Sunday, May 22, 2016

Placebo Effect and The Choice of Placebo

Recently, NPR had a discussion about the influence of Placebo effect on consumer product purchases.  While the concept of the placebo effect comes from clinical trials in medical contexts, the placebo effect also have impact in non-medical research fields. For example, just by telling the people it is a brand name product (even though it is not), it affects the results of using the product.

Placebo effect is also called the placebo response. It is a remarkable phenomenon in which a placebo -- a fake treatment, an inactive substance like sugar, distilled water, or saline solution -- can sometimes improve a patient's condition simply because the person has the expectation that it will be helpful. Expectation to plays a potent role in the placebo effect. The more a person believes they are going to benefit from a treatment, the more likely it is that they will experience a benefit.”

Placebos have to be identical in all respects to the active drug, except that the active drug is absent. This means for oral placebos that they are not distinguishable by color, size, shape, taste, or texture. For oral therapy there are several modes of application available: capsules, tablets, or liquids. Since many active drugs have a distinctive taste, the use of capsules is most often feasible.

Sometimes, it is not so easy to make the placebo that are not distinguishable from the active drug. For example, liquids containing therapeutic protein generate foam but the placebo using buffer solutions do not. The treatment can be differentiated by observing the foam by shaking the test products. In this case, the ideal placebo will contain also an inactive protein in low concentrations such as human albumin. Some clinicians are very clever in breaking the code; they not only shake or identify the drugs by taste and smell but even photometers were used to determine differences between active drug and placebo. For such reasons it is sometimes very difficult to guarantee blinding.

While majority of the placebos used in the clinical trials are true inactive substance (at least we think so), there are times that the placebos may not be true inactive substance. This is especially true in the clinical trials using the protein therapies, for example, the therapeutic proteins derived from human plasma.

Table below listed several randomized, placebo controlled clinical trials using the therapeutic proteins derived from human plasma. Various concentrations of albumin (also protein) were used as the placebo assuming that the low dose of albumin has no therapeutic effect.

Study
Experimental Treatment
Choice of Placebo
IGIV
Biological: 0.25% human albumin solution (Placebo)
IGIV
Biological: Placebo solution: Human Albumin 0.25% - 4 mL/kg
Biological: Placebo solution: Human Albumin 0.25% - 2 mL/kg
 IgPro20
2% human albumin administered by weekly SC infusions during the SC treatment period of the study
Prolastin
Albumin (Human) 20%, United States Pharmacopeia (USP)
IGIV
Placebo will be supplied as 20% or 25% Albumin and diluted with glucose 5% to a final concentration of 0.1%


If the results of these clinical trials are positive (statistically significant), people will not question the use of albumin as the placebo. If the results of these clinical trials are negative (not statistically significant) or the placebo group perform much better than expected, people may question the adequacy of the use of albumin as the placebo. This happened in one of the clinical trials in relapse-remitting multiple sclerosis (RRMS).  

Prior to the clinical trial below, previous studies had shown that the benefit of IGIV in treating the RRMS. The expected response rate in placebo group is about 50%.

Randomized, Double-Blind, Placebo-Controlled Study to Compare the Effects of Different Dose Regimens of IGIV Chromatography (IGIV-C), 10% Treatment on Relapses in Patients With Relapsing Remitting Multiple Sclerosis
  
The study results were published in Fazekas et al (2009) “Intravenous immunoglobulin in relapsing-remitting multiple sclerosis: a dose-finding trial”. For the primary efficacy endpoint, after 1 year, the proportion of relapse-free patients did not differ statistically according to treatment (IVIG 0.2 g/kg: 57%; IVIG 0.4 g/kg: 60%; placebo: 68%). Surprisingly, the response rate in placebo group is much higher than expected and than what was assumed when the study was designed. This prompted the discussion of the potential therapeutic effect of albumin used as the placebo. The article by Hommes et alIVIG trials in MS. Is albumin a placebo?” argued that albumin may have properties that make it unsuitable as placebo.

Since using the albumin as the placebo may have therapeutic effects, something other than albumin (for example, saline with Polysorbate) may be an option as the placebo in clinical trials with these therapeutic proteins. Polysorbete is added to mimic the foam created by the therapeutic protein.  However, the blinding may still remain as a concern when saline is used as the placebo because of the difference in foaming. In clinical trials using the therapeutic proteins, we need to balance the risks of potential unblinding of using Saline as placebo versus potential placebo effect of using albumin as placebo. 

For blinded studies, we always assume that the blinding is maintained during the course of the study. However, this can be wrong in many trials where the investigators and/or patients can tell which treatment group the patients are on by the distinguishable features between the actual product and the placebo, by the different adverse event profiles, and others. It is important to assess the potential unblinding by asking the investigators or patients to guess the treatment assignments – a practice seems to be reasonable, but pretty much nobody wants to do. If the assessment indicates the potential unblinding or high percentages of guess-it-right, there is no easy way to mitigate the impact and make it difficult to interpret the results from the clinical trial.


After the study, it is a good practice to compare the results from placebo group back to the original assumptions. During the design stage of the clinical trial, we usually make assumptions about the response rate or mean/standard deviation in placebo group. After the study, it will be interesting to compare the actual results from the placebo group with the assumptions in the protocol sample size calculation section. I see the situations where the final results are way different from the assumptions in the protocol sample size calculation – in this case, we must have some further investigation what goes wrong and why this happens. 

Monday, May 02, 2016

Dose Cohort Expansion Study - A Never Ending Phase I Study Design

Clinical trials are conducted in a series of steps, called phases - each phase is designed to answer a separate research question. In Phase I studies, Researchers test a new drug or treatment in a small group of people for the first time to evaluate its safety, determine a safe dosage range, and identify side effects. Phase I trials are usually conducted in healthy volunteers, however, phase I trials are often conducted in patients for 1) Trials in life-threatening diseases (such as cancer, AIDS); 2) clinical trials with human plasma derived products; 3) clinical trials with some biological products; 3) clinical trials with gene-therapy

Many phase I studies are designed as a dose escalation study. In previous articles, I have discussed the Phase I Dose Escalation Study Design: "3 + 3 Design" and Alternative phase I dose escalation study designs: CRM, BLRM, mTPI, and PGDE. The purpose of these dose escalation studies is to find the maximum tolerable dose (MTD). However, the efficacy signal can also be detected/assessed.

In a recent issue in New England Journal of Medicine, Prowell et al published an article titled “Seamless Oncology-Drug Development” and discussed the suddenly popular study design so called ‘dose cohort expansion study’. The study on pembrolizumab or MK-3475 or Keytruda (a programmed cell death 1 receptor) by Merck was mentioned as the example.  


The study was initially designed as a typical dose escalation study with a title “Phase I Study of Single Agent MK-3475 in Patients With Solid Tumors. The anticipated number of subjects was 32. “When impressive response rates and durations of response were observed early in the trial, particularly in patients with metastatic melanoma or non–small-cell lung cancer, the sample size was rapidly increased. Cohorts were added to assess efficacy in these two patient populations and to evaluate alternative dosing regimens and candidate predictive biomarkers — a move that resulted in the enrollment of more than 1200 patients in the trial”

This phase I study is still active, but not recruiting participants at this point. The latest description on clinicaltrials.gov is so much different from the original study design. The official title of the study has been revised to “Phase I Study of Single Agent MK-3475 in Patients With Progressive Locally Advanced or Metastatic Carcinoma, Melanoma, and Non-Small Cell Lung Carcinoma”. The change history documented on clinicaltrials.gov indicated 60 updates. One of the FDA review documents stated that the protocol was amended 50 times and counting.

While the protocol is still staged as a phase I study, many of the sub-studies have been added to this phase I study protocol. The study design is way beyond the initial dose escalation. The purpose of the study has been shifted to demonstrate the efficacy with an appropriate dose level. Within some added cohorts, the control group was added and subjects were randomly assigned. Essentially this is a combined phase I/II studies with multiple sub-studies. If it is designed in traditional phased clinical trial approach, this one study will have to be split into multiple phase I and phase II studies.

With portion of the results from this phase I study, Merck has obtained the approval from FDA initially for Advanced Melanoma, and last October for Non-small cell lung carcinoma. FDA review documents for its approval for advanced Melanoma have been posted on FDA’s website and the review documents for its approval for non-small cell lung carcinoma should be posted soon.

Cohort expansion design has now been used by many other study sponsors. We can easily see these studies with cohort expansion design from the clinicaltrials.gov website. In cohort expansion study, the decision on which cohort to add and how many subjects to expand depends on the results from the previous cohorts. However, it is very different from the adaptive design where the adaptation rules are pre-specified. In cohort expansion study, the changes to the study design are all post-hoc.

Cohort expansion study can be replaced with multiple phase I and phase II studies. However, keeping the study under the same protocol (but with many protocol amendments) will probably still save time and budget in study conduct, patient recruitment, IRB/EC submissions, and regulatory submissions. In a very competitive environment for oncology drug development, the time has the ultimate importance. 

Cohort expansion design becomes a viable approach in drug development perhaps due to following special circumstances:

Further reading:

Sunday, May 01, 2016

Free SAS e-learning courses in Programming 1 and Statistics 1

Great news to new SAS learners! SAS have made their introductory e-learning courses available to everyone at no cost. Don’t miss this chance to learn – or brush up on – programming and statistics basics.
  • Programming 1: This course will teach you the basics of writing SAS programs. You'll learn how to navigate SAS, create and combine SAS data sets, create reports and more.
  • Statistics 1: This introductory course is for SAS users who perform statistical analyses using SAS/STAT® software. You'll learn about ANOVA and linear regression and get a brief introduction to logistic regression.
These e-learning courses are the same material SAS teach on-site and via Live Web. 

To access the free version of the e-learning courses, use the following web link. You will need to enter your SAS profile account user name and password. if you don't have a SAS profile, you can create one for free. 

In addition, if you don't have a SAS software, you can get a SAS University Version for free. 
Als try SAS new video portal loaded with tons of free SAS tutorials
SAS user guides for all SAS products are always free on the web. 
Youtube also has many videos for SAS tutorials