meds for OA, including placebo

the lancet just published a network meta-analysis of NSAIDs, acetaminophen, or placebo for hip or knee osteoarthritis (OA) pain (see osteoarthritis lack of acetaminop benefit lancet2016 in dropbox, ordoi.org/10.1016/ S0140-6736(16)30002-2​). a network meta-analysis is a mathematical device which allows one to infer the comparative effectiveness of different interventions from different randomized clinical trials, even though there may not have been actual trials with direct comparisons. details:

--74 RCTs were included in the analysis with a total of 58,556 patients with predominantly knee or hip OA. all trials had at least 100 patients per group and had prespecified primary and secondary outcomes of pain and physical function.

--they developed a network comparison chart with 23 nodes, each with a treatment (several nodes had the same drug but in different doses), and the relationships between them.

--results:

    --when looking at what was considered a clinically significant medication pain response vs placebo, the following were significant  (anything <-0.35 is considered clinically significant):

        ​--of the 7 the most clinically effective, rofecoxib and etoricoxib at varying doses represented 6 of them (neither available in the US)

        --of the ones available, diclofenac 150mg was the best, with effect size of -0.57 (the highest was rofecoxib with effect size of -0.62). none of the lower doses of diclofenac were effective

        --naproxen 1000 mg/d and ibuprofen 2400 mg/d trended to being clinically significant, but did not make the cut

        --lower doses of naproxen and ibuprofen weren't even close

        --acetaminophen was much worse, with the dosage of 3900-4000mg having a clinically-nonsignificant effect size of -0.17, and lower doses being no different from placebo

        --celecoxib at 100mg was not effective, but at higher doses came close to having a clinical effect

    --when looking at physical function, the results were much worse:

        --only rofecoxib and diclofenac unequivocally reached clinical significance.

        --naproxen 1000 mg/d and ibuprofen 1200 or 2400 mg/d trended to a benefit

        --acetaminophen wasn't even close

    ​--of importance: several of these are COX-2 inhibitors -- rofecoxib (vioxx) was pulled from the US market because of increased heart attacks and strokes; etoricoxib has not been approved and has associated severe skin reactions; and lumiracoxib, which is structurally different from the other COX-2 inhibitors (and more similar to diclofenac) causes liver failure and was never approved​. celocoxib is the only of the COX-2 inhibitors available in the US, though there is real concern of increased cardiovascular problems, esp at the higher doses (which were the ones that worked)

    --several studies looking at adverse cardiovascular effects of NSAIDs have found that diclofenac (which i've read in the past has a higher COX-2/COX-1 inhibition, though not sure this is true) is the worst. (naproxen is usually found to be the best, high dose ibuprofen is not so good. see, for example, Prescrire Int. 2016 25 (167): 14; or PLoS Med 2011; 8(9): e1001098)

so, what does this all mean??

--as mentioned in several blogs, there is typically a significant placebo effect of pain meds, often on the order of 20-30% (though one interesting study found, not surprisingly, that people who had positive expectation of treatment had a much higher placebo response than those with negative expectations (see placebo treatment expectations scitransmed2011 in dropbox, orScience Translational Medcine 2011; 70(3):70ra14). and, in this study, subjects  exposed to experimental heat pain​ but had a negative expectancy​ did have some benefit from an infusion of a m-opioid agonist remifentanil; but, when they thought the infusion had stopped, they had a full restoration of pain intensity.

--so, though the network meta-analysis showed not much effect of acetaminophen, or even lower dose ibuprofen of naproxen, if the patient expects some benefit, they may well respond (which means: to a large extent this benefit may depend on how well/enthusiastically we can present the options -- ie, we may be able to enhance the effectiveness of an intervention by providing a very positive vibe).

--and the many adverse effects of high dose diclofenac (used a lot in Europe), naproxen or ibuprofen, which have a higher likelihood of benefit in RCTs, may be avoided by using even low-dose acetaminophen

--i still use a lot of non-systemic treatments first. Exercise/PT is a mainstay. topicals often work (eg topical capsaicin, lidocaine, or topical dicofenac). injections often work well (though i leave intra-articular hip injections to specialists who can use ultrasound-directed injections), and they work well at many different sites (A-C joint, hand joints, etc).

--and, in the final analysis, i do sometimes use lower-potency opiates (eg tramadol, codeine), even in elderly people who either experience or at higher risk of experiencing adverse effects of long-term NSAIDs (eg heart failure, bad hypertension, higher risk of GI bleed, renal insufficiency) and who are not at risk of abusing or diverting them, or having them taken by a younger family member. unfortunately, as we age we tend to have the combination of more pain and more adverse effects of meds. and i realize that, as clinicians, we all have different prescribing thresholds, but many of my older patients are barely able to get around/function because of pain, and they really do tolerate and function well on low doses of less potent opioids (i also realize this is a tad heretical in the current anti-opioid atmosphere...)