Tuesday, April 12, 2016

The results you want to see.


In recent years, there's been a great deal of controversy surrounding antidepressant medications. In 2007, the FDA issued a black box warning advising against the prescription of selective serotonin repute inhibitors (SSRIs) for treating patients under the age of 25. While this age group encompasses multiple developmental periods, adolescence is particularly important in the context of depression and other psychiatric disorders. Why? As figure 4 from Paus et al., 2008 shows (above) about half of all mood disorders (including depression) emerge during adolescence. While there are still multiple treatment options are available to adolescent patients with depression, as mentioned earlier, one of the main treatment approaches is known to cause depression symptoms to worsen in this population.

While there has been debate about whether SSRIs actually do more harm than good in treating adolescents with depression, one recent meta-analysis not only confirmed increased suicide ideation following SSRI treatment in children and adolescents, but also noted cases of increased aggressive behavior and advocated for close monitoring of all patients being treated with SSRI's - regardless of age. While these findings from the meta-analysis are new, these drugs have been prescribed for multiple decades.

The authors of the meta-analysis points out how the drug company Eli Lilly either misrepresented, or omitted, incidences of suicide ideation and suicide attempts in their summaries of patients who were assigned to receive SSRIs during drug trials. This reminded me of the section of our textbook that dedicated to discussing outliers. Specifically, Motulsky notes, "Even if you try to be fair and objective, your decision about which outliers to remove will probably be influenced by the results you want to see" (p.210). Furthermore, the author warns against occluding outliers if the outlier simply reflects biological variability. While we may not yet know the precise biological mechanism by which SSRIs contribute to worsened depression symptoms and suicide in adolescents, we do know that the adolescent brain undergoes significant changes during this developmental period. For this reason, when you compare the adolescent brain with the adult brain, you are really comparing organs with two completely different biologies.

One of the focuses of the lab I work in is geared towards finding antidepressant alternatives to SSRI's, and testing the short-term and long-term therapeutic potential of administering these drugs during adolescence. I believe that reports such as Sharma et al., 2016 shed light on skewed outcomes from prior drug trials that could be partially explained by excluding certain data points. The prior exclusion of data collected from patients treated with SSRIs during drug trials may play a role in terms why, only in recent years, risks associated with SSRIs have been revealed in adolescent populations and are still viewed as debatable.

4 comments:

  1. This is a very interesting blog post, as it emphasizes Dr. Murphy's point about not eliminating outliers, given the subjectivity of this exercise. Even though there are tests, like the Grubb's test, for identifying outliers, the researcher has to take into consideration certain assumptions, as well as the significance level. As you pointed out, biological variability should also be considered. This is why I agree with you that data in previous drug trials has been skewed by excluding what the researchers thought were outliers. This blog post makes me further understand the importance of not eliminating any data point, given it can have a high impact and affect future experiments being done on the subject by other scientists.

    ReplyDelete
  2. This is a very interesting blog post, as it emphasizes Dr. Murphy's point about not eliminating outliers, given the subjectivity of this exercise. Even though there are tests, like the Grubb's test, for identifying outliers, the researcher has to take into consideration certain assumptions, as well as the significance level. As you pointed out, biological variability should also be considered. This is why I agree with you that data in previous drug trials has been skewed by excluding what the researchers thought were outliers. This blog post makes me further understand the importance of not eliminating any data point, given it can have a high impact and affect future experiments being done on the subject by other scientists.

    ReplyDelete
  3. I think your argument about eliminating outliers is very important. What may be even more important in your example is the idea of bias. The development of drug therapies is both the most critical step in treating patients, but also the most susceptible to bias. As you stated, when quoting Motulsky, even if you are trying to be as fair and as objective as possible, there are times when people have to make decisions about outliers. We all feel the stress of publishing and getting the data we may want to see, but in drug development the stakes are much higher. This is true for people all on all sides, those developing the drugs, those who will profit, and those who are competing for their own drug therapies. Therefore it is important, as you point out, to be extremely cautious of all bias from all sides. Interestingly, there are now companies, or consultants for lack of a better term, that will independently analyze your data in order to make unbiased decisions about outliers and significance. It is considered unbiased because they profit from doing the analysis, regardless of the outcome. Even this system is not perfect, and could have issues of their own. But it does highlight that people are starting to recognize the impact that bias can have on such a high stakes experiment.

    ReplyDelete
  4. It always amazes me when drug companies do not properly do their statistical analyses or when try to sweep glaring errors under the rug. I understand that drug development is a costly process, and they want to get products out on the market as soon as possible, but you think they would (in an ideal world) also want to avoid any detrimental effects or inevitable future lawsuits. It should be a standard practice to present the whole data ("outliers" included) to the FDA, but again that would be in an ideal world. I do think it's a tricky road, though, to work in suicide ideation; would it be a "yes" or "no" to suicidal thoughts or a scale? How may patients would state the truth? That being said, though, I am glad that there are studies out there that point out the flaws in some of these drugs development statistics, and can hopefully improve the pharmacology climate so that other companies do not exclude critical data, especially given the significant impacts on human life.

    ReplyDelete