 This chart appeared in the 2017 President's Commission on Combating Drug Addiction and the Opioid Crisis to demonstrate a link between legal opioids like OxyContin and overdoses. One line shows opioid prescriptions, another the number of people being admitted to treatment centers for opioid abuse, and the third charts deaths from opioid overdoses. It demonstrates absolutely nothing. By using a chart like this as evidence that the increase in opioid prescriptions caused the overdose crisis, government officials committed the most rudimentary statistical error of all, which is to assume that correlation proves causation. Global avocado production also climbed in these years. My age also went up. Does that mean avocado production of my age might have something to do with the increase in overdose deaths? No. While it's at least plausible that an increase in opioid prescriptions led to an increase in overdoses, you can't determine that simply by noting parallel increases. The purpose of statistical analysis is to test our theories and assumptions. And this chart fails miserably in that regard. The way it's drawn in the commission's published report also gives the misimpression that the rates of increase for these three lines were about the same. Actually, all three lines are measured in different units, and the chart should look more like this. Another red flag. This chart doesn't go past 2010, even though the report came out in 2017. There's a reason for that. In the years that followed, the pattern rapidly reversed, with opioid prescriptions falling dramatically while drug-related deaths continued to increase. The use of charts like this one, which is part of the government's failure to correctly understand and respond to the opioid overdose crisis, has caused enormous human suffering. The government has pushed doctors to severely limit opioid prescriptions for pain patients through laws, regulations, and the threat of civil and criminal penalties. And today, thousands of Americans with chronic and debilitating illnesses are unable to access drugs that could alleviate their pain. This chart from the CDC gives us a clearer picture of events. Drug-related deaths involving commonly prescribed opioids like hydrocodone and oxycodone did rise in the first decade of the 21st century, but then leveled off. For the first 10 years, the increase in overdoses correlated with increases in legal prescriptions. But those prescriptions fell after 2010, with no reduction in overdoses. What we did see in 2010 was an increase in illegal heroin use, and in 2013, an even sharper rise in overdoses from fentanyl and other synthetic opioids rarely used in legal medicine. This was likely caused by a change in the business practices of drug cartels, and the fact that fentanyl is cheaper and easier to smuggle than heroin. During this period, fentanyl, which is more potent than heroin, became increasingly common as a heroin booster and substitute. If heroin were legal, most of these overdose deaths probably would not have occurred. The governments clamped down on prescription opioids also probably exacerbated the overdose crisis. The blockbuster pain drug oxycontin became available in 1996. Three years later, we do indeed see a rise in deaths involving prescription opioids, but that does not necessarily mean those deaths involve people who obtain prescriptions from doctors. National survey data from 2013 and 2014 indicated that most people who misused prescription pain relievers reported getting them from a friend or relative. Then a quarter said they obtained the drugs from doctors. A survey of oxycontin-addicted people entering rehab from 2001 to 2004 found that 78% had never been prescribed the medication, and among the 22% who had been prescribed oxycontin nearly all had substance abuse problems before taking the drug. So if they weren't taking oxycontin, they likely would have been taking something else. The stereotype of a clean living person getting hooked on opioids originally prescribed for pain and then suffering a fatal or non-fatal overdose is not common in the data. As prescription opioids became harder to find on the black market, many non-medical users of these drugs likely turned to heroin contaminated with fentanyl, putting them at a much higher risk of overdosing. They were placed legally produced, reliably dosed pharmaceuticals with black market drugs whose potency is highly variable and unpredictable. If we widen the time span, we see that overall drug overdose deaths have been increasing at a fairly steady rate since 1979, or 17 years before oxycontin became available, and continuing along after legal use of opioids was severely cut back. During the same period, we see parallel increases in deaths from suicide and alcohol abuse, or so-called deaths of despair. There have also been steady increases in diagnoses of depression. This suggests that many deaths ruled accidental overdoses could be deliberate suicides, or perhaps they were partially deliberate, although this is correlation, not causation, so we don't know for sure. The data also suggests that the underlying cause of the overdose crisis was more fundamental than overprescribing of opioids. Back around 2010, when opioid prescriptions were still rising alongside overdoses, it was plausible to investigate a link, but none of them took the correct approach, which would be to compare similar patients, one of whom got a prescription for opioids, and the other who did not. Such match pairs shouldn't just be checked for overdose deaths, which after all affect only 0.04% of patients, but for overall health and quality of life. Instead, researchers compared unmatched samples of patients, some of whom received higher doses of opioids and others. One influential and heavily cited study published in 2011 in the Journal of the American Medical Association began with a sample of 750 veterans health patients who received opioid prescriptions for pain and later died of opioid overdoses. It compared them to a random sample of 155,000 other veterans health patients who also received opioid prescriptions and didn't die of overdoses. It found that the high dose opioid prescriptions were more likely to lead to fatalities, a finding that was used to support the CDC guidelines for strongly discouraging large doses. The problem with this study is that the high dose and low dose patients differed in many ways. In particular, the high dose patients presumably had more severe pain and more serious medical issues. They may have also differed in the type of pain they endured or the length of time they suffered. Controls have to match subjects as closely as possible. This study did not meet that basic requirement of good science. It is plausible that the 750 patients who overdosed did so for reasons other than the amount of opioids they were prescribed. Reasons the study didn't control for. 40% had been diagnosed with substance abuse disorders in the 12 months before getting an opioid prescription. 66% had been diagnosed with a psychiatric disorder. A population with these characteristics is likely to have a higher than normal frequency of overdose deaths with or without legal opioid prescriptions. Perhaps most important, this study ignores the benefits of opioid pain relief to the 155,000 patients who didn't overdose. Even if taking prescription opioids does increase the risk of overdose deaths, which this study doesn't demonstrate, many patients would gladly accept the risk for the quality of life improvement of diminished pain. An influential 2010 study, published in the Annals of Internal Medicine, also attempted to show that higher prescribed doses of opioids correlated with a higher risk of death from overdose. But within the study group there were just six patients who died of opioid overdoses, fewer than the number of study authors. Yet the authors claimed statistically significant results after adjusting for smoking, depression, substance abuse, comorbid conditions, pain sight, age, sex, recent sedative, hypnotic prescription, and recent initiation of opioid use. This is absurd. It would take at least 100 patients who died to estimate the adjustments, much less make any conclusions with statistical confidence. The authors even conceded that it was not established whether their results were due to patient differences or direct effect of higher dose. In other words, the study demonstrates nothing. Although they are more than a decade old, these studies continue to be cited by public health officials and policymakers, including the 2017 President's Commission on Combating Drug Addiction and the Opioid Crisis, which relied almost completely on studies published before 2014, which generally used data collected before 2011. Again, there's no mystery as to why. Only by relying on the outdated studies could the prescription opioids cause the crisis narrative be sustained. Government officials haven't reckoned with the abysmal failure of limiting access to prescription opioids. Rather, they've doubled down on the policy while citing and promoting statistically flawed studies based on stale data and promoting charts that conspicuously cut off before the data start to undermine their chosen policy. No one has yet demonstrated that patients prescribed opioids are more likely to suffer overdoses than match patients who were denied them. No one has shown that overall quality of life and life expectancy are lower for opioid patients. The first step to arriving at a serious policy response to the overdose crisis is to take an honest look at the problem by using statistical techniques to gain a deeper understanding of its root causes. Instead, government officials want us to look at three lines going up over a set time period, now long in the past, and assume that they must have something to do with each other.