meds for diabetic neuropathy

The Annals of Internal Medicine had a recent meta-analysis of medications for painful diabetic neuropathy (see dm neurop meds annals 2014 in dropbox, or doi:10.7326/M14-0511). as most of us know, this is a very common clinical issue (30-50% of diabetics get it, predominantly distal symmetric sensorimotor polyneuropathy). The first approach is to optimize glycemic control (data are clear that nerve conduction improves, less clear that symptoms improve), but patients typically need meds to control the burning, painful discomfort. many classes of agents have been used, including tricyclics (TCAs, most often studied is amitriptyline), anticonvulsants (gabapentin, pregabalin, carbamazepine), serotonic-norepinephrine reuptake inhibitors (SNRIs, eg duloxetine or venlafaxine), opioids, opioid-like substances, and topicals (eg capsaicin cream). results:

--65 RCTs analyzed, with 12632 patients and 27 pharmacologic interventions. though 1/2 of studies with high or unclear risk of bias. network meta-analysis done (which combines effect sizes for all possible pairwise comparisons)
--9 head-to-head trials showed greater pain relief with  SNRIs than anticonvulsants; also greater improvement with TCAs than with topical capsaicin 0.075%
--SNRIs, capsaicin, TCAs and anticonvulsants were better than placebo for short-term pain control; in assessing individual drugs, carbamazepine, venlafaxine, duloxetine, and amitriptyline are better than placebo
--adverse effects common: somnolence and dizziness with TCAs, SNRIs, and anticonvulsants; xerostomia with TCAs, and peripheral edema and burning sensation with pregabalin and capsaicin
--of note, when looking at pain score reductions of drugs vs placebo, from highest to lowest, were SNRIs, topical capsaicin, anticonvulants (though notably carbamazepine was the best), TCAs, and last (by a fair margin) opioids. 

it is hard with these types of complex analyses to be sure about the relative effectiveness of meds, since head-to-head trials were limited.   My personal bias is to use TCAs as first-line, with really very impressive results in the vast majority of patients, even long-term. because of the really common adverse effects of amitriptyline, especially in the elderly (which is the majority of patients with diabetic neuropathy), i have been using desipramine (in the 25-75mg range) and nortriptyline (in the 10-50mg range) pretty much exclusively, with desipramine having the fewest adverse reactions of any TCA and nortriptyline being useful in those who have trouble sleeping. the reason i chose these is from a New England J Med article in 1992 -- dating myself again -- (see dm neurop tca nejm 1992​ in dropbox, or  NEJM 1992; 326:1250-6), which found that desipramine was as good as amitriptyline, and fluoxetine actually was as good but only in those with depression (see figure below from that article). i mentioned this article to a really phenomenal neurologist i knew then, and he said that he almost exclusively used desipramine for all neurologic indications where amitriptyline had been used in studies (eg migraine)​ and had great results. My guess is that desipramine may even be better than amitriptyline since many patients have adverse effects with amitriptyline and drop out of treatment. My experience with  gabapentin and duloxetine are limited, but in the few instances i've used them i have had few patients who tolerate the adverse effects.  i've had very good results with carbamazepine or oxcarbazepine, but only with trigeminal neuralgia (I have not used with diabetic neuropathy, but they may well be useful, from the above study).