Assessing Provectus' Pivotal Melanoma Phase 3 Trial, Part VI

The sixth in a series of blog posts and news items assessing the company's pivotal Phase 3 trial for unresectable locally advanced cutaneous melanoma (Stage III patients, and not metastatic [Stage IV] melanoma).

1. Win or lose: Will the trial succeed in meeting or fail to meet its primary and/or secondary endpoints?
2. Time-to-success or -failure: How long might it take for the trial to succeed or fail?
3. Good or bad process: Are there process steps and aspects thereof that may provide hints of potential future trial success or failure.

This blog news item, (i) 1. Win or lose: Predicting outcomes, (ii) 2. Time-to-success or -failure: An unusual response in one of the randomized controlled trial arms, and (iii) 3. Good or bad process: Multiple triggers.

Blog post takeaways

• Provectus' pivotal melanoma Phase 3 trial should be successful, meeting its primary endpoint of progression-free survival (PFS).
• Meeting the secondary endpoint of overall survival (OS) in the context of PFS as a surrogate for OS, assuming the beneficial support of patient-reported outcomes (PROs), is worth a separate discussion and blog post or news item,

• It appears the single planned interim analysis for efficacy and safety (based on a prescribed number of events) is very unlikely to (can't) be triggered. There [probably] won't be enough events. As a result, trying to project the timing of trial success, on an interim data readout basis, ultimately must rely on speculating what triggers, boundaries, safeguards, etc. have been designed into the trial to compensate for PV-10's non-normal or unexpected distribution of events to assure a timely readout of interim study results.

• If Provectus management believe they can adequately predict the outcome of the pivotal trial, then such adequacy may extend beyond whether the trial would be successful to when they think it could be so. Fundamental to these beliefs is the trial's actual enrollment rate.

An article like Seeking Alpha contributor Steven Giardino's June 2014 one (h/t InvestorVillage PVCT poster leave_the_gun) discussing the role, responsibilities and mandate(s) of a pivotal trial's data monitoring committee (DMC) could encourage us to take a step back when assessing Provectus' Phase 3 trial, before rejoining more analysis and opinion-making as well as prior deep dives.

1. Win or lose: Predicting outcomes 

The trial's outcome of winning or losing, succeeding or failing, may simply boil down to observing that PV-10 just kills cancer, which is how I believe the company's Chairman and CEO Dr. Craig Dees, PhD would see and say it. The preclinical and clinical dataset of PV-10 ablating a cancerous tumor or lesion continues to grow, while also demonstrating the drug's agnosticism to disease presentation: e.g., clinically in potentially different melanomas (mutations or variations, but lesions don't appear to be categorized by Provectus), hepatocellular carcinoma (Hepatitis B and C, cirrhosis), breast cancer, and liver metastases (colorectal, non-small cell lung, melanoma, ovarian), and preclinically in at least melanoma, liver, renal, breast and pancreatic tumors. If PV-10 just kills cancer, then the Phase 3 trial ultimately should win or succeed. If it doesn't, it won't.

Provectus' CTO Dr. Eric Wachter, PhD on the other hand already said "[w]e firmly believe that Phase 3 testing should not be started unless you can adequately predict the outcome" when considering what PV-10 does or can do, what chemo does, the trial's patient population, the primary endpoint, secondary endpoints, etc., all in the context of the company's pivotal trial. If Eric thinks he can predict the trial's outcome, then he clearly also believes the trial will win or succeed.

"It just kills.." or "being able to predict the outcome..." could present itself, in my over-analytical or analyzed manner, as:

• In the melanoma Phase 2 trial the overall response rate (ORR) for patients who had all their existing melanoma lesions injected with PV-10 was 71%, with 50% achieving a complete response (CR) despite receiving PV-10 up to four times over a 16-week period (i.e., week nos. 0, 8, 12 and 16) [1],[2],
• In the Phase 3 trial patients in the PV-10 treatment arm would receive the drug every 4 weeks until CR or disease progression (e.g., week nos 0, 4, 8, 12, 16, 20, etc.).

• In the Phase 2 trial the ORR for the lesions of patients who had all their lesions injected with PV-10 was 80%, with 74% achieving a CR [1], and

• In the Phase 2 trial the progression-free survival (PFS) of patients who had all their lesions injected was 9.8 months [1].

• In the pivotal Phase 3 trial the control arm is systemic chemotherapy of either dacarbazine (DTIC) or temozolomide (TMZ), which have PFSs of less than 2.5 months for DTIC/TMZ [2].

• I believe Eric designed the trial for a 90% power to detect a 70% improvement in median PV-10 PFS (over DTIC/TMZ PFS) — if the trial's DTIC/TMZ median PFS is 1.5 months then PV-10's must be >2.55 months, or if 2 then >3.4, or if 2.5 then >4.25.

[1] Subgroup Efficacy in Patients Receiving Intralesional Rose Bengal to All Existing Melanoma in Phase II Study PV-10-MM-02, European Society For Medical Oncology, Abstract #1120P, September 2014. [2] Janet Fricker, PV-10 delivers greatest effects when all lesions are injected, Pharmiweb.com, October 15, 2014

• As the pivotal trial's principal (lead) investigator, St. Luke's Cancer Center's Dr. Sanjiv Argawala, MD, said last year following ESMO 2014, “The progression free survival of 9.8 months compares favourably with historical progression-free survivals of less than 2.5 months for DTIC/TMZ.” [2]

2. Time-to-success or -failure: An unusual response in one of the randomized controlled trial arms

Compare and contrast PV-10 and chemotherapy above, or the distribution of responses ("events") for chemotherapy and PV-10 below.

Click to enlarge.
See July 23rd blog post Assessing Provectus' Pivotal Melanoma Phase 3 Trial, Part IV. 

I think it's fair to describe PV-10's historical objective tumor response results for melanoma as non-normal distribution curve or very likely to produce an unexpected distribution of events, where the baseline of what is or should be expected being chemotherapy's distribution curve. Thus, trying to estimate the timing of early and/or interim readouts of Provectus' pivotal study results must rely heavily on what if any as yet unmentioned multiple triggers, boundaries or safeguards Eric designed into the trial to augment the publicly discussed planned interim analysis for efficacy and safety based on a prescribed number of events (which is probably never going to [won't] get triggered).

In my July 23, 2015 Assessing Provectus' Pivotal Melanoma Phase 3 Trial, Part IV blog post I tried to make the argument Eric designed a planned interim analysis of efficacy and safety based on prescribed time, in addition to the planned one based on prescribed events. The analysis suggested the timing of this second trigger (starting in July 2016) was generally in line with (but later than) public comments by management of when an interim analysis might occur ("likely in the first half of 2016"). If there can be two interim analysis trigger, why cannot there be more for a drug with an effect size large enough to withstand the statistical dings or alpha spend of additional efficacy looks?

3. Good or bad process: Multiple triggers

Planned/prespecified or unplanned/not prespecified reviews of safety data of randomized controlled trials do not cause bias or incur alpha spending, and thus do not result in statistical dings of study results. Reviews of or "looks" into efficacy data, however, do incur dings or spend alpha (if you take multiple looks at the data, there is an increase in the risk of Type I error). Thus, if the addition of a second trigger (a prescribed time one) indeed is true (the first being a prescribed event trigger), then Provectus' statistical analysis plan (SAP) would have to address and thus account for three efficacy looks: (i) a final analysis, (ii) the single planned prescribed event interim analysis and [potentially] (iii) a prescribed time interim analysis. As an aside, the aim of an SAP is "to minimize bias by clearly stating the proposed methods of dealing with protocol deviators, early withdrawals, missing data, and the way(s) in which anticipated analysis problems will be handled as well as many other possible issues."

Using current parameters (such as the trial's N of 225), Eric's design yields a 90% power to detect a 70% improvement in median PFS. As I noted above, if the trial's actual DTIC/TMZ median PFS turns out to be 1.5 months then for the trial to win PV-10's actual median must be >2.55 months, or if 2 then >3.4, or if 2.5 then >4.25. PV-10's median Phase 2 trial PFS of 9.8 months (out of a 12-month follow-up period) compared to a 70% improvement of 4.25 over Agarwala's 2.5 month historical median PFS for DTIC/TMZ should mean PV-10 has a lot more alpha to spend if Eric wants to spend it. And if he spends his alpha early (on a much earlier interim analysis than expected) and never has to spend it again (on the planned interim analysis), then he can continue through the final analysis with the planned spend equalling the actual spend.

What might these additional triggers be and when (if at all) could they facilitate earlier study readouts than the interim analysis (prescribed event or time)? My guess tries to keep simple: There may be efficacy looks that of course cost alpha, presumably planned or prespecified but not publicly discussed, that occur after a certain number of events (one event per person). The planned interim analysis requires 113 events for a trial N of 225 (50%). Maybe another trigger is 20 events, 30 events, etc. The analysis of what "early" eventually could be defined as may boil down to the number of patients (and thus events) required to show statistically significant PV-10 and chemotherapy PFS curve separation acceptable to a DMC and/or the FDA. Returning to 2. Time-to-success or -failure above, the timing of success then becomes the sum of (i) the enrollment time of the patients (e.g., 60, 90, etc.) required to generate the smaller number of events (e.g., 20, 30, etc.) and (ii) the separation time required of the last patient treated in that smaller trial n.

Click to enlarge. 20 events.

Click to enlarge. 30 events.

Eric has been repeatedly and consistently wrong in the past with regard to the timing of things. Peter's guidance of a 1H16 interim data readout, based presumably on Eric's guidance, thus is questionable because it requires the trial's actual enrollment rate to track Eric's projected monthly enrollment rate figure. A interim analysis trigger requiring a lower number of events, and thus patients, might fit into Peter/Eric's sense of timing given lower actual enrollment rates. Or it may not.

Previous entries in the series:

1. 3. Good or bad process: Patient enrollment. See Assessing Provectus' Pivotal Melanoma Phase 3 Trial, Part I (July 13, 2015) on the blog's Current News page,
2. 3. Good or bad process: Designing an interim analysis for efficacy into Provectus' pivotal melanoma Phase 3 trial. See July 16, 2015 blog post Assessing Provectus' Pivotal Melanoma Phase 3 Trial, Part II,
3. 3. Good or bad process: Patient crossover. See Assessing Provectus' Pivotal Melanoma Phase 3 Trial, Part III (July 20, 2015) on the blog's Current News page,
4. 2. Time-to-success or -failure: Triggering the interim analysis. See July 23, 2015 blog post Assessing Provectus' Pivotal Melanoma Phase 3 Trial, Part IV, and
5. 3. Good or bad process: Patient-reported outcomes. See Assessing Provectus' Pivotal Melanoma Phase 3 Trial, Part V (July 24, 2015) on the blog's Current News page.