The liveliest e-mail streams I have ever encountered are the ones which are currently coming out of the Overdiagnosis Group, set up by Margaret McCartney last year. The group is now a standing body within the Royal College of General Practitioners and most of its members are working GPs. But there are participants from all over the world too, and from other medical or non-medical disciplines.

A lot of recent discussion has centred on how we actually share decisions with patients. Since I retired from full time general practice five years ago, it’s something I have thought a lot about. It is far more complex than just using evidence synthesis and decision aids, though these can be valuable in their context. Andrew Spooner, an experienced GP and RCGP Council member, suggested that we use the example of spironolactone as add-on treatment for heart failure to debate the key issues. This pressed several buttons for me and I found myself writing a rather long case example:

Let’s say that your patient is an 80 year old woman with diabetes, angina, and arthritis and breathlessness which has been attributed to heart failure because her LVEF is 32%. She is already taking furosemide, carvedilol, and losartan (she could not tolerate her ACEI due to cough). She has just been to the cardiology clinic and the SpR has written to say that “Mrs B reports worsening dyspnoea on exertion. Could you please start her on spironolactone 25mg daily and monitor her electrolytes in two weeks. Today her creatinine was 189, K 4.8, Na 131.” The letter has been copied to the community heart failure nurse.

So Mrs B comes to see you and tells you that the clinic doctor said she should start some new pills to help her breathing.

So right away you are in a trap and your easiest way out is to do what you are told (obedience bias) and what the patient expects (expectation bias). You may also be worried about what might happen if you decide not to follow the SpR’s instruction and the patient gets worse (risk aversion bias). You are aware that adding spironolactone is often used as an add-on drug for heart failure but you don’t know the precise trial evidence (ignorance bias). You could look it up, but not feasibly in the middle of a consultation in which Mrs B has already moved on to talk about her knee pain (time pressure bias, knowledge access bias). You have just begun to think what mixed messages the patient might get if the heart failure nurse tells her that she should have been started on new tablets (team pressure bias) when you realise that Mrs B is talking to you. “It’s been a struggle getting to the shops ever since you stopped my diclofenac tablets a while back doctor. So my sister told me to get some ibuprofen tablets from Boots and I can get around better now. Can I have some on prescription?” The penny drops: her worsening breathlessness was probably due to restarting an NSAID and it would actually be dangerous to start her on spironolactone. You could call that incomplete history bias or tunnel-thinking bias.

So you advise her to use paracetamol instead for her knee pain, and you get the usual reply, “I’ve already tried them and they’re rubbish.” The ten minutes have ticked by and you suggest that she comes back in two weeks, having stopped taking ibuprofen. You jot down her name and “spirono” on a yellow sticky and promise yourself to look up the evidence before she comes back.

Let’s not dwell too much on how satisfactory Mrs B found this consultation. You did your best under difficult circumstances. You may have saved her from renal failure or fatal hyperkalaemia. Twenty minutes have since passed, and you have moved on to discuss the marital difficulties of a local accountant (whose wife is also coming to see you, with a completely different story) and the contraceptive needs of a temporary resident.

Mrs B returns after a fortnight, unconvinced that her breathing is any better and miffed that she can’t keep taking ibuprofen. In the meantime, you have looked up the RALES trial. It was stopped early because of a mortality difference in favour of the intervention (early stopping bias). Participants were selected on the basis of severe (NYHA grade IV) heart failure and an EF <35%. All were taking a loop diuretic, 95% were taking an ACE inhibitor (usually captopril), 70+% were on digoxin, and only 10% were on a beta-blocker. Their mean age was 65, and 72% were male. The primary outcome measure was all-cause mortality.

You become aware that although there are some similarities between Mrs B and the trial population, there are actually rather more dissimilarities. There is a pretty gross mismatch in age and concomitant drug therapy. And yet you are inclined to go with the flow because of optimism bias. This bias also leads you to assume that because the drug reduced mortality in this very sick population, it also improved symptoms. But if you look carefully at the text, you see that “in the spironolactone group, the condition of 41 percent of the patients improved; it did not change in 21 percent, and it worsened in 38 percent.” So from Mrs B’s point of view, the chances are equal that she will feel better or worse.

Now we get to the really difficult bit. “In the placebo group, the condition of 33 percent of the patients improved; it did not change in 18 percent, and it worsened in 48 percent… The difference between groups was significant (P<0.001 by the Wilcoxon test).” Blinded by statistics and misled by optimism bias, you might be inclined to believe that spironolactone is likely to prevent a worsening of Mrs B’s symptoms. In fact you haven’t a clue whether it will.

The pressure is still on you to prescribe – it is the path of least resistance. But as well as considering benefits, you should also consider harms. You have a vague unease about the risks of hypotension and hyperkalaemia, which the RALES report does its best to allay. “The incidence of serious hyperkalemia was minimal in both groups of patients.” But you may wonder if a lot of it is missed in real life, and the observational literature on this is not reassuring. In Canada, the RALES trial was followed by an increase in deaths from hyperkalaemia in heart failure patients started on spironolactone.

So, if you are honest with yourself, you conclude that for Mrs B, you do not know the number needed to treat or the number needed to harm. To be sure, you can calculate them for the RALES trial population, but you would actually be doing patients a grave disservice if you put them in any kind of shared decision aid, because they are inapplicable to almost all heart failure patients in 2015. And at best they are population-specific estimates which cannot predict what will happen to individuals.

The practical question that now faces you is how to share this information with Mrs B. It’s an important question for her. You conclude that you might give the drug a try but you decide that it might be best to start on a lower dose and check her BP sitting and standing and find out about her potassium and renal function sooner than the SpR advised. You ask Mrs B in and start the conversation.

By a leap of imagination you decide to put the rules of EBM, the NICE guidelines on heart failure and the renin-angiotensin-aldosterone pathway right out of your mind. You start with an open question – “How are things since I saw you?” for example.

“Well, they’re not any better, I can tell you doctor, since you stopped me taking any ibuprofen. I can hardly get to the shops at all now.” “But how’s your breathing now?” “It’s not my breathing that’s stopping me doctor, it’s my knees.”

And now, at last, you start asking the right questions and discover that Mrs B is not concerned with how long she will remain alive or even about her breathing but how she can be the least burden to her daughter and grand-daughter who have to take her on every journey out. She does not want to visit the practice for regular BP checks and blood tests. She’ll take the tablets the clinic doctor said she should have, “but only if you think they’re worth it doctor.”

Do you?