Sunday, July 7, 2013

The agony & the ecstasy of EBM in symptom management

So, I decided I might blog a little again. Probably the occasional Journal Club of the Cloud-type posts. Christian and fellow bloggers, thank you for all you do in keeping Pallimed thriving and relevant.

So, Eduardo Bruera & colleagues at MD Anderson have published the results of their long-awaited follow-up trial to their 2006 double-blinded, placebo controlled trial suggesting that methylphenidate (MP) is no better than placebo for cancer-related fatigue (CRF).

In the 2006 trial, most patients had a marked improvement in fatigue across the 2 weeks of the trial. This lead the investigators to hypothesize that perhaps it was the frequent research nurse calls, that both the active and placebo groups received, that caused the fatigue to improve.

Thus, the 2013 trial is really a 2x2 trial comparing both methylphenidate and placebo, as well as a nursing telephone intervention vs a control telephone intervention (ie there were 4 groups: MP + nursing call, MP + control calls, placebo + nursing call , placebo+control calls). The investigators and patients were blinded to allocation. Notably, the main drug intervention was giving patients 5mg MP tabs (or matching placebo), and encouraging them to use it q2h prn, up to 4 doses a day, for fatigue. This is the identical protocol to the 2006 trial. The patients all had cancer and were recruited from the MD Anderson clinics (median age 58 years, 67% woman, 72% white, baseline fatigue was self-reported as 6-7/10). As far as I can tell there was no restrictions to enrollment based on stage of cancer or chemotherapy history. 

The nursing calls involved both symptom assessment, plus 
During each call, the research nurse asked open-ended questions regarding general well-being of the patient and family. The research nurse listened empathetically, answered the patient's questions, provided supportive statements, and then ended the telephone call.
The control calls were just the administration of a symptom assessment instrument and questions about medication use - any concerns the patient expressed were directed towards their physician.

The primary outcome was improvement on the FACIT-F fatigue subscale at 14 days; as usual many secondary outcomes were evaluated. They estimated a sample size of 212 to detect a 33% difference between MP and placebo outcomes (190 were actually allocated; 151 were evaluable).

Basically, the results were a wash. MP was no better than placebo, therapeutic nursing calls were no better than the control calls for fatigue. All 4 groups' fatigue improved, basically by 2 points on a 0-10 scale. Several non-fatigue symptoms significantly improved in the nursing call group that did not improve in the control call group (nausea, anxiety, drowsiness, appetite, sleep, and 'feeling of well-being'). Notably, the median number of MP capsules used was 18 over 14 days (=6.2 mg/day) and the maximum used was 36 (=13mg/day). To be clear, while the trial is under-powered somewhat, there really is not even a trend towards improvement with MP. As usual, MP was well-tolerated and safe.

So, what to do with this? First, there have been a couple recent studies on MP for CRF that are relevant. First is a RCT of controlled-release MP (pushed to 54 mg daily) for cancer patients which did not show any benefit overall. However, for the subset of patients with advanced cancer (defined here as stage III or IV), MP did improve fatigue vs placebo (statistically significant; probably clinically significant). Next is a RCT of dexmethylphenidate ("Focalin," which I must remark, reminds me so much of this) in150 patients with fatigue after cytotoxic chemotherapy (unclear to me how many of these patients had advanced cancer; one's sense after reading the study is that probably most of them had neo/adjuvant chemotherapy but I don't know that for sure). These patients were pushed to ~25mg of dexmethylphenidate (presumably equal to 50mg of racemic MP?) for 8 weeks and were found to have improved fatigue compared to placebo.

I guess, despite this blog post's title, there's really not much ecstasy when it comes to EBM in symptom management. We have 3 trials, all using different drug protocols, all looking at at overlapping-but-different patient populations, with conflicting results. This is why I employ the cardinal rule that one should never even consider making a practice change based on a single research paper, without actually reading the entire paper. Reading Bruera's abstracts, one could conclude that MP is worthless. Digging deeper, those studies are using MP in ways different than I do in my actual practice - in particular at substantially lower doses than I generally employ. The dex-MP study seems promising, but I can't even figure out from reading it if those patients resemble most of mine (I see some patients with curable cancers receiving adjuvant chemotherapy, but most of my patients have advanced cancers). And the controlled release MP study makes me tempted to break cardinal EBM rule #2: one should, in general, treat secondary outcomes as hypothesis-generating more than clinically relevant.

My own experience is that most patients, ~2/3, don't respond to MP. They stop it after a few weeks, after they've let me push it to 20-30 mg daily.  ~1/3 however seem to respond, evaluate their quality of life as better on it, and we continue it. Maybe 10% it's a wonder-drug: patients feel markedly improved on it and it makes a huge impact on their quality of life. These numbers are gross estimates-I have not systemically evaluated my own outcomes.

Which has got me thinking, what data would I need in order to stop using MP? Which is another way of asking the question what is the purpose of EBM, clinical trials, in symptom management, where everything is always an N=1 trial.  Compare for instance statins for the secondary prevention of cardiovascular disease and, say, amitriptyline to treat post-herpetic neuralgia. We give statins because there are large, RCT telling us that if we do we'll prevent a few MIs over the years in a few patients. We don't know for any one patient if giving them a statin will actually do anything. We can measure their LDL, but that's really of secondary importance to actually preventing them from having another heart attack, or stroke, etc. In fact, we have no real way of measuring efficacy of statin therapy in any individual patient. If they never have another MI or stroke, good, but maybe they wouldn't have anyway, or if they do have one, maybe it's two years later than it would have been without the statin, etc. The point is, the only way we can measure the efficacy of that statin is across a population of people. Which is why we have large, well-designed RCT. This is why for a decade we were all going for very tight glycemic control for diabetics - out of the belief that across a population of diabetics we would be reducing sufficient number of MIs, strokes, kidney failure, retinopathy, etc to be worth all the extra patient hassle, expense, and morbidity (even death) associated with aiming for tight control (I personally think it remains unclear if we've actually done our diabetics a favor by doing this).

Amitriptyline for PHN: you give it to the patient, ask them in a few days if it's working or not, maybe a week, then increase the dose, reassess in a few days or a week, then maybe readjust up again or abandon amitriptyline completely for another agent.  Ie, with symptom management we basically have immediate & patient-relevant data on efficacy.

So what, then, is the point of EBM and doing these studies? I think it's several-fold. One is to give us some guidance as to what's more likely to work (what we should go to first); eg, SNRIs for chemotherapy related neuropathy. The fact that SNRIs have been the only agents ever to be shown to be effective for the pain from chemo neuropathy has not stopped me from trying other things, it's just that now I go to the SNRIs first. Two is to give us a sense of safety, side effects, toxicities - incredibly important. Three is to give us a sense of what probably works via placebo mechanisms vs other mechanisms. The bind I get into is with this one - with interventions that I 'witness' working, sometimes dramatically, albeit on occasion, that all the evidence points to works via placebo mechanisms. AB(H)R gel, infamously. Lidocaine 5% patches for everything except painful syndromes involving damaged peripheral axons such as diabetic neuropathy, post-herpetic neuralgia. Ketamine, after the Currow trial, which to me was more damning due to evidence of harm from ketamine than lack of efficacy (still waiting, for blinded, controlled depression data for ketamine)? Kyphoplasty?

Personally, I don't include methylphenidate on this list - I think the data, as a whole, support the idea that it is helpful, probably at higher doses (ie higher than 15mg a day) for some patients with severe disease or fatigue.

Please leave your thoughts, particularly about the use of interventions that, in your heart and EBM-brain, you think actually work entirely or mostly via placebo mechanisms.