Pulse has a recent article on how a GP surgery tackled the problem of frequent attenders with no obvious pathology. Such patients were offered a referral to the local MIND centre and there seems to have been benefit for those who accepted:

“…Nine patients out of the 20 we wrote to attended the special one-to-one appointments. The meetings were felt to be useful by both parties and the historical need to attend the surgery was discussed. Resolutions involved a range of ideas including volunteer work, subsidised reflexology, anxiety groups and subsidised counselling.

The main attraction of this project was the possibility of a dramatic reduction in our workload but the implications of somatisers for secondary care are significant and raised some commissioning possibilities.”

It’s possible the ‘subsidised reflexology’ was suggested/provided by MIND, rather than GPs, and it’s probably mean-spirited of me to wonder if it is really necessary to include a modality which makes claims about human physiology and disease which have been found to have little basis in reality.

Beyond that however, although there can be one-off good placebo responses in controlled settings like the experiment – I wonder if, paradoxically, they are the people who might rely on ‘luck’ to tide them through and might not be the people who would reliably take medication – as per diabetes, asthma, epilepsy.

Intriguing research area that must be getting much harder to explore in a ‘real world setting’ because Helsinki is (rightly) stacked against placebo where an appropriate best-practice treatment exists.

“…Recent functional brain imaging by University of Michigan researchers, headed by neuroscientist Jon-Kar Zubieta, has confirmed this finding. Their model is straightforward. Healthy young volunteers agreed to receive a painful injection into a jaw muscle. Prior to the injection, the volunteers were asked to gauge how much pain reduction they would experience if they were also administered a painkiller. Would they feel a 20 percent reduction in pain? Fifty percent? Then, simultaneously, they received the painful jaw muscle injection and a painkiller injection — only the “painkiller” was a placebo — an inert saline solution.

In patients reporting pain relief, PET scans showed an activation of their opioid receptors, where both narcotics and endorphins ply their trade, and a decrease in activity in several brain regions vital for feeling pain. The scans also showed that the placebo effect isn’t confined to endorphin release; the neurotransmitter dopamine receptor sites critical for the brain’s primary reward system also lighted up.

In an elegant follow-up study, Zubieta and colleagues looked at how subject expectation of pain relief was associated with a more generalized expectation of good results. The volunteers were invited to play a game; if they won, they’d get a small monetary reward. Before playing the game, they were asked to estimate their chances of winning. The study showed that those who felt they were most likely to win were the same folks who predicted the greatest relief from their “painkiller” injection. Scans confirmed that the activation of brain reward regions correlated with both general anticipation of winning and degree of expectation of pain relief…”

I wonder how this interacts with the risk perception/compliance questions discussed above.

the study of how culture influences our understanding of health, illness and medicine.

While we tend to think of illnesses as specific encapsualted ‘things’ that happen to the body, it turns out that our culture and psychology has a huge influence on not just what we think of illness, but how we actually become ill.

Culture also shapes what we think of as ‘healthy’ and ‘unhealthy’, ‘normal’ and ‘abnormal’ and this is one of the main driving forces behind how we express physical or psychological distress and expect it to be treated.

As if that didn’t offer enough placebo-related goodness – they have a fascinating link to Vasectomania, and Other Cures for Sloth with some novel and fad placebos from early C20.

Several authors have commented that randomized controlled trials in duodenal ulcer have given substantially different placebo healing rates [7–9]. It has not been postulated that in duodenal ulcers the frequency of placebo administration might influence the healing process. By means of a systematic review we aimed to examine the influence of frequency of placebo administration on duodenal ulcer healing. Therefore we calculated the pooled ulcer healing rate of the placebo arms of all randomized clinical trials with a four times a day regimen and compared this with the pooled placebo healing rate in trials with a twice a day regimen…
Relevant data for the evaluation were extracted from text, tables, and figures of the publications. In many trials the ulcer healing rate was computed by dividing the number of patients with a healed ulcer by the total number of evaluable and compliant patients. Our analyses are based on these reported healing rates. Differences in healing rates between placebo regimens were assessed by equal, fixed, and random effects models [10]…
In the four times a day regimen, 805 of 1821 patients (44.2%) were healed after 4 weeks of placebo treatment, while in the group that took a placebo twice a day 545 of 1504 patients (36.2%) were healed (difference, 8.0% [equal effects model]; 95% CI 4.6%, 11.3%).

So, it looks like there was a higher rate of placebo response among people who were compliant enough to take 4 tablets a day than those who were willing to take 2 per day: your Girl Scouts v. your muddle-through+s as it were.

However, the authors do remark:

Third, in numerous trials the statistical analysis was based on the number of evaluable patients, while some trials also excluded patients from the analysis who did not comply with the treatment regimen. In theory, patients on a four times a day regimen are more likely to be noncompliant. So analysing healing rates of compliant and evaluable patients only, could have resulted in biased healing rate estimates. However, many reports did not specify the number of excluded patients because of noncompliance. Therefore we were not able to examine the impact of noncompliance on the healing rate. Using the total number of randomized patients in the denominator instead of the number of evaluable and compliant patients did not change the healing rate difference between the two placebo regimens.
Fourth, it would be possible that the physician who decides to include a patient in a trial, makes an assessment whether an individual patient will comply with a certain drug regimen. Therefore the group of patients in the trials with a four times a day regimen might be different from the patients in the twice a day trials. In a randomized trial to evaluate the efficacy of lipid-lowering drugs in the therapy of coronary heart disease, patients in the placebo arm of the trial who complied with the prescribed treatment regimen (took at least 80% of placebo capsules) had a lower 5-year mortality compared with the noncompliers [14].

Which looks like they think that compliance may be an interesting confounding factor. But, it is almost impossible to track the figures because these trials happened in the 70s and 80s and it doesn’t look like they were at all hot on establishing NNTs or using the intention to treat analysis what with NNT not being popularised until 1988.

Of course, with Helsinki and everything – it’s probably very tricky to come up with a study design involving placebo for a significant illness that would pass an IRB nowadays. By and large, lots of people (rightly) frown on using a placebo in such trials, because you are expected to evaluate your proposed novel treatment against the current best practice treatment.

Which is why there probably needs to be some creative thinking about data-sets that exist for another purpose.

And, I think evaluating placebo/compliance issues in young men, in particular, would be a nightmare.

However – it might be worthwhile having a poke around. Not that there is much chance of getting access to the raw data, but this is where the ‘allergy’/intolerance studies would be interesting because of the remarkable level of compliance that is needed to follow an exclusion diet. (However, iirc, Hunter argued that if the Gut paper had used an intention to treat analysis, then the NNT was 9 and therefore less effective than the usual exclusion diet and not the claimed 2.5 – which the Gut authors accepted). However, as the Gut paper demonstrated, it is not that straightforward to come up with a plausible sham diet – and measuring diet compliance would be considerably more difficult than evaluating whether someone has taken their scheduled drugs/placebos.

one group I’ve seen characterised as at risk of poor compliance/prone to risk taking (in the context of chronic disease such as asthma or epilepsy where daily meds are prescribed) is young people in teens/early 20s, particularly males.

Also true of type I diabetics, in my (anecdotal) experience of the half-dozen I have known reasonably well. Of course, it is understandable there in the context of young diabetics having to “restrict” the behaviour that is normal for their age / peer group (irregular eating, eating crap food, getting boozed up / hung over, taking illegal substances etc etc). You can call all of this “risk-taking” behaviour, but it is pretty standard stuff from 15-25 (at least!).

Another important distinction in any study would have to be “diseases with perceptible and inconveniencing symptoms” (like a peptic / duodenal ulcer) versus “diseases with no very perceptible symptoms at present but possible nasty consequences down the line”.

Diabetes in young people might serve as an example of the latter – not-the-very-best blood glucose control may not have that many immediate consequences for a teenager , but it would probably be speeding them towards experiencing all the nasty “co-morbidities” (see DVNutrix’s post above) significantly earlier than someone with “tight blood glucose control”.

As has already been mentioned, the medical sociologists are clear that you have to approach compliance issues through this “risk vs. benefit as perceived by the patient“ framework. But the question of how that interacts with the effectiveness of placebo interventions is definitely interesting.