As cannabis use surged in the late 1960s and ’70s, psychiatrists who had rarely dealt with the drug before saw its effects up close—and reported their findings:
Forty-six cases of psychosis in cannabis abusers (International Journal of the Addictions, 1972).
Psychotic reactions following cannabis use in East Indians (Archives of General Psychiatry, 1974).
Cannabis-associated psychosis with hypomanic features (Lancet, 1982).
Cannabis psychosis in south Sweden (Acta Psychiatrica Scandinavica, 1982).
For the most part, the reports didn’t make sweeping conclusions about marijuana’s effects on the brain. Psychiatrists and emergency room physicians were simply trying to manage psychotic breaks in smokers. Did they need antipsychotic medicines? Or could they simply be left in seclusion until their hallucinations and delusions faded? Fortunately, the crises seemed temporary, at least in people without underlying mental illness. After a day or two, most patients calmed down.
At the same time, doctors saw cannabis provoking new psychotic episodes in people who already had schizophrenia. A 1978 case report in the American Journal of Psychiatry discussed four patients whose symptoms worsened when they smoked. “The sole substance abused was marijuana,” the report said. “Each time marijuana use at moderate levels began, there was exacerbation and deterioration.”
But many psychiatrists viewed the reports as a curiosity. Then as now, the specialty leaned to the left. In 1964, almost 1,200 psychiatrists had signed a letter arguing that Barry Goldwater, the Republican candidate for president, was “psychologically unfit” for the office. (The criticism that followed led the American Psychiatric Association to say none of its members should offer a diagnosis of a public figure unless he or she had examined the figure personally. Known as the “Goldwater rule,” the restriction remains today.)
The dislike ran both ways. In his book The Selling of the President, the journalist Joe McGinniss recounted that Richard Nixon’s campaign wouldn’t let a psychiatrist on a panel that would question Nixon. “Nixon hates psychiatrists,” a Nixon advisor told McGinniss.
Three years later, Nixon’s own Oval Office tapes would catch him complaining to an aide about psychiatrists—specifically Jewish psychiatrists—and the nascent cannabis legalization movement:
Every one of the bastards that are out for legalizing marijuana is Jewish. What the Christ is the matter with the Jews, Bob, what is the matter with them? I suppose it’s because most of them are psychiatrists, you know, there’s so many, all the greatest psychiatrists are Jewish . . .
Nixon was wrong, at least about the two men who would become the most important advocates during the 1970s. Neither Keith Stroup nor Tom Forcade was Jewish—or a psychiatrist. But the fact that psychiatrists like Lester Grinspoon were among the doctors best-known for favoring marijuana use couldn’t be argued. (Grinspoon’s cannabis advocacy makes him a hero to modern-day legalizers; there’s even a marijuana strain called Dr. Grinspoon, “for connoisseurs and intellectuals.” His similar work on behalf of cocaine has been forgotten; in 1978, he said cocaine was less dangerous than alcohol or tobacco and the penalties for its use much too harsh.)
The chasm between Nixon and psychiatry meant that many psychiatrists in the 1970s were in no hurry to sound an alarm on marijuana’s potential dangers. But the debates about marijuana’s risks in the 1970s suffered from an even more fundamental problem. Case reports don’t prove anything. Yes, some people with schizophrenia broke down after smoking. But maybe others didn’t. Without knowing the denominator, the numerator was meaningless.
Scientists make a distinction between “hypothesis-generating studies” and “hypothesis-testing studies.” Hypothesis is a fancy word for theory. It’s easy to generate a hypothesis: Maybe preservatives are good for you. They preserve hot dogs, don’t they? Almost any theory can look plausible if you squint hard enough.
Figuring out whether the theory is true is much harder, especially if the theory is that a drug or activity is harmful to the people who use it. Scientists can’t ethically test something on people to see if it might hurt them. Ever since the disclosure of the repulsive experiments that German and Japanese scientists conducted during World War II, that restriction has been a core principle of medical research—and it should be.
The case reports on cannabis in the 1970s were hypothesis generating. By showing that cannabis could cause temporary psychosis, they raised the bigger question of whether the drug might have serious long-term effects, worsening or possibly even causing schizophrenia.
But for a decade, no one made any progress in testing that theory.
Then Sven Andréasson had an idea.
Andréasson is a Swedish physician who specializes in addiction medicine and alcoholism. He practices at the Karolinska Institute, the Stockholm medical university responsible for choosing the Nobel Prize in medicine each year.
In the early 1980s, Andréasson had just started practicing. He noticed that the schizophrenic patients who relapsed after being released from his hospital were often the ones who used cannabis. The patients came back “with much more florid hallucinations or disordered thinking,” he told me. Of course, Andréasson wasn’t the first doctor to wonder if cannabis might be connected to schizophrenia. But unlike everyone else, Andréasson and his supervisor, Peter Allebeck, had a way to test their theory.
At the time, Sweden had a military draft, with universal male conscription. Sweden being Sweden, this draft wasn’t an American-style process that well-connected men could avoid. Practically every male Swede served. During the intake process, conscripts filled out two questionnaires, one about their education and upbringing, the other about drug use. The military used the information to figure out what jobs to give recruits. The questionnaires were saved, but the personally identifying information in them was supposed to be destroyed.
However, the questionnaires from the 1969–70 year hadn’t had their identifiers removed. A Swedish military psychologist had wanted to create a database for future use, Andréasson said. But the psychologist never used the files. The tapes that contained their results sat moldering in a basement. They had been scheduled to be destroyed until Allebeck saved them.
When Allebeck mentioned the files, Andréasson realized their value. They covered almost fifty thousand conscripts, a full year’s population of Swedish men. Under normal circumstances, a researcher would be thrilled to have a dataset one-tenth that size. Many studies cover only a few hundred people. But teasing out the factors behind an uncommon disease like schizophrenia is difficult without a big dataset—a random 2,000-person sample might include only 15 people who develop the disease.
Further, Andréasson didn’t have to worry that the data might have a hidden bias because researchers hadn’t attracted a fully representative sample. Every Swedish conscript had been surveyed, full stop. Ninety-two percent of those had filled out the survey about their drug use. The background questionnaire was comprehensive, too, as one would expect from a survey created for military use. And the men were the right age. They had mostly been born in 1951 and been 18 or 19 when they filled out the questionnaires. They’d come of age in the late 1960s, as marijuana use was rising, and been surveyed exactly when men began to develop schizophrenia.
“We had access to so much data about these conscripts—there were hundreds of items in these questionnaires that they all answered,” Andréasson said.
But the questionnaires were only half the puzzle. For them to mean anything, Andréasson had to have outcome data. In other words, he needed to know what had happened to the conscripts since they’d served, so that he could see which of them had wound up with mental illness. Fortunately, Sweden’s national health care system gave him the chance to find out. Sweden tracks hospitalizations, and because the military surveys contained personal identifying data, Andréasson could see exactly which conscripts had been hospitalized.
“We had a complete inpatient registry for this group, and in particular a psychiatric inpatient registry, which in the case of schizophrenia is very good,” Andréasson said. “If you really develop schizophrenia, you’re likely to have been treated in a hospital.” Even criminals wouldn’t be missed, since Sweden screens prisoners for mental illness.
To be sure he had captured most cases, Andréasson sampled the inpatient registry. “We didn’t take it for granted—we did a number of investigations.” He found it was comprehensive.
He also checked the honesty of the self-reported drug use on the conscript questionnaires and found they were accurate, too. That fact was less surprising than it seemed, because draftees had been given the option not to fill out the survey if they didn’t want to reveal their drug use at all.
With a bit of luck, Andréasson had found every researcher’s dream. He had a large database that contained not just the key variable he needed but lots of secondary data too. Plus, another database that exactly captured the outcome at issue. All that was left was to compare the two.
So, he did, running the numbers to see whether conscripts who used cannabis before 1970—ever, once in a while, or frequently—had developed schizophrenia by 1983.
Wow. That was Andréasson’s first reaction when he saw the results.
Use of cannabis was strongly correlated with schizophrenia. And the risk was dose-related. In all, the questionnaires covered 45,570 conscripts. By 1983, 246 of them had been diagnosed with schizophrenia, or 0.54 percent. That figure was relatively low by international standards, reflecting the fact that Sweden set a high bar for the diagnosis.
Though cannabis use was rising in Sweden in the late 1960s, it was still relatively low. Only 4,290 conscripts had used the drug even once. They accounted for 49 cases of schizophrenia—a risk of 1.16 percent. Put another way, smoking even once more than doubled the risk.
But it was the risk in the heavier users that really jumped at Andréasson. Of the 752 conscripts who said they’d smoked fifty times or more, 21 later developed schizophrenia, a 2.8 percent risk—six times as high as people who had never smoked.
“We had a hypothesis that cannabis was a contributor to psychiatric disorder,” Andréasson said. “But this was very much more powerful than we had expected.”
The next step was trickier. Andréasson had to check to make sure the results weren’t confounded. In other words, he had to check that the association that seemed to stem from cannabis didn’t actually come from some other factor.
Confounding sounds complicated, but a hypothetical example may help explain it. Say, researchers find that people who buy air fresheners get lung cancer more often than those who don’t. Uh-oh! Febreze causes cancer. Procter & Gamble’s in trouble! Well, maybe not. Before those researchers blame air fresheners, they’d better figure out whether the people who buy them are also cigarette smokers. Because if they are, then fresheners don’t cause cancer—those are just two things you get if you smoke.
In the case of schizophrenia and the Swedish conscripts, Andréasson knew several factors resulted in higher odds of schizophrenia—most important, a family history of mental illness. He had to adjust for those risks, along with other factors that might plausibly cause the disease, such as other drug use.
Andréasson knew most conscripts who developed schizophrenia were unlikely to have only cannabis use and no other risk factors. Some might have a family history of mental illness. Others had other risks. Each adjustment might reduce the strength of the statistical link that he had found. But he had to check all the potential confounding variables he could imagine. Otherwise the initial finding hardly mattered.
Fortunately for Andréasson, the huge size of the database—and the strength of the initial finding—worked in his favor. “Even when we considered all these background factors, we still saw the risk,” he said. Use of cannabis more than ten times raised the risk of developing schizophrenia by 2.3 times even when Andréasson accounted for eleven different confounding factors. That adjusted risk wasn’t quite as high as the “raw” figure—but it was still big.
Andréasson wrote up the findings and submitted the paper to the Lancet, the prestigious British medical journal. He highlighted the sixfold risk increase among heavy cannabis users. “Persistence of the association after allowance for other psychiatric illness and social background indicated that cannabis is an independent risk factor for schizophrenia,” he wrote.
He expected he wouldn’t hear anything for months. Winning publication in a top medical journal is very competitive. Many researchers spend their entire careers hoping to place a single article somewhere like the Lancet. Instead, the journal’s editor wrote back almost immediately to say he wanted to run the paper as soon as possible.
“I was a young investigator,” Andréasson said. “I had no experience with being treated with that kind of interest.”
The Lancet published the paper, “Cannabis and Schizophrenia: A Longitudinal Study of Swedish Conscripts,” on December 26, 1987. For the first time, a researcher had moved past case reports to demonstrate an actual statistical link between marijuana and schizophrenia.
The fact that risk increased with use was especially important. A core principle of epidemiology is that if X causes Y, then more X should cause more Y, exactly as Andréasson’s data showed. “The more you used, the more risk you had,” he said.
Later, Andréasson conducted a follow-up study to examine whether people who developed schizophrenia after smoking were different than schizophrenic patients who had not smoked. He found that smokers tended to be relatively high-functioning before their illness, while nonsmokers had been troubled from a much younger age—a more classic presentation of schizophrenia.
Based on his data and later findings, Andréasson says he believes that cannabis is responsible for between 10 percent and 15 percent of schizophrenia cases. Few people develop schizophrenia solely because of smoking, he thinks. But many who would not have become sick do so because marijuana pushes their vulnerable brains over the edge.
“Without cannabis, fewer people would develop the disorder,” he says.
Andréasson hardly has rose-colored glasses about alcohol. He authored another paper based on the conscript data showing that even light drinking increased death rates. He is chairman of the Alcohol Policy Forum, a nonprofit group that helps craft Sweden’s restrictive alcohol laws. But Andréasson believes that cannabis is significantly more dangerous. If people used marijuana as much as they drank, “we would see an enormous amount of morbidity from cannabis,” he says.
The Lancet paper generated immediate interest. In the pre-internet era, many researchers wrote Andréasson about the findings. He even recalls an imprisoned child molester writing to ask if marijuana could have caused his behavior. In the thirty years since, other scientific studies have cited the article more than a thousand times, a huge number for any academic paper.
But the paper didn’t end the argument on whether marijuana caused mental illness. In fact, as the 1980s came to a close, the debate was only beginning.
In the United States, so was an ultimately successful attempt to rebrand marijuana—a drug used as an intoxicant for thousands of years—as medicine.