• Piantadosi describes Phase 2 as assessing (i) safety and activity and (ii) feasibility.

What question(s) is trial seeking to answer?

1. "Should comparative trial be conducted later?"

2. "Which subpopulation should experimental therapy be used in?"

3. "Does experimental therapy work at all?"

4. "Does experimental therapy work better than standard of care?"

5. "There are multiple experimental therapies currently in pipeline – which should be carried forward?"

Go to https://pollev.com/philipboonst814 or text PHILIPBOONST814 to 22333. Type the numbers as a single string (e.g. 125, nothing else) of the questions that you think would support doing a randomized phase 2 study.

• How many eligible patients are there?

• How many eligible patients can be enrolled?

• How many patients can we afford to enroll?

• Would patients accept randomization to non-experimental therapy in middle development?

• What is disease prognosis for patient?

• What is potential for large experimental treatment effect?

• Pipeline operating characteristic: recall (HW4) odds of true positive are \(\tfrac{\pi}{1-\pi} \prod_{i=1}^3 \tfrac{\theta_i}{\alpha_i}\)

  • Implies that even error-proned randomized phase 2 increases the pipeline odds as long as \(\theta_i > \alpha_i\), i.e. power exceeds type I error rate. Could even do single arm phase 2, then error-proned randomized phase 2, then comparative randomized phase 3!

  • This perspective assumes that nominal values of \(\theta_i\), \(\alpha_i\) are correct

• Bias-variance trade-off

• One-arm (non-randomized) phase 2 trials make implicit assumption that historical response rate \(\textcolor{orangeish}{p_0}\) is correctly described, i.e. known, when comparing to \(\gamma\). This buys us sample size (reduces variance).

• However, \(\textcolor{orangeish}{p_0}\) may not be correctly described

• How could this be violated?

1. Difference in study popuations (selection bias)
2. Changing therapeutic landscape (time bias)
3. Difference in outcomes (measurement error)

• I know a little about clinical trials but am not a physician
• My questions are those that I would pose as a protocol or manuscript reviewer. Investigators do not have opportunity to respond

• Hepatitis B is infectious disease spread through blood or bodily fluids. Symptoms related to abnormal liver function

• Experimental agent is yeast cell intended to stimulate additional immune response to HBV

• Phase I study in healthy volunteers demonstrated safety and potential mechanism

• Ultimately seeking 'functional cure' for HBV

• Study enrolled chronic Hep B patients whose disease was already responding to oral antiviral (OAV) therapy

• Design:

• Endpoint:

• Sample Size:

#Arm A (we haven't learned these formulas yet)
#(1 + 1/r) *(z_(1-alpha/2)+z_(1-beta))^2 / (delta / sigma)^2;
(1.5) * (qnorm(0.016/2) + qnorm(0.20))^2 / (0.15 / 0.179)^2

## [1] 22.56966

#Arms B-D
#r * (1 + 1/r) *(z_(1-alpha/2)+z_(1-beta))^2 / (delta / sigma)^2;
2 * (1.5) * (qnorm(0.016/2) + qnorm(0.20))^2 / (0.15 / 0.179)^2

## [1] 45.13932

• Results:

• Discussion:

• Looking to establish that experimental therapy works better than standard of care using non-clinical endpoint
• Also seeking to establish dose-efficacy relationship
• Minimal risk to patients

• In retrospect, anticipated signal was optimistic (which is point of conducting trial)

• Psoriasis is chronic skin disorder. Painful, comorbidities, psychological and quality of life problems

• Foam formulation of a Vitamin D + steroid combination

• !!!:

• Background:

• Design:

• Endpoint:

• Sample size?? (427 patients enrolled; no justfication)

• Results:

• Discussion focuses on statistically significant finding of foam treatment being superior ointment treatment

• Why were controls necessary?

  • Cal/BD already established as potent treatment (I think)
  • No stated hypothesis regarding interactions between treatment and vehicle
  • 100 patients were randomized to control because…why?
• Concern about lack of sample size justification
• Is this underpowered phase 3?

• I don't understand difference between this trial and referenced PSO-FAST trial

• Untreated metastatic esophageal cancer or gastroesophageal junction cancer

• Background:

• Design:

• Endpoint:

• Sample size (??)

• Results:

• Discussion:

• Randomized selection design: which chemotherapy regimen to carry forward?

  • Postscript: all three chemo regimens are currently treatment options, but Cetuximab has failed subsequent validation

• Differences in response rates, but overall survival differences went away after about 15 months

• Insufficient information to reproduce sample size (information probably in protocol)

• No dispute that randomized designs controlled against standard of care are uniformly superior. Rarely (or never) see uncontrolled phase 3 design

• Randomized phase 2 designs should not be used underpowered, error-proned phase 3 design

• Chapter 6.5.2, (Piantadosi, 2017)

• Chapter 5, (Friedman, Furberg and DeMets, 2010)

• Chapter 3.2, (Cook and DeMets, 2007)

• (Rubinstein et al., 2009), (Mandrekar and Sargent, 2010), (Sargent and Taylor, 2009)