Maybe the docs figure their elderly patients will be too fatigued to complain?

I kid, I kid.

The system is crazy. Once we see this insanity what else is there to say, explanation wise? Not much.

I do agree that we may be looking at some lazy research here as far as the reporting goes. Could be someone stuck on the beat that doesn't want to be there and doesn't give much of a hoot.

Who knows.

The only solution is to keep shining the light on the shadows. Hopefully as awareness grows, change will come.

According to the abstract, the effect size was 0.93 (CI 0.50-1.36), not 0.18.

And if the medication is also sedating, then it should be given at night, especially if insomnia is a problem.

Philip Dawdy responds: my understanding of effect size is that it's the difference between the effect of the drug and the effect of the placebo, which gets you to .18. how is that wrong dguller?

There are other problems with this trial. This trial was successful (marginally, as you mention) only if you looked at the peole who stayed in the study for the full time and did not drop out. If you count all the people who entered the study, completers and drop-outs, there was NO difference between the two groups (drug v placebo) on the main outcome measure (anxiety). The correct way to analyze and make conclusions about a trial is to use ALL the people entering the study, not only the completers (called intent to treat analysis, ITT). If you use only the completers, you get a distorted view of the results: you do not count all the people who had problems with non-response or stopped because of adverse effects of drug (or other) treatment. As the number of drop-outs increases and the analysis is limited to only those who complete the study, the results become more and more distorted. (High drop out rates are big problems with psychiatric research in general). In any case, whether there are a small or large number of drop-outs, it is still scientifically correct to use ITT: ie, to base the conclusions on the people who enter the trial, not on those who complete it. So in this study using ITT, they found NO difference in response for anxiety but only adverse effects.

The problems of how trials are analyzed and waht methods are used have been have been discussed for some time by editors of medical journals. The recommendations are to use ITT to analyze all trials.

From Medpage Today:
An SSRI for Anxiety in Older Adults?

Older adults with generalized anxiety disorder may benefit from the SSRI escitalopram (Lexapro), according to a preliminary study in JAMA.

Researchers randomized nearly 180 patients aged 60 and older to receive escitalopram (up to 20 mg daily) or placebo for 12 weeks. In the primary analysis (which included all patients, but censored data after any dropouts), escitalopram recipients were significantly more likely than placebo recipients to be much improved or very much improved by the end of treatment (69% vs. 51%). However, in an intent-to-treat analysis (in which dropouts were considered nonresponders), there was no difference between the groups.

Noting that nonadherence is common in patients with anxiety, the authors speculate that the latter finding "may more accurately portray the results of treatment in clinical settings." They call for additional studies with more participants and longer follow-up.

Link(s):
JAMA article (Free abstract; full text requires subscription) http://click.jwatch.org/cts/click?q=227%3B67122456%3BI9qx2nuGIGGsc6DwLol%2FRam18hcSjTMmb445avMV8B0%3D

And what happens after the twelve week "intervention" I'd like to know? This is not a long term solution (even without the immediate adverse side effects) and should be acknowledged as having some serious issues beyond the very limited time frame in the study.

fatigue and somnolence

It always seems to me that the list of side effects don't give a clue of what do they really mean.

It seems that the patient can overcome them with a little effort.
Everybody has fight fatigue and somnolence, the natural feeling but when it is drug-induced we are talking about something quite different.

Philip:

Basically, effect size is calculated by the difference between the means divided by a standard deviation of the data. However, a better measure instead of a standard deviation is a pooled standard deviation. It is not just the difference between the means of two groups, but rather also incorporates their standard deviations and the number of subjects in each group.

Philip Dawdy responds: thanks. that's not how it's been explained to me by researchers who are generally ok with me using the differences between group means. but i'll look into it.

Eileen:

There was a difference between the two groups.

You are correct that using ITT analysis that used everyone who was randomized in the trial, found no statistically significant difference between the two groups in terms of the primary outcome measure, i.e. cumulative response rate using the CGI.

However, when a modified ITT analysis was performed that excluded subjects who only provided a single data point during the initial assessment (7 subjects, out of 33 who dropped out in total for a 18.5% dropout rate) there was a significant difference found between the two groups.

They also found a statistical difference between the two groups in terms of anxiety symptoms and role functioning.

It is important to note that ITT provides a conservative estimate of efficacy by purposefully lowballing the results. In other words, if something can be shown to be effective in an ITT analysis, then a stronger case can be made for its overall efficacy. That is because some people who dropped out of the study may not have improved on the medication, and thus should be included.

However, there are also problems with ITT analysis. You can read Evidence Based Mental Health (2008; 11: 3-5) for details.

Take care.

"Philip Dawdy responds: thanks. that's not how it's been explained to me by researchers who are generally ok with me using the differences between group means. but i'll look into it."

I'm not understanding.
Where has Philip said it?

From the abstract, it looks like the .93 "effect size" comes from the Clinical Global Impressions-Improvement scale. I suspect this is a clinician rating of the patients' progress... the effect sizes and confidence intervals for patient self-report questionnaires and measures of actual activity limitations are much less impressive. Which means, I guess, that the most enthusiastic ratings came from the doctors, rather than the patients.

It is hardly surprising that significant numbers of sedated and somnolent patients would be rated as "much less anxious" by doctors... although maybe not by themselves. Sedated patients certainly talk less, fidget less, and are easier to "manage." Hallelujah. Not.

Johanna:

From the study: "The main outcome assessment was the Clinical Global Impressions-Improvement scale for which raters synthesized anxiety rating scale scores, participant self-reports, and the rater's determination of degree of improvement."

In other words, it was a COMBINATION of clinician and subject information.

Hi dguller:

We agree that the study says that if you use one method of analysis which excludes some participants, there was a difference between the two groups, but if you use another method (intent-to-treat ITT) which includes all participants, there was no difference. The ITT method is recognized as being more unbiased and it is the standard recommended by medical journals and associations of medical journal editors, based on CONSORT recommendations. (Sorry- I cannot find the citation now). I am not sure what you mean when you say ITT will "purposefully lowball" the results of trials. ITT is more rigorous and that is why it is supposed to be used in clinical trial analysis. Yes, the results may be different (lower) than with less unbiased methods and that is an important issue to be considered in evaluating the results of the trial. One might just as well say that less rigorous methods may "purposefully overstate" the efficacy. There are inevitable problems in all methods of data analysis. Obviously we disagree about how to interpret the results based on the different methods of analysis. I understand the need for researchers to show difference between groups in clinical trials. But some of us believe that using less rigorous methods to show diferences between groups results in overstatement of efficacy and overpromotion of drugs with limited and marginal efficacy. Put another way: there are so many limitations in most psychiatric research over and above this problem of how to handle the data (ex., selected populations who are statistically expected to respond; many exclusions for populations not expected to respond like people who did not respond to similar drugs in the past or people with common comorbidities like substance abuse; low bar for "success" in outcome measures; short duration; no standard way to measure adverse events; poor coding of adverse events; unblinding due to obvious adverse events like weight gain; many outcome measures some of which will be positive; high drop-out rates limiting conclusions; few studies on effectivness; few studies on polypharmacy despite common use, etc) that some people have come to believe that the results of many trials are not valid, reliable, generalizable and are not scientific. We also need to factor in researcher conflict of interest. I guess some of us have concluded that advocating for or focusing on the use of certain methods of analysis that show efficacy over others that show no efficacy is another way that results of trials may not be as objective and transparent as we think they should be.