Richard Lehman’s journal review – 23 April 2012

Richard LehmanJAMA  18 Apr 2012  Vol 307
1583   George Orwell predicted a nightmare world where soothing words would mean their opposites, and gave his dystopia the date of 1984. It was about that year that the term patient centred first appeared in the medical literature, coinciding with the time when the medical-industrial complex went totally out of control in the USA and patients were thrown entirely to the mercy of the market. Books and papers about patient-centeredness (sic) proliferated in America during the 1990s, but the momentum of medicine there has continued to career in the opposite direction. Now that total chaos and unaffordability loom, the US government has set up the Patient Centered Outcomes Research Institute with a hefty budget to find out how to put things right by finding out what systems of care work best for patients. A laudable aim and a fine-sounding name, certain to arouse suspicion among cynics everywhere; but this particular cynic is amazed and optimistic. To find out why, listen to the visionary speech which Harlan Krumholz gave to the PCORI Patient and Stakeholder group a few weeks ago. This goes way beyond the usual rhetoric of being nice and involving patients, and commits PCORI to a radical agenda of patient empowerment – the only way that health systems the world over can reclaim the true purpose of medicine. This article shows how Harlan’s vision is shared by others in the developing organization.

1585   But the moment that you attempt to empower patients, you run into problems. Patients as well as doctors like to believe that there must be a single right answer for every problem, when very often there is not. As I’ve said before, Harlan’s surname (meaning crooked wood in German) always reminds me of Kant’s famous dictum, “out of the crooked timber of humanity, no straight thing was ever made”. And it’s no good torturing the evidence by exercises in subgroup analysis and modelling: in most of medicine, there is irreducible uncertainty. Here is a nice short philosophical piece by David Kent and Nilay Shah, headed with the splendid observation of George Box that All models are wrong, but some are useful.

1587   Three non-clinicians discuss the problems of continuous patient engagement in comparative effectiveness research. Now comparative effectiveness research is actually fiendishly difficult, for reasons I will try to outline very briefly in a moment; and securing patient involvement in research is also difficult, but absolutely essential. In fact it will be a measure of PCORI’s success if it can demonstrate that every aspect of its research is genuinely patient-centred – i.e. that it listens to the patient voice at every stage, and that every output has direct bearing on decision making with patients and society. The ultimate measure of its success, ironically, will be the disappearance of the concept of the patient altogether.

1593   In this hefty themed issue of JAMA, there now follow five examples of comparative effectiveness research (CER), followed by a knotty editorial with the title Is It Time for Medicine-Based Evidence? And here is the problem for you and for me, dear Reader: you cannot properly assess a paper on outcomes research or CER without some understanding of the following methods – multiple linear regression or analysis of covariance for continuous (dimensional) outcomes, logistic regression for binary (dichotomous) variable outcomes, proportional hazards analysis or Cox regression when a time interval is relevant to a binary outcome (i.e. survival analysis), and Poisson regression when outcomes are measured as counts. Moving on, you then need to employ these techniques in one or both of two conceptual processes which can help to balance the characteristics of unmatched groups in observational studies: propensity scores and instrumental variables. There are plenty of statistics texts to confuse the unwary, but there is no simple, comprehensible guide to outcomes research for the non-specialist. I know, because I am trying to help write one. And I am hoping somebody else will deal with all this while I write about patient-centredness. So finally, back to this study. You need not read it: it is simply a good teaching example for those who want to understand the use of propensity scoring in retrospective cohort studies. The study concludes that without needing a randomized controlled trial, we can be pretty certain that adding bevacizumab to carboplatin-paclitaxel chemo for advanced non–small cell lung cancer makes no difference. And that is useful knowledge for decision-making.

1602    So we’re getting accustomed here to the idea of extracting useful knowledge from unbalanced observational data. To do this requires both sides of the brain. Your left brain can immediately busy itself with the data, using extension tools like statistics software packages and tabulation methods. Here we’re looking at nearly a quarter of a million American adults with serious trauma transported to hospital either by ground or by helicopter. What does your right brain tell you about this problem? Mine tells me that you cannot match these groups because there are simply too many confounders. But the left brain goes ahead and tries, using every gizmo it can lay its hand on. After performing all its tricks, it reports that there is an absolute mortality benefit of 1.6% in those transported by helicopter, and a small benefit in functional outcomes. And what does my right brain say about that? That you still cannot be certain you have really corrected for confounders to that level of difference, and that even if you had, it would be no argument to buy more helicopters as a strategy for improving trauma outcomes.

1629   Let’s skip to the final study of the 5, which uses an instrumental variable approach to account for measured and unmeasured differences between patients with clinical stage T1a kidney cancer treated with partial or radical nephrectomy.

Briefly, the instrumental variable approach identifies an instrument (variable) that is thought to be associated with the treatments of interest but not with the outcome. Here there is a very striking difference in long-term outcomes: the hazard ratio for death in those treated with partial nephrectomy rather than radical is 0.54. Although my left brain struggles to follow every stage of the methodology, my right brain tells me that a difference of this magnitude is unlikely to be due to skewed assumptions or residual confounding.

NEJM  19 Apr 2012  Vol 366
1467    Now that we’ve finally escaped from JAMA and all this stuff about CER methodology, let’s look at this first paper in the New England Journal. Being in the NEJM, funded by the NHLBI and conducted by a distinguished team of researchers, it must be right, and it concludes that “In this observational study, we found that, among older patients with multivessel coronary disease that did not require emergency treatment, there was a long-term survival advantage among patients who underwent CABG as compared with patients who underwent PCI.” Proof at last of what we all suspected: new tubes must be better than stents. But hang on, what was the absolute mortality difference between these groups? The median follow-up period was 2.67 years, at which time the survival lines were beginning to diverge in favour of CABG, but not by very much. In the minority of patients followed to 4 years, the difference was statistically significant and stood at an absolute value of 4.4% provided one accepts the methods of the study. And what are these methods? Why, our new friends propensity scores and inverse-probability-weighting adjustment. So we are back to the problems of comparative effectiveness research with a vengeance. The left brain, without the help of complex statistical computation, cannot interrogate these results; while my creaky old right brain tells me that I cannot make use of this information in decision-making with patients, because there are too many variables to rely on such small differences. In fact I think we may need new methods of describing the confidence limits when using these two-stage weighting adjustments with unbalanced groups. So do we need another RCT comparing CABG with PCI using current methods? The editorial discusses this question, but not with any satisfactory conclusion. I think equipoise still best describes the clinical situation.

1477   Yippee! We’ve finally got away from CER and on to the best kind of medical paper – a randomized controlled trial conducted without industry funding, with a clear result that will be of benefit to thousands of patients with muscle-invasive bladder cancer. And British too! The simple trick is to give chemotherapy using fluorouracil and mitomycin C at the time of radiotherapy. This provides a sustained survival advantage without a significant increase in adverse effects.

1489   And now, like a Common White Butterfly, we must return to the field of cabbage. CABG can, as all of you know, be performed with a cardiopulmonary bypass pump or without. Off-pump CABG is technically more challenging but is supposed to reduce the amount of debris reaching the brain during surgery. This trial (given the unoriginal acronym CORONARY – how much jollier BRASSICA might have been) randomized 4752 patients in 79 centres to have their cabbage done one way or the other. At 30 days, there was no significant difference in gross outcomes, but they acknowledge that “Neurocognitive outcomes and economic data may have an important effect on and substantially influence the ultimate interpretation of the primary findings.”

1515   Here’s a good update on alopecia areata, a T-cell–mediated autoimmune disease in which the gradual loss of protection provided by immune privilege of the normal hair follicle plays an important role. But I must leave you with these bald facts and rush forward to the remaining journals.

Lancet  21 Apr 2012  Vol 379
In his Offline column this week, Richard Horton tells us that this physics-themed issue of The Lancet is timed “to coincide with the death of Albert Einstein on April 18, 1955.” Now this is an idea that Einstein would appreciate: what, after all, are 57 years and 3 days in the continuous fabric of space-time? In fact Einstein once wrote a letter of consolation to a bereaved friend using this idea. Or, as TS Eliot more gloomily declares as the beginning of Burnt Norton,

Time present and time past
Are both perhaps present in time future,
And time future contained in time past.
If all time is eternally present
All time is unredeemable.

RH reaches similar heights of mysticism as he tells us why physics is special. He had discovered that underlying everything in the material world there is physics. It follows, he declares, that “Physics is at the heart of our society and so our understanding of health.”… “all of us interested in the future of healthcare, should declare and implement a passion for physics. Our Series is our commitment to do so.” Yes indeed. Perhaps it is also time for our column to be renamed Offwall.

1489   My old practice used to number amongst its diabetic patients a Canadian lady who had been treated by Banting in the 1920s: she owed a 60 year extension of her life to the insulin he had just isolated from the pancreatic cells of animals. Animal insulins were still the only kind in use when I first took up doctoring, and very good they were. Many patients complained of hypoglycaemia without warning, and erratic control, when they were replaced by human insulins in the 1980s. But despite their lack of demonstrable superiority, these had almost entirely replaced the cheaper, older insulins within a few years. Insulin manufacturers managed to develop a highly effective mechanism for disseminating their expensive new products by means of trials sponsored by industry, peer pressure from academic diabetes centres receiving large funds from industry, and primary care nurses trained by industry. This mechanism was put into action again in another huge wave of marketing once the patents on human insulin expired, and the so-called analogue insulins – modified by a peptide of two – took over, further ratcheting up costs without any improvement in outcomes. The effect has been to make insulin treatment unaffordable in some developing countries. But still the search for profits goes on. The latest trick is to produce ultra-long acting insulins such as NovoNordisk’s insulin degludec and seek to prove their advantage over existing basal insulin regimes. This trial in type 1 diabetes shows overall equivalence with insulin glargine, including in the incidence of hypoglycaemia. But fewer of these episodes occurred at night with degludec. In a trial in type 2 diabetes, the rates of hypos with degludec just managed to squeeze under the statistical bar and come out lower than glargine (95% CI 0.58-0.99). On open-label, manufacturer-sponsored trials of this sort do billions of dollars’ worth of sales depend. The Lancet chooses to devote most of its research space this week to them, perhaps expecting good sales of reprints. It would be good to be told.

1551   Having declared his passion and commitment to physics, complete with the Royal We, Richard Horton has managed to pull in a singularly weak set of contributions for his Physics and Medicine series, the strongest of which is the last – mainly because it is much more about biology than about physics. It’s a fascinating exploration of such things as fractal patterns in nature and the problems of scale in biological systems, and I would strongly recommend it.

BMJ  21 Apr 2012  Vol 344
As always, there are plenty of good things to read in this week’s BMJ. I would particularly recommend Iona Heath’s beautiful lament on the demise of the NHS -how a war-ravaged generation strove to create a fairer society, and how we are strangely set on destroying it; and Margaret McCartney’s piece on why screening for streptococcus B in pregnancy may not be the unmixed good it is portrayed as in British newspapers. As for original research, there is PhD student-led systematic review of metformin plus insulin versus insulin alone in type 2 diabetes. “There was no evidence or even a trend towards improved all cause mortality or cardiovascular mortality with metformin and insulin, compared with insulin alone in type 2 diabetes. Data were limited by the severe lack of data reported by trials for patient relevant outcomes and by poor bias control.” I used to think that the last 40 years of diabetes research had yielded just one fact that one could rely on: metformin is a good drug. Now, sadly, I’m not even sure about that.
Physicist of the Week: Michael Faraday

I – or rather We – declare and implement our passion and commitment to physics by this unique celebration of the life and work of Michael Faraday, timed to coincide with his death on 25th August 1867, give or take 145 years and a few months.

Faraday is the physicist everyone can love because he was kind to children and animals, and bad at maths. He led a life of blameless application, humility, piety, kindness, and good sense: in fact his life, shorn of the physical discoveries, is literally too boring to read about. If you want proof, try Michael Faraday (1864), by John Hall Gladstone, free on Kindle. You will do better with The Electric Life of Michael Faraday (2009) by Alan Hirshfeld, a physics professor who explains some elements of Faraday’s conceptual achievements, but fails utterly to match the drive and luminosity of Richard Holmes describing Faraday’s mentor Sir Humphry Davy in The Age of Wonder. Faraday needs a biographer to match his stature. A really able writer with a knowledge of science is needed to explore the paradox of a man who achieved amazing feats through a mixture of conceptual freedom and endless meticulous experiment. Although Faraday was a religious fundamentalist, his God was the very opposite of Newton’s (or Milton’s) determinist tyrant, playing with billiard balls of all sizes from the atomic to the cosmic. Faraday dismissed ball-atoms and action at a distance: for him forces existed as fields and vibrations, and without a single mathematical equation he worked out the basis of electromagnetism and went a long way towards relativity and the modern view of the atom.

Einstein always kept a portrait of him in his room. It took the genius of Thomson and Clerk Maxwell to create what we now recognize as the mathematical physics of the later nineteenth century out of the qualitative experimental descriptions of Faraday. While they toiled on the equations, he slipped gently into senility – childless, blameless, and finally wordless: a strange and rather haunting hero of science.