PV-10's actual Phase 2 DOR nor estimated Phase 3 PFS CI overlap DTIC CIs. While the bottom DOR intervals are within spitting distance of the top DTIC intervals, there is substantial daylight between the PFS endpoint bottom intervals and the top DTIC ones. This would suggest the likelihood of the Phase 3 trial meeting its primary endpoint of PFS.

Trial Math: Meeting the Primary Endpoint, Pt. 1

We firmly believe that Phase 3 testing should not be started unless you can adequately predict the outcome. It's critical to understand what the drug is doing, which patients are most likely to benefit, what other options those patients have, and which endpoints would be most convincing for government agencies to approved the labelled indication for the drug. -- Provectus' Eric Wachter, June 19th conference call

There are important aspects of Provectus' upcoming pivotal Phase 3 trial for unresected locally advanced cutaneous melanoma to consider and assess, including but not limited the design itself, total trial cost and cost per patient, likelihood of success at completion, and likelihood of success at or by the interim assessment. The trial design is discussed on Provectus' ASCO 2014 poster (bottom right hand corner). Could the trial be successful, and if so why (or why not)? Could it be stopped early, and if so when and why (or why not)? How much could the trial cost?

The sample size (N) of the pivotal trial is estimated at 210 patients and randomized two-to-one (2:1). There would be 140 patients in the PV-10 arm and 70 patients in the comparator/systemic chemotherapy (DTIC [dacarbazine]/TMZ [temozolomide]) arm. The trial's primary endpoint is progression free survival ("PFS"). The study assumes the null or base hypothesis of the two arms having the same response -- that is, PFS for both PV-10 and the systemic chemotherapy (DTIC/TMZ, or DTIC for short) would be the same. The trial is powered to detect the alternate hypothesis or outcome of the PV-10 and DTIC arms having different responses -- that is PFS for PV-10 would be different from PFS for DTIC.

The alternative hypothesis or difference in response -- this difference in PFS between patients in the PV-10 arm versus those in the control or comparator (DTIC) arm -- may be expressed as a hazard ratio ("HR"), which essentially is the ratio of the response (PFS) of the control arm divided by the response (PFS) of the PV-10 arm. The smaller the HR, the larger the effect size (or impact of investigational drug), and thus the more clinically relevant the observed difference between the two arms is.

Click to enlarge.

In the example above, for this pivotal trial, a target DTIC PFS of 1.5 months and a predicted HR of 0.545 would lead to a predicted PV-10 PFS of 2.8 months -- that is, assuming a target control arm PFS of 1.5 months and using a set HR of 0.545, the goal of the trial would be for patients in the PV-10 to demonstrate a PFS of at least 2.8 months.

DTIC and TMZ are well known systemic chemotherapies (the former is administered intravenously, the latter is a pill). Their performance is well documented. For example, refer to Middleton et al., where median PFS was 1.9 months for TMZ and 1.5 months for DTIC. There are various other randomized control trials and studies utilizing DTIC/TMZ as control arms (in similar settings to Provectus' pivotal trial), and PFS may be as high as 2 months (so x above might be 3.7 months). Thus far the pivotal Phase 3 trial's HR is not known. I'm using the HR from 2012/2013 when Eric was discussing a Phase 3 trial design under a sought after special protocol assessment with the FDA (see below). The target HR for the current trial may be higher or lower, but I doubt it is materially different.

Click to enlarge. Provectus corporate presentation, March 15, 2013

In Provectus' metastatic melanoma Phase 2 trial, median PFS for Stage III patients (similar to those who would be enrolled in the pivotal Phase 3 trial) was at least 9.7 months (median PFS for Stage III subjects was not reached during the 12-month study interval) (see below).

Click to enlarge. Immuno-chemoablation of metastatic melanoma with intralesional rose bengal, October 2012

Another way of looking at potential PV-10 PFS in the pivotal Phase 3 trial is considering the mean PFS of the all disease treated subgroup of the Phase 2 trial (see below).

Click to enlarge. Locoregional Disease Control in Metastatic Melanoma: Exploratory Analyses From Phase 2 Testing of Intralesional Rose Bengal, September 2013

9.7-9.8 months for PV-10 PFS > 2-3.7 months for PV-10 PFS "predicted" > 1.5-2 months for DTIC/TMZ PFS.

Historical PV-10 PFSs may be indicative of potential pivotal trial success because the figures are substantially greater than the possible or likely DTIC PFS; however, the above analysis is very rough and not the lease bit "loose." Additionally, trial success is achieved when the PV-10 arm's confidence interval ("CI") for response -- the CI for PV-10's PFS -- does not overlap or cross DTIC's CI for response -- the CI for DTIC's PFS. Nevertheless, the large difference between PV-10's possible pivotal trial's PFS and DTIC's may present a large enough cushion between the two figures, and the potential for trial success when the time comes.

Perhaps this is what Eric might have meant, in part, when he said "[w]e firmly believe that Phase 3 testing should not be started unless you can adequately predict the outcome" on the June 19th conference call.