Painful Diabetic Neuropathy: How the AAN Graded the Evidence

If you are looking for the one best drug for the burning, tingling foot pain of diabetic nerve damage, the 2022 American Academy of Neurology (AAN) guideline has an answer that sounds like a non-answer: there is no single best drug, and that is exactly the point. Tricyclic antidepressants, serotonin-norepinephrine reuptake inhibitors (SNRIs), gabapentinoids, and sodium-channel blockers all lower pain by a similar amount. The guideline (Price and colleagues, Neurology, 2022) tells clinicians to offer any one of them, to switch to a different class when the first one does not work, and to avoid opioids. Read carefully, it is a short lesson in what to do when several treatments tie.

Key points#

• Four oral drug classes ease painful diabetic neuropathy, and their measured effects are close enough to be considered comparable.
• Tricyclics posted the largest number on paper, but the AAN rated that estimate as low confidence, so it did not become the front-runner.
• Because efficacy is similar, side effects, other conditions, drug interactions, cost, and patient preference decide the pick.
• When one drug fails, the guideline advises trying a different class rather than a cousin of the same drug.
• Opioids are not recommended, and the head-to-head OPTION-DM trial backed the guideline's tie by showing that combining two drugs helps when one alone falls short.

What "works about equally" actually means#

To compare drugs tested on different pain scales, the AAN converted each trial's result into a standardized mean difference (SMD), which puts every scale onto one common ruler. Think of it as translating grades from schools that use letters, percentages, and points out of ten into a single shared scale so you can line them up fairly.

On that ruler, gabapentinoids came in near 0.44, SNRIs near 0.47, and sodium-channel blockers near 0.56. Those numbers sit close together and their confidence intervals overlap, which is the statistical way of saying the differences between them are within the range of ordinary noise. When intervals overlap like that, claiming one class is truly better than another is reading a signal that the data does not support.

Why the biggest number did not win#

Tricyclics looked like the standout, with a point estimate around 0.95, roughly double the others. Yet the guideline did not name them first line. It flagged low confidence in that figure, because the tricyclic trials were older, smaller, and more varied in how they were run. A big effect measured loosely is not the same as a moderate effect measured well.

This is the heart of evidence grading, and it is worth slowing down on. Grading systems deliberately keep two things apart: how large an effect appears, and how sure we are that the effect is real and repeatable. A precise medium result can earn a firmer recommendation than an eye-catching large result built on shaky trials. Collapsing both into a single ranking, and simply crowning the biggest number, is the mistake the AAN's method is designed to prevent.

What the guideline asks clinicians to do#

From that reading, the practical advice falls into place. Clinicians should explain to patients that the benefit is genuine but partial, then offer a trial of a tricyclic, an SNRI, a gabapentinoid, or a sodium-channel blocker. Each of those carries a Level B recommendation, the AAN's phrasing for a step that should generally be taken given the evidence behind it.

Two further instructions shape the guideline. First, because the classes work about equally, the decision hinges on everything other than raw potency: the side-effect profile, the person's other medical conditions, interactions with their current medications, cost, and what they are willing to live with day to day. Second, if a drug fails or the side effects are intolerable, the next move is a drug from a different class, not another member of the same one. Switching the underlying mechanism is more likely to help than trading one gabapentinoid for its near-twin.

The advice on opioids also sits at Level B, but it points the other way. The guideline concludes that opioids should not be used here, because the evidence for lasting benefit is weak while the harms are well documented. Grading can rule a category out just as clearly as it can support one.

OPTION-DM: putting the tie to the test#

A guideline can only pool the trials that exist; it cannot run comparisons no one has done. Direct head-to-head data on these pathways was thin, and that is the gap the OPTION-DM trial set out to close. Published in The Lancet in 2022 (Tesfaye and colleagues), it was a double-blind, randomized crossover study across 13 UK centers that pitted three sequences against each other: amitriptyline then pregabalin, pregabalin then amitriptyline, and duloxetine then pregabalin. In plain terms, a tricyclic-first, a gabapentinoid-first, and an SNRI-first strategy.

The pain relief was nearly identical. Average scores dropped from about 6.6 at baseline to roughly 3.3 by week 16, with no meaningful difference among the three pathways on the primary outcome. What separated them was tolerability, not results: dry mouth showed up more with amitriptyline, dizziness with pregabalin, and nausea with duloxetine. That is the guideline's logic confirmed in a trial. If the pain relief is a tie, then side effects and personal factors are the fair way to choose.

OPTION-DM also supplied something the guideline could only hint at. For people whose pain stayed high on a single drug, adding a second agent produced a further, worthwhile drop in pain compared with staying on one. The combination was reasonably tolerated. That is directly useful for a common bind in clinic, when the first drug helps a little but not enough, and it fits neatly into the stepwise approach.

A better way to read a tie#

Faced with a table of drugs, the reflex is to rank them and move on. The AAN guideline and OPTION-DM both argue against that reflex. When several options land within the noise of one another, the useful questions change shape. Which side effects can you tolerate? What else are you taking? How confident are we in each estimate, and what did the trials actually measure? A finding of comparable efficacy is not a failure to find a winner. It widens the set of reasonable first choices and hands the real decision to the individual sitting in the room.

It also protects against a subtle trap: betting on the largest number simply because it is largest. The tricyclic result is the cautionary example here, impressive on the page yet marked low confidence. Grading exists so that magnitude and certainty are weighed apart rather than fused into a single verdict. Both the guideline and OPTION-DM sit with that uncertainty honestly, and both arrive at the same disciplined place.