Not finding something doesn’t mean it doesn’t exist

A favorite citation used by the marijuana lobby to promote the idea that marijuana doesn’t impair driving is one published in 2015 by the National Highway Traffic Safety Administration (www.nhtsa.gov/behavioral-research/drug-and-alcohol-crash-risk-study).  And when the study is cited, the results are usually badly misquoted. 

 

Los Angeles Times: “Good news (?): marijuana doesn’t increase the likelihood of car crashes.”

USA Today: “New study shows no link between marijuana use and car accidents.”

Reason: “Landmark study finds marijuana is not linked to car crashes.”

Nonsense.  It’s true the study failed to find a statistically significant link between marijuana use and car crashes.  But failure to find a link is not the same as finding there is no link. Particularly in a study not designed to find that link in the first place.  The study also failed to find a link between crash risk and heroin or methamphetamine, both much more impairing than marijuana’s THC.

It’s like your failure to find your lost keys is not proof that they don’t exist.

The NHTSA study has been widely and properly commended for its excellent design of controls.  But the study itself was badly designed.  Here are the primary problems with it:

  1. The sample size was too small to find a statistically significant link for any drug other than alcohol.  This is because of both the baseline prevalence for use of the drug and the modest Odds Ratio for that drug causing crashes.
  2. All study subjects were not included in the study – only those who volunteered to be included.  It’s not clear why someone who knew he/she was impaired would volunteer to be studied, but some did.
  3. At least 413 of the 3,095 subjects were innocent victims who were involved in the crash, rather than being the crash initiator.  One might expect that such subjects would have a drug prevalence similar to controls. Including them in the study pool weakened the signal provided by the subject pool.
  4. NHTSA has published strong data showing the Odds Ratio for causing a non-fatal crash is less than that of causing a fatal crash.  Yet of the 2,682 crashes studied, there were only 15 fatal crashes.  The study has been dismissed by some as merely studying “fender benders.”
  5. The study site Virginia Beach is atypical.  It’s a military town with a 14.4% drug prevalence compared with a 19-22% drug prevalence nationally according to studies done at that time.

The study concluded,”This study should not be interpreted to mean that it is safe for individuals who have used substances to operate a vehicle.”  Yet that is exactly what the marijuana lobby and many in the media have tried to promote.

 

Categories

DUID Articles