In the mid-1990s, an incredible thing happened. The National Institute of Mental Health (NIMH) decided to fund major clinical psychopharmacology studies. About 90% of NIMH is for other purposes, so the world of clinical psychopharmacology always has suffered from neglect in academic research. The field has relied on the pharmaceutical industry for its largest and most definitive studies, with all the limitations already discussed. NIMH leadership, almost always laboratory researchers without background in clinical psychopharmacology, has tended to see this problem as a solution. Government funding wasn’t needed where private industry funding was present.

By now it should be obvious how harmful this attitude has been.

Besides the scientific reasons for government funding of psychopharmacology research, the mid 1990s provided an opportunity in the United States for the first and only time in the past half century. For a few brief golden years, the federal government under the Clinton administration had a budget surplus. The NIMH had extra money, and it decided to spend a little more, finally, on clinical psychopharmacology.

Those budgets may have seemed large to academic researchers, but they weren’t huge – about $20 million each for the three studies funded in the three major psychiatric diseases of severe depression, bipolar illness, and schizophrenia. Much has been made of these studies, and they will be discussed in detail here, but it is relevant to note that $20 million is tiny in the world of pharmaceutical research. That amount is spent on just a few phase II trials, with about 100 subjects or so, just to get to the point of deciding whether to do much larger phase III trials for marketing approval. Those phase III trials tend to run in the hundreds of millions of dollars in psychiatry for each drug. So $20 million to study most available major drugs for a major illness is not much.

But these few drops of water felt wonderful to the parched throats of the field of government-funded clinical psychopharmacology.

Of course, the federal budget quickly fell into debt and has remained there for the past three decades. These studies will not be repeated, and with all their limitations, they are and will remain for the near future the best data available on clinical psychopharmacology with current agents outside of pharmaceutical industry studies.

These studies thus have been called the “canons of psychiatry” when used for teaching in a family medicine setting; they are a canon, but one that has been misread.

The new canon of psychopharmacology

So what is this new canon of psychiatry, or more properly, the new canon of psychopharmacology?

We have three studies – for schizophrenia: Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE); for so-called major depressive disorder: Sequenced Treatment Alternatives to Relieve Depression (STAR*D); and for bipolar illness: Systemic Treatment Enhancement Program for Bipolar Disorder (STEP-BD). These grants were contracts given to consortia of universities: $20 million or so would be spread about between 10 or so universities. The universities would be sites in a single trial, designed by a core site. CATIE was headed and centrally designed out of Columbia University, while STAR*D and STEP-BD were headed and centrally designed out of Massachusetts General Hospital (MGH)/Harvard Medical School. I participated in some of the early work on STEP-BD at MGH, and later was a subsite in that study at Cambridge Hospital/Harvard Medical School; so I observed some of the sausage-making process, as well as the later presentation of results.

Most of the research was conducted in the early 2000s, and then published later in that decade and into the 2010s. Hundreds of papers have been produced, and careers made and advanced with this work.

Let’s review how the studies mainly have been interpreted and presented, and then look at alternative interpretations of those studies, focusing on the statistical methods that will adjudicate what are the most correct interpretations.

The main rationale for these studies

There are some common rationales for these three studies. The first is that pharmaceutical industry funding tends to take a cookbook approach to satisfy regulatory standards (from the Food and Drug Administration, FDA, in the United States) to come to the commercial market. These standards require that each drug be shown to be more effective than placebo in two separate trials. In American psychiatry, there generally is no FDA requirement that a drug be shown to be equal or better than prior proven treatments. Hence we have hundreds of studies comparing single antidepressants or antipsychotics versus placebo, as reviewed previously. But we have very few studies directly comparing individual antidepressants or antipsychotics to each other. “Network” meta-analysis is a false solution to this problem.

Hence clinicians in practice do not know if one drug is better than another, or what they should do if one drug fails. They basically have to guess.

These studies had a general principle of trying to help clinicians answer those questions. Hence the CATIE and STAR*D didn’t have placebo arms, since efficacy of individual drugs over placebo was considered proven already. Instead, they were designed mainly to compare each drug to another.

A related rationale for conducting these studies was to address the concern that pharmaceutical industry-conducted or sponsored studies tend to be biased in favor of the drug being studied. For those who harbor doubts regarding the biasing effects of the pharmaceutical industry, the conduct of studies by academic groups without commercial consequences, funded by the federal government, would seem to make it more likely that results would be less biased in favor of drugs.

A final comment is that STEP-BD differed from CATIE and STAR*D in including a placebo arm. Unlike the immense literature in schizophrenia and depressive illness, the bipolar literature no antidepressants was sparse. Hence STEP-BD had much more modest goals than CATIE or STAR*D; it simply wanted to see if antidepressants were better than placebo. Despite its simple design, even STEP-BD has been misinterpreted in practice.