Friday, June 27, 2008

On Not Getting It...



Here is the scenario: you have cancer, and your doctor has told you there is no way to treat it. But there is an experimental drug that is being offered in a phase 1 study. Your doctor asks if you're interested.

You join the study because its your best shot. You know there are likely to be side effects–there always are with cancer drugs. You know the drug might not work–no cancer drug works all the time for all people.

If your doctor knew that entering the phase 1 study was unlikely to eradicate your cancer-- or even extend your life– would you want to be told, or would you prefer that the doctor keep this information from you?

The question came up in a recent research ethics committee (IRB) meeting I attended, in which a phase 1 cancer study consent form contained language to the effect of "you may or may not benefit from joining this study." I proposed that the language be changed to "this study is unlikely to significantly affect the course of your cancer." My preference was based on the following reasons:

1- "may or may not benefit" does not contain any information about probability. Indeed, there is a subtle rhetorical partity, wherebt most readers view "may" as having equal weight as "may not" such that they view this as suggesting a 50% probability of benefit.

2- two recent, large mete-analyses showed that studies of the type reviewed here result in tumor shrinkage for about 10% of patients. There is no evidence that tumor shrinkage in these studies actually translates to longer life or improved comfort. So, in fact, we do have information about probability of benefit.

3- Patients enter these studies not because of a desire to further cancer research, but rather, because they are seeking cure. This has been shown in numerous surveys.

4- Patients who enter these studies significantly overestimate the probability that they will benefit clinically. Also shown in numerous surveys.

5- Informed consent requires, at a minimum, that volunteers be provided information that a reasonable person would consider material to their decision to enroll.

A reasonable, terminal cancer patient would probably want to have some information about the probability that a drug that is likely to have strong side effects is unlikely to work against their cancer.

Incredibily, most of the members of my IRB disagreed, saying (among other things) that it would discourage patients (hmmm. I thought medical paternalism went out with Buicks with tailfins); that nothing in the original language implied benefit was a sure thing (yes, the original language made it very clear that instead of 100% chance of benefit, chance of benefit was somewhere between 0 and 99.999%);  that no one would enroll in the study with my language (wow! I thought our job was to protect volunteers, not investigators); that we've never done that before (let me make sure I understand this.  It's ok to withhold relevant information from volunteers, as long as you've done it before).

Yes, Virginia, the IRB system really is broken! (photo credit: Kenoir 2007)

No comments: