Two new reviews of cannabis for neuropathic pain
Today I stumbled across two new scientific reviews of the same thing — cannabis for neuropathic pain — with different conclusions about more or less the same data. I thought I’d have some good nerdy fun comparing and contrasting them.
If nerdy fun is a non-starter for you, here’s the zero-nuance bottom line: surprise surprise, cannabis does not work well for neuropathic pain, if at all. There’s about seven asterisks on that, but that’s the reasonable broad strokes conclusion. Discouraging evidence never stops anyone from shouting “it works for me,” of course. Mostly not about anything, but cannabis is in another evidence-denial league. Maybe it works for you and maybe it doesn’t, but I can guarantee — based on this evidence — that it doesn’t work very well for very many.
(And no, there probably are no juicy outliers missed by the studies. A good benefit for a lucky few mostly cannot “hide” in weak average results. If it were good, it would boost that average.)
Perhaps the most interesting and controversial plant in the world.
REVIEW #1: Ateş et al was a meta-analysis from the Cochrane Collaboration — which has a reputation for high standards and getting this kind of thing at least less wrong. They delivered a low to very low certainty conclusion: barely a conclusion at all! Really just the classic “garbage in, garbage out” pattern for so many meta-analyses of pain treatments. What the data (barely) shows is a minor bump in some softer secondary outcomes (and only for a THC/CBD combo, not for either THC or CBD on their own), which the reviewers themselves flagged as not clinically relevant. Plus an entirely predictable increase in adverse events. This is how they frame their bottom line:
“There is no clear evidence for an effect of [cannabis] medicines on pain relief of 50% or greater …”
REVIEW #2: Mitchell et al is another meta-analysis of mostly the same studies, with another barely-positive result with a much more positive-sounding interpretation:
Cannabis-based therapies offer modest analgesic benefit and should be reserved as an adjunctive option for select patients with refractory symptoms, guided by a shared decision-making process that considers individual risks and preferences.
But “modest analgesic benefit” isn’t justified by their data … and “should be reserved” is just editorializing loosely based on that unjustified conclusion. We can recommend practically anything that isn’t blatantly a poison as “an adjunctive option” for patients with tough cases; “it’s worth a shot” doesn’t even really require good evidence of anything other than a lack of harm.
Of course, there is evidence of harm in this case. From both studies. THC has never been harmless, and opinions to the contrary have never really been anything but propaganda. (See my full cannabis article for more information.)
So were Mitchell et al. just making lemonade out of lemons?
Not entirely. There is a legit case for their more positive conclusion, even if it’s still unimpressive. Standard mean differences (SMDs) are arguably the more meaningful measurement than the Ateş headline outcome that focused on a threshold (not counting any pain relief that wasn’t at least 50%); they focused on an interpretation that was more sensitive to small effects.
But sensitivity to small effects is only valuable if small effects are big enough, and SMD −0.30 in pain research mostly isn’t.
And what’s more damning is that the “modest analgesic benefit” disappears with another analysis (change-from-baseline). Mitchell’s headline conclusion leans on one of two pre-registered analyses, the one that produced a positive result, with no reason given for preferring it — so we’re left to assume the obvious, namely that the authors just liked one answer better than the other. When two legit interpretations disagree, “modest analgesic benefit” is not a fair summary unless it also acknowledges “but it depends on how you squint.”
Meanwhile, the harms were the same no matter how they crunched the numbers. The downsides of cannabis are confirmed more robustly in both reviews. So that’s not a great look.
This is what we get from the mundane optimism bias that pervades pain research. We are constantly hearing the best news that can possibly be justified by the data, or tortured out of it, and even that is unimpressive. The cynical salamander’s rule of thumb: supposedly “promising” research almost never actually is. And in this case we’ve got a second review for the reality check, instead of just my skepticism.