Toward the end of the 1990s, hiding the suicide risk on antidepressants by unearthing ghost suicides and suicidal acts from the early washout phase of trials looks like it might have seemed to company and FDA officials as problematic as Macbeth’s invitation to Banquo to make sure he came back to the feast later that evening. A new strategy came to the fore. Again one of the earliest versions of this had been pioneered by Lilly. (It would be interesting to nail down the very earliest version.)
This was a statistical significance strategy — or how many bullets does the barrel of a gun have to contain before you say it’s loaded? The first step in this strategy was to focus on placebo-controlled trials only. This greatly reduces the numbers of patients under consideration, which was crucial to the success of the strategy. Reducing the numbers of patients meant that even if there were more suicides on Prozac or Paxil, the figures could never be statistically significant.
In the case of Prozac, the placebo-controlled trials had shown that there were five suicidal acts from 1398 patients on Prozac versus one (ghosted) suicidal act in 645 patients on placebo. These data show double the number of suicidal events on Prozac, but the data are not statistically significant. This is what was published in the BMJ where it was read by up to 100,000 people, most of whom had no links to the pharmaceutical industry, many of whom might have been hostile to industry. But when the company wrote that data from these trials do not show that fluoxetine is associated with an increased risk of suicidal acts, there was not a peep — no one said, “Wait a minute, there is clearly an increased risk – a 12-year-old schoolboy could tell you there is an increased risk.”
|Number Acts/Number Patients
(95% C.I, 0.27, 19.7)
It seems likely regulators and company personnel took note. SmithKline soon afterwards took the same approach to Paxil, where the figures looked as follows.
|Number Acts/Number Patients
(95% CI 0.4, 25.7)
There is no need to breach regulations by moving dead bodies or suicidal acts around when taking this approach. The old way involved moving one dead body in Lilly’s case, and four suicides or suicidal acts in SmithKline’s case. This way 10 suicidal acts vanish.
Omitting the ghost suicide from the Prozac figures and adding the two sets of figures together gives a 5.2 fold greater risk on Prozac/Paxil than on placebo. The 95% confidence interval is 0.66, to 40.3. This means that if you’re dead set against conceding there is a risk, you can argue there is a small chance that these results happened by chance; but you have to concede there is a roughly equal chance that Prozac and Paxil are up to 40 times more dangerous than placebo.
A failure to tackle the issues
A lot of people in companies or regulatory agencies must have put these figures together and added in the Zoloft figures also, which make the problem look even worse. The regulators likely justified not doing so on the basis that their bureaucratic brief was/is to consider each drug in isolation. Company reviewers are regularly moved from desk to desk and not encouraged to remember what they might have been doing last week or to put pieces of the jigsaw together. Doctors are busy.
This applies not just to selective serotonin reuptake inhibitors (SSRIs) and suicides but to heart attacks or other events on pain-killers, like Vioxx, and hypoglycemics, like Avandia. There is nobody in the system it seems with an incentive to get to grips with the issues.
The consequences are not just avoidable deaths from heart attacks, suicides, and drug-induced homicides, but more to the point a close to North Korean level mass psychosis. Both patients and doctors report suicides, heart attacks, or other problems, but are told there is no scientific evidence to back up what they are saying, when in fact the science and their personal testimonies match precisely. Their heart attacks or suicide attempts are backed up by the clinical trial data, which shows an increased risk from the drug they are on. But they are told by regulators such as Bob Temple of FDA that while their stories are harrowing, the agency has to go by the science and the science does not show a risk.
The evidence of our own eyes
Freud led us to disbelieve the obvious and go by what experts told us stories about abuse meant. In an entirely comparable manner, we are now being told we cannot believe the evidence of our own eyes — that our experience is anecdotal. In fact 80% of the time, possibly more, when doctors report on adverse events they turn out to be right — from heart attacks to suicides to changed hair on oral contraceptives. The changed hair on oral contraceptives probably means that when hairdressers talk about side effects of treatment they are also likely to be right, and it’s patients rather than doctors who typically make the links to strange side effects from suicide to wetting the bed on a new drug.
The question is how do we treat this psychosis?
This does not seem to be a time to desert science but rather to embrace it. But it is a time to recognize that words come before numbers.
Doubt is our product
It may also be a time for humanists and post-modernists to come into their own. They have spent decades trying to deconstruct science but never did the job as well as drug companies now do. Holding out for statistical significance until you concede that a drug might have caused something is the ultimate manifestation of the Doubt is our Product strategy pioneered by the tobacco companies. And Doubt is our Product is the perfect caricature of post-modernism.