Can FDA be relied upon as guarantor of drugs it approves?

 

Can FDA be relied upon as guarantor of drugs it approves?

Posted by Y Rabinovitz

Apr 12, 2022

Scientific, or rather, pharmacologic deception predates COVID era

• How do researchers assess psychiatric medication with nothing tangible to observe?
• Are clinical trials for new psychiatric drugs really placebo-controlled?
• Can researchers be relied upon to present their data in an accurate manner?
• Are accepted standards of efficacy adhered to when approving new (and very expensive) drugs?
• Can the FDA be relied upon as a guarantor of efficacy of the drugs it approves?

As medicine moved into a new age beyond treating the obvious – visible injuries, nasty diseases with obvious symptoms and so forth – new ways of assessing treatment efficacy developed, chief among them the clinical trial. To cut a very long story short, the basic idea today is that when you pick up a drug from the pharmacy, you can feel confident that a bunch of scientists tested it rigorously for efficacy and safety and that the results of their trials were then passed to the FDA or its foreign equivalent. The FDA reviewed trials for accuracy and bias, approved the drug – and there you go, a solution to whatever was ailing you.

That was potentially simple when dealing with a nasty germ that was exterminated on a petri dish by the latest antibiotic. But what happens when the “disease” is entirely intangible – nothing to be seen on a monitor or in a test tube? In such cases, clinical trials have to rely on the reporting of the trial participants themselves. Modern psychiatry has developed a variety of scales to rate the severity of mental illness and gauge how well a drug alleviates its features.

For schizophrenia, the main scale in use is called PANSS, the Positive and Negative Syndrome Scale. It assesses things like delusions, hallucinations, apathy, lack of spontaneity, disorientation, poor impulse control, and social avoidance, assigning a score between 1 and 7 for each of the 30 items on the list for a total that can range between 30 and 210. The question then arose: When gauging the efficacy of a new drug to treat schizophrenia, how much improvement should one look for in a PANSS score? Many studies have been made to develop what is called the MCID – Minimum Clinically Important Difference. The basic consensus is that an improvement of at least 15 points is necessary to declare a drug effective.

On a scale of 30 to 210, 15 really isn’t very much, although of course depending on where the added points stack up, it could actually translate into significant improvement. But at least now that there are numbers to rely on, and the FDA to check that they actually exist, we can feel confident that any drug prescribed by a GP or psychiatrist is actually going to help.

Right? Let’s see.

The example we will be looking at here is a drug called Rexulti (brexpiprazole), used to treat schizophrenia, although many more examples are available. Rexulti went through the usual stages of clinical trials.  You can see the summary of its Phase 3 study here. Around 700 people participated in the study, randomized into four groups: placebo, 0.25 mg/day, 2mg/day, and 4mg/day. The study ran for six weeks, and the authors made it easier to assess the results (a long series of tables with lots and lots of figures) by printing them nicely in a graph.

Here it is:

It’s quite clear from here that placebo did worst of all and that any amount of the drug did significantly better. Or is it?

Take a look at this graphic instead:

This graph doesn’t appear in the article describing the Phase 3 study. It was produced by Robert Whitaker of Mad in America (see here).

To produce the more dramatic graph, the study’s authors took a small section of the total graph and magnified the y-axis by a factor of 7.2 (180 divided by 25). The question is: Why?

After all, the study results went to the FDA; the graph wasn’t produced for advertising to the general public. But they approved Rexulti in 2015 and over the next four years, Otsuka and Lundbeck made $1.4 billion from Medicare and Medicaid sales alone.

 

Placebo effects

Looking back at the graphic, we see that a placebo improved PANSS scores by around 12, the lowest dose of Rexulti by 15, and the 2mg and 4mg doses by around 21. In other words, just 9 points of improvement can be attributed to the drug, significantly less than the 15 noted above. Yet the FDA still approved Rexulti.

Will your psychiatrist tell you this as he writes out a prescription for this latest wonder drug? Will your GP tell you? 

But there’s more. What does “placebo” actually mean in a trial like this one?

One might have expected that a study such as this would examine the effects of Rexulti on people experiencing schizophrenia symptoms and free of all drugs (or, at least, pharma drugs) prior to the study’s onset. That is not what happened (and this is typical of psychiatric drug trials):

Almost all patients were taking antipsychotics before the study (579 of 636, 91.0%).

The placebo cohort comprised people who had been on powerful drugs for an undisclosed (and possibly unknown) length of time. To participate in the study they were abruptly weaned and given no substitute. While it is common (although not advisable) to transfer psychiatric patients from one drug to another without a long weaning period, what is less common (although unfortunately also found) is to transition people off psychiatric medication abruptly (aka going “cold turkey”). The results of abrupt discontinuation of psychiatric medication (and this includes commonly prescribed drugs in the benzodiazepine class) can be dramatic, even life-threatening.  They are physiological, not just psychological.

The study makes no effort to describe this possible effect on the trial results, nor to control for it. Instead, it dismisses this effect as being “related to the underlying condition”. (This too is common in psychiatric practice, when withdrawal effects from a drug are interpreted as “signs that the disease is returning” and used as justification for restarting the drug, even in cases where the person had ceased using the drug due to intolerable side effects.)

What is noted is that the “overall incidence of treatment-emergent adverse events was lower in the three brexipiprazole groups than in the placebo group.” This is an interesting way of terming side effects.  The problem, of course, is that the trial has conflated side effects with withdrawal effects from those discontinuing their previous medication in the placebo cohort.

There is also no information on how long trial participants had been on antipsychotics, but what is noted is that many were on a cocktail of other drugs as well and continued to use drugs such as lorazepam (an anti-anxiety drug) throughout the trial, regardless of whether they were in a trial or a placebo cohort. How this may affect the trial results is similarly neither described nor controlled for.

“Overall, 410 patients completed the study.” The reasons for which the others dropped out are not detailed; they are not even listed. Many trial participants will have dropped out after realizing that they were on a placebo and experiencing the significant discomfort that accompanies withdrawal. Others will have dropped out because of intolerable side effects. Given the divergent experiences of those in the trial, can one truly call it blinded?

 

Conclusions

The trial’s authors concluded that

… brexpiprazole at a dosage of either 2mg or 4mg/day resulted in statistically significantly greater improvement than placebo as indicated by the primary outcome measure, change in PANSS total score…

and that this was

… not only statistically significant but also clinically relevant, indicated by a significantly greater change in CGI severity score…

An increase of 9 points on a 180-point axis would not seem significant or clinically relevant to many people, but the FDA evidently though it was.  This may be because they are not expecting much better, since

… the numbers needed to treat for response (>30% improvement in PANSS total score…) were 6 for 2mg and 7 for 4mg. Both the effect sizes and numbers needed to treat are within the midrange across examined antipsychotics…

In other words, most drugs don’t do any better.

 

The study’s authors also noted that

… the results of this study need to be interpreted within its limitations. These include the lack of an active comparator, short study duration, and inclusion of patients with schizophrenia without other psychiatric comorbidities. However, this study was designed as an acute, short-term regulatory approval study…. Longer-term studies that include a comparator, as well as patients with common comorbid conditions, will be needed to further evaluate the efficacy, safety, and effectiveness of brexpiprazole…

A trial with an active comparator is one in which the trial cohort is matched with a cohort on another, similar drug, rather than those not on medication. While also not ideal, this at least eliminates much of the concern regarding withdrawal symptoms muddying the trial results. From the perspective of the FDA, the consideration might also be whether there is any justification for approving a new (and very expensive) drug with a long list of side effects when pre-existing drugs are just as effective at treating the same disorder.

Several further studies have been conducted to test Rexulti; none has shown a 15-point improvement in PANSS scores. The FDA has not reconsidered its approval of Rexulti; it remains a much-prescribed, and highly lucrative drug.