NEJM 7 July 2016 Vol 375
Bayesian cancer trials
I’m glad I haven’t got any cancer that I know of. I’m not unique in that, but I’m of an age when friends start getting cancer, and when they ask for advice it really brings home the current realities of oncology. This is the Wild West, with everything from futile treatments offered purely for the sake of doing something to expensive genomic buccaneering based on the belief that a new treatment might offer some sudden cure. There may be one best option for each individual, but for the time being there is no short cut to finding out what it is. In twenty years we might have cracked it, but what is going to happen to my friends (or perhaps me) in the interval? I guess one option is to get recruited into a trial using adaptive randomization. It may not cure you but it might give you a clue about what to try next and it might help similar patients in the future. That’s the sell. Now what of the reality?
I-SPY more than my little eye can handle
There are two reports in this week’s NEJM of breast cancer treatment guided by treatment response, and they both form part of a trial called I-SPY-2. When I read the abstract, I thought I understood what was going on. Then I read the papers, and decided otherwise. “The I-SPY 2 trial, a multicenter, adaptive phase 2 trial of neoadjuvant therapy for high-risk clinical stage II or III breast cancer, evaluated multiple new agents added to standard chemotherapy to assess the effects on rates of pathological complete response (i.e. absence of residual cancer in the breast or lymph nodes at the time of surgery).” But actually it was a lot more complex than that. Firstly the patients were selected into 10 groups on the basis of complex biomarker evaluations. And although pathological complete response at the time of eventual surgery was the primary outcome, treatment response evaluation during the trial depended heavily on repeat MRI and core biopsies, a process which sounds very costly and burdensome. These evaluations changed the prior probability of a positive result in the course of the trial, allowing patients who showed a poor response to drop out and be allocated to some other treatment. But actually we know very little about the predictive value of these measurements in this context. I’m a great fan of Thomas Bayes and his theorem, but it depends critically on the prior probabilities you feed into it. These seem too fuzzy. I’m also a great fan of Henri Poincaré who pointed out that beyond a rather low threshold of complexity, you cannot make causal predictions. My Poincaré complexity meter was going bonkers when I read this paper. But the NEJM allows a nice clear treatment recommendation into the end of the abstract: “Neratinib added to standard therapy was highly likely to result in higher rates of pathological complete response than standard chemotherapy with trastuzumab among patients with HER2-positive, hormone-receptor–negative breast cancer.”
Where is Poincaré when you need him?
Poincaré was an amazing genius in mathematics and several associated fields. Most of us are not. Our minds bounce off complex topics and hope that someone else is doing the hard work on our behalf. If there is delusive certainty on offer, we love to grab it and move on; and to my mind it should be the chief role of medical journals to make us stop doing that. If they can’t, what are they even for? There is a reasonably good commentary on the I-SPY-2 trial(s) which gives many of the necessary caveats. But this second paper—using the same methods as the first—still has the following conclusion to its abstract: “The process used in our trial showed that veliparib–carboplatin added to standard therapy resulted in higher rates of pathological complete response than standard therapy alone specifically in triple-negative breast cancer.” The process showed? Proved? Suggested?
Adapt and gain power
The general principles of an adaptive trial design are laid out in an article which is worth reading by methodologists but is pretty hard going for the less nerdy. The general principle of adaptive design can be used across all clinical areas. Essentially it is a way of ramping up the power of a trial on the basis of a pilot phase which provides an estimate of the size needed. The statistical design problems need not concern us here but they will make for some pretty chewy reading in years to come (e.g. “The initially planned enrollment of 10,900 patients, with possible early stopping for efficacy on the basis of a gamma [−5] alpha spending function [which generates group-sequential boundaries that resemble the O’Brien–Fleming boundaries] when 70% of the patients had been enrolled, provided the study with 86% power to detect a 24% lower relative risk, from an event rate of 5.1% in the control group to an event rate of 3.9% in the experimental-therapy group.) To gauge the number of participants you need to get a dependable result, you can either use real clinical events, or a surrogate measure such as FEV1, blood glucose, blood pressure or a cancer biomarker. Much depends on the predictive value of these measurements in the context of the trial population, as I mentioned. I’m not decrying it as an approach by any means, but it means that honest thinking, open data and collaborative working will be needed more than ever if medical science is not going to sink into a conceptual and statistical morass.
JAMA 5 July 2016 Vol 316
Heavy water hype
Not long ago, I read Diana Preston’s account of the history of nuclear physics from Curie to Hiroshima, Before The Fall-Out (2005). As war loomed and the military potential of nuclear fission became clear, all the combatants suddenly became interested in heavy water, or deuterium oxide. Mighty acts of derring-do punctuated the years from 1939-45: commando raids in Norway, secret convoys from Paris to North Africa, a final submarine load travelling thousands of miles from defeated Japan. All this centred on preserving, capturing, or destroying the heavy isotope of hydrogen, because it moderates neutrons sufficiently to allow controlled nuclear fission. Nowadays, anyone can buy heavy water online at $1000 per litre. And pharma companies are starting to use it as a fashion accessory to replace good old hydrogen which comes from water costing one cent a litre. The logic is that it “attenuates CYP2D6 metabolism and increases active metabolite half-lives and may therefore lead to stable systemic exposure while preserving key pharmacological activity.” Yeah, right, so you can take your pill twice a day instead of three times? That about the sum of it, in a 12-week pilot study of deuterium-substituted tetrabenazine to reduce chorea in people with Huntington’s disease. But no, hang on, we don’t even know that because the deutetrabenazine was compared with placebo, not ordinary tetrabenazine. The editorial on the trial states, “Assuming deutetrabenazine is not priced to be significantly more expensive than tetrabenazine, the favorable profile of deutetrabenazine would offer an additional option for patients and clinicians.” Great sense of humour.
Doctors helping dying patients to die
“Attitudes and Practices of Euthanasia and Physician-Assisted Suicide in the United States, Canada, and Europe” is a really useful summary of the evidence which concludes that “Euthanasia and physician-assisted suicide are increasingly being legalized, remain relatively rare, and primarily involve patients with cancer. Existing data do not indicate widespread abuse of these practices.” The wonder is that its first author is Zeke Emanuel, who has in the past expressed his opposition to doctors helping dying patients to die at a time and place of their choosing. Anyway, here again is the bottom-line message of this open-access review: “Between 0.3% to 4.6% of all deaths are reported as euthanasia or physician-assisted suicide in jurisdictions where they are legal. The frequency of these deaths increased after legalization. More than 70% of cases involved patients with cancer. Typical patients are older, white, and well-educated. Pain is mostly not reported as the primary motivation. A large portion of patients receiving physician-assisted suicide in Oregon and Washington reported being enrolled in hospice or palliative care, as did patients in Belgium. In no jurisdiction was there evidence that vulnerable patients have been receiving euthanasia or physician-assisted suicide at rates higher than those in the general population.”
JAMA Intern Med July 2016
Wealth and diabetes outcomes
OL Let’s go to a wealthy developed country with universal health coverage, low social inequality, and low levels of obesity. Is this the UK? No. Is this America? It certainly is not. Yet even in Sweden, socioeconomic status is a powerful predictor of all-cause and cardiovascular mortality in people with type 2 diabetes. Strangely, the risk was much lower for non-Western immigrants with T2DM. The best way to improve your diabetes outcomes is to earn more money than other people. We need a new metric for this, with units called Marmots.
The theology of eating fat
OL Priests have often dictated matters of diet to their followers, and the priesthood of public health is certainly no exception. Two articles on the JAMA IM website discuss the current theology of fat intake. The first is based on two familiar cohorts of US health professionals: the 83 349 women from the Nurses’ Health Study, and the 42 884 men from the Health Professionals Follow-up Study. These nurses and doctors kept food diaries from time to time and they can be used to calculate quintiles of intake for various kinds of fat. Comparing the bottom with the top quintile has become something of a fashion in the literature, so here are the hazard ratios for overall mortality: 1.08 for saturated fat, 0.81 for polyunsaturated fatty acid (PUFA), 0.89 for monounsaturated fatty acid (MUFA), and 1.13 for trans-fat. Add a bit of leeway for confounding and this comparison of extremes doesn’t provide the priesthood with very much to preach about, in my opinion. Though it’s amazing how zealously some people can preach when they are paid to do so.
The second study looks at the fatty acids which are regarded with greatest favour by most priests and their followers. You can actually measure intake in free-living populations by using biomarkers of seafood-derived eicosapentaenoic acid (EPA; 20:5ω-3), docosapentaenoic acid (DPA; 22:5ω-3), and docosahexaenoic acid (DHA; 22:6ω-3) and plant-derived α-linolenic acid (ALA; 18:3ω-3). By pooling 19 cohort studies, the authors were able to detect a weak (typically <10%) reduction in fatal coronary events in those with the highest level of ω-3 biomarkers ALA, DPA, and DHA. But when you next think of dinner, I would leave these papers behind.
Lancet 9 July 2016 Vol 388
For the first time in 18 years, I can’t find anything new of real generalist interest in The Lancet or its website. I’ve covered several items previously.
TheBMJ 9 July 2016 Vol 354
The best thing for me in this week’s The BMJ is The Junior Doctor Past v Present. Until recently, I could talk to Alex Paton about his experiences of housemanship in the 1940s, but when reading these reminiscences I feel that I myself am now a fossil of geological antiquity. In fact there was no single junior doctor experience in the mid-1970s: every hospital was quite different. Some were slightly nice and gave you toast, but most were already hostile places where you grabbed bad food and sleep when you could amid conditions of squalor.
What most strikes me on my rather infrequent talks to medical audiences in the UK is how little has actually changed in the NHS hierarchy. The system is still almost entirely provider-driven. The macropolitics of the NHS ensures that there is no link between demand and the supply of money to meet it; and the micropolitics ensures that anyone who seriously believes in the idea of patient involvement in service design either goes mad or is replaced within months. Yet the NHS enshrines everything that could and should be good about caring for sick people (I refuse to use the term “healthcare” in this context). The people are there; the data are there; the ownership is (theoretically) there as in no other system. So now that everything in Britain has reached the point of reductio ad absurdum, let’s hope that the self-evidently good ideas in Eugene Nelson et al’s piece about patient focused registries get adopted so that the NHS can start its long rebirth.
Plant of the Week: Meliosma pinnata subsp. arnottiana var. oldhamii
I was wandering through Kew Gardens as a birthday treat a few days ago when I spotted this unheard-of tree. It was by far the most flower-covered in this great paradise of trees, a great mass of foaming white about 12 metres high. Thirty years ago, I would have tried every means to get hold of one. But even then it would probably have been too late to stand much chance of seeing it flower. This is very much a tree to donate to some public location where your great-grandchildren might chance to see it.
I consulted the Venerable Bean (WJ Bean Trees & Shrubs Hardy in the British Isles) and discovered that the Kew tree came from a seed sent from Lushan Botanic Garden in 1935. Lushan has certainly seen some history since then: it was the place where Chiang Kai-Shek launched his counter-offensive against Japanese invaders in 1937 and it later became the “summer palace” of Mao Zedong and his government.
I do wish that our great gardens propagated plants like these. Collections like Bodnant, Savile, Batsford and others used to be run by people (often ex-army) who would enjoy taking cuttings from their choice rarities and running little nurseries as a side-line. If you showed a bit of arcane knowledge, they would take you to their secret treasure houses and sometimes even agree to sell one or two of their plants to you. A far cry from the lurid garden centres which now surround these places with their expensive commonplace offerings.