September 18, 2018.
After drinking water was fluoridated in the 1950s, the rate of tooth decay dropped by 50%. As more and more states mandated the use of seat belts in the 1980s, motor vehicle fatalities (per miles traveled) steadily declined. Should we not apply the same measure of success to drugs that Big Pharma dispenses on a grand scale -- like antidepressants?
In 1987 most patients with Major Depressive Disorder (MDD) were treated with psychotherapy; only 37% were prescribed antidepressants. But that rate jumped to 75% by 1997 and reached 80% in 2016, thanks to Big Pharma's triumphant marketing of the new Prozac strain of antidepressants (SSRIs). Figure 1 depicts the industry's success: the percentage of Americans wielding a prescription for an antidepressant rose from less than 2% in 1991 to 11% in 2014.
However, along with the "Prozac boom" came serious doubts about the effectiveness of antidepressants. Fortunately for the purveyors of pills, a reanalysis of past clinical studies concluded that SSRIs manage to quell symptoms 20% of the time. Since this finding was touted in industry trade journals, I will accept it as gospel for the purpose of this post (even though the real figure is about 12%).
But here's the mystery: If antidepressants are truly effective, why has the dramatic increase in their use NOT reduced the prevalence of major depression? To borrow from Robert Solow, "We see antidepressants everywhere, except in the mental health statistics."
The fact is, the percentage of Americans diagnosed with MDD has not changed over the course of two large national surveys (figure 1). While antidepressant prescriptions quadrupled between 1991 and 2001, the NCS surveys for those two years showed no change in the prevalence of MDD. Likewise, no decline in MDD is evident in the NSDUH surveys from 2005 to 2016; yet prescriptions climbed another 60%.
Furthermore, depressed women are significantly more likely to seek treatment than their male counterparts. In the US, 65% of the long-term users of antidepressants (>5 years) are women over 45. This suggests that the "Prozac boom" engaged women more than men, so we should have seen a narrowing of the huge gender gap in the the prevalence of major depression. But there's no evidence of that. In 1993 the female to male prevalence ratio for major depression was reported as 1.76, and in 2016, a NSDUH report put the ratio at 1.77.
Figure 1
The same pattern has been observed in Canada, England, Australia and New Zealand. If anything, depression rates there seem to have gotten worse with the rising popularity of antidepressants.
As a quick and easy alternative to talk-therapy, SSRIs raised the rate of treatment for major depression. In 1991, only 46% of major depressive episodes were professionally treated; by 2014, the treatment rate reached a high point of 69%. Treatment was now being extended to minor depression, which is another reason we should have seen a decline in the prevalence of Major Depression. Treatment of mild cases should have prevented escalation of the disorder.
Assuming that SSRIs really are 20% effective, this huge increase in treatment since 1991 should have cut the prevalence of MDD to 3.5% by 2014, according to my calculation (dotted line on figure 1).
Why did this decline fail to materialize? One explanation is that the causal risk factors for depression were actually worsening, but their effect was counteracted by a corresponding rise in antidepressant treatment. In other words, were it not for the "Prozac boom," rates of major depression would be getting a lot worse.
To test that hypothesis, I assembled time-series data on the prevalence of the "big-five" risk factors* for major depression. Listed in descending order of their causal potency* they are: Economic Distress, Substance Abuse, Personal Trauma, Low Education and Social Isolation.
The plots on figure 2 show the change in the risk factors' prevalence since 1991. Only two of the five got worse after 1991: the percentage of the population afflicted by Social Isolation and Substance Abuse increased 2 percentage-points by 2016.
However, that bad news was far outweighed by the declining prevalence of Personal Trauma and Low Education since 1991. Finally, the Economic Distress index stands out because it is strongly correlated with the business cycle. The economic boom that followed the recession of 1991 produced big declines in the rates of unemployment and poverty. That was temporarily reversed by the great recession in 2008-2010; for all the other years, economic distress was much lower than in 1991.
Figure 2.
Clearly, figure 2 does not support the notion that a rising tide of depression has been held back by the dissemination of antidepressants. To the contrary, the overall change in risk factors is in a direction that should have reduced the prevalence of depression, especially from 1994 trough 2007. So the predicted drop in MDD depicted in figure 1 is quite conservative.
Then why haven't we seen even a hint of decline in the prevalence of MDD?
The obvious explanation: antidepressants are NOT as effective as the syndicate of psychiatry and Big Pharma claim. Estimates of efficacy are based on selective clinical experiments, and cannot be generalized to real world patients. Clinical studies deliberately exclude subjects that come with baggage: alcoholism, drug abuse, psychiatric disorders, heart disease, and so on. The excluded are less likely to be helped by antidepressants, but are more representative of the people who show up in doctors' offices with depressive symptoms.
Furthermore, the results of clinical trials are inflated by bias...
Some clinical studies have actually excluded subjects known to be unresponsive to antidepressants. Then there's the "short-term bias" because clinical studies usually run 6 or 8 weeks. The estimate of 20% efficacy cited above comes from a meta-analysis that summarized the results of 41 trials at the 6th week only. So it doesn't report the results of long term trials, such as the famous TADS study of adolescent depression. It compared Prozac with non-drug treatment over the course of 36 weeks. While Prozac exhibited superiority in the first six-weeks, its advantage diminished to nothingness by the 24th week, and at week 36 the non-drug treatment group finished with the highest remission rate. However, the one thing the Prozac group did excel at was suicidality. The same lack of long term effectiveness applies to adults as well, especially when measurement-bias is removed.
Also dismaying is the (unreported) failure of clinical trials to achieve blinding. Blinding is supposed to prevent researchers and subjects from identifying members of the treatment and control groups. Even though blinding is absolutely essential for achieving unbiased results, only 2% of health-related clinical studies bother to test whether it is actually achieved.
So what's the big deal? Well, that "2%" was calculated by researchers who scoured thousands of health-care trials to identify the cases where a test was actually performed. Of those cases, blinding was successful in 45%, uncertain in 32%, and failed in 23%. This suggests that the probability of a "broken blind" is around 33%. For antidepressant trials in particular, the risk of a broken blind is probably greater because they fail to use an active placebo. (Members of the treatment group are tipped-off by the early side effects of the antidepressants, while the those on an inert placebo are not similarly cued).
Finally, it is important to note that many authors of clinical studies receive fees and honoraria from Big Pharma.
All of the above conspire to inflate estimates antidepressant efficacy. The newest estimate comes from a blockbuster meta-analysis published this year. Its conclusion: the average SSRI is more effective than a placebo by an odds ratio of about 1.67, which means...
For the carefully vetted patients in clinical trials the chances of significant improvement from taking Zoloft are 12% better than taking a placebo. In other words, the drug helps just 1 out of 9 subjects in the treatment group.
And that estimate is NOT adjusted for the research flaws enumerated above. So if you are a typical "real world" depressive, and want a treatment that outperforms a sugar pill over the long, betting on an SSRI is really a long shot.
_______________________________________________________
*Risk Factors:
Economic Distress = percent of adults who are unemployed >27 weeks OR below poverty line OR whose net worth <= zero. The strongest evidence for a causal connection to depression is for unemployment and poverty, which is why Economic Distress ranks first.
Substance Abuse = percent of adults who are heavy users of cannabis OR alcohol OR opioids
Personal Trauma = percent of adults who are victims of violent crime OR in poor health OR died before reaching age 65 (i.e., death rate for children and non-elderly adults)
Low Education = percent of adults with less than a bachelor's degree (which closely tracks the percentage with less than a high school degree). The causal potency of this factor ranks low because it so highly correlated with Economic Distress.
Social Isolation = percent of adults who are living alone OR widowed OR divorced OR stay-at-home-moms. Its causal potency ranks last because it is a surrogate for the real risk factor - loneliness. Social isolation turns out to be a poor measure of loneliness.
Comments