NEJM 16 October 2014 Vol 371
1507  I hate military metaphors for cancer as much as anybody, but here is a study which describes hell in the leukaemia trenches. The 30 patients in the trial had acute lymphoblastic leukaemia. The youngest was 5 years old; most were under 20. All of them had relapsed after initial chemotherapy, which heaven knows is bad enough. Eighteen had then endured the horrors of allogeneic stem cell transplantation, and then relapsed again. Some had been through other experimental treatments. Now they were given an infusion of autologous T cells transduced with a CD19-directed chimeric antigen receptor (CTL019) lentiviral vector. In all of them, this caused a cytokine release syndrome, which was severe in over a quarter of cases and had to be treated with the anti–interleukin-6 receptor antibody tocilizumab. The complete remission rate at the end of all this was 90%. So victory at a high price: the challenge now is to operationalise this treatment in a way that is bearable to patients and affordable to health systems.

OL  Two years ago, persuading the pharmaceutical industry to share full data on all its clinical trials seemed an impossible task. Now this has become routine practice for many companies. One of the leaders was the UK based giant GlaxoSmithKline, which started giving access to de-identified data from its trials in May 2013, through an independent review panel. In a Perspective piece the members of this panel report on progress in the first year. Currently, there are over 1200 trials on the site, but the number of applications has been small—58 in the first year. I have had the privilege of being part of a similar initiative (YODA), in which our Yale based team will become the completely independent holder of data released by Janssen Pharmaceuticals. Our website has just opened and we will facilitate access to any researchers who offer a reasonable proposal for analysis. I do hope that there will be growing realisation of the importance to science and medicine of these huge resources for improving the understanding of the treatments that we use for so many millions of people.

JAMA 15 October 2014 Vol 312
1513  This Viewpoint gave me a guilty start because it sets out to “discuss the meaning of patient centeredness in research and reasons for conducting patient centered outcomes research” on behalf of PCORI, the Patient Centered Outcomes Research Institute, which put out its first call for bids two years ago. And in the year before that, Harlan Krumholz invited me over to Yale with the remit of putting together a book that would address those precise questions. I suppose I should take comfort that in this piece the leading figures of PCORI show that they too are not finding this an easy task. My working definition of patient centred outcomes research was “any form of enquiry that would help patients to decide on the best course of action in matters relating to their health.” The reason for doing this kind of research is that so much current effort is of little or no direct relevance to decision making by patients, health professionals, or society at large. So far, it remains to be seen whether even so great and commendable a project as PCORI will do anything to redress the balance.

1531  Here’s an example of one kind of patient centred outcomes research: a study of so called quality indicators in New York obstetric units, and their relationship to maternal and neonatal morbidity. Result: “there were no correlations between the quality indicator rates and maternal and neonatal morbidity.” Now that really is a bit startling.

1542  Another example, although this study is more directed towards informing policy than individual choice. Do patients over 65 discharged to “skilled nursing facilities” in the United States show a difference in hospital readmission rates? The researchers again found that current metrics did not predict much. “Available performance measures were not consistently associated with differences in the adjusted risk of readmission or death.”

JAMA Intern Med October 2014
OL  Talking about antibiotics last week, I said that if I was ill enough to be in hospital with an infection, I would want my treatment to work first time round. “Community acquired pneumonia” is a common cause for hospital admission, but in Europe at least only a small proportion of people who get pneumonia end up in hospital. So we are looking at the worse end of the spectrum in this study from Switzerland, in which a β-lactam antibiotic alone was compared with a β-lactam and macrolide combination in moderately severe community acquired pneumonia. The conclusion states that: “We did not find noninferiority of β-lactam monotherapy in patients hospitalized for moderately severe community-acquired pneumonia.” The non non-inferiority was found only in patients with atypical organisms causing their pneumonia. This makes not nonsense.

Lancet 18 October 2014 Vol 384
1429  This week’s Lancet is big on outcomes research too. An opening editorial heralds: “A renaissance in surgical research,” but all we get are two big observational studies of knee replacement and an article about anaesthesia. I do believe that there is a lot of excellent surgical research going on, encouraged by the IDEAL collaboration, but it would be hard to guess from these papers. The investigators in this first paper have gone through the data relating to 467 779 primary knee replacements carried out in England and Wales between 2002 and 2011. In that time, the mortality for such operations fell from 0.37% to 0.2% even after adjustment for age, sex, and comorbidity. The conclusion of the paper states that efforts to further reduce mortality should concentrate more on older patients, those who are male, and those with specific comorbidities, such as myocardial infarction, cerebrovascular disease, liver disease, and renal disease, because these groups showed the highest risk. All nice to know but not easy to act on.

1437  Let’s judge the next paper by my PCORI criterion of whether it would help me decide on what to do about my bad knee. The conclusion of the abstract makes it sound straightforward: “In decisions about which procedure to offer, the higher revision/reoperation rate of unicompartmental than of total knee replacement should be balanced against a lower occurrence of complications, readmission, and mortality, together with known benefits for UKR in terms of postoperative function. If 100 patients receiving TKR received UKR instead, the result would be around one fewer death and three more reoperations in the first 4 years after surgery.” Very nice, but what room is there for patient choice when the “offer” is made? Surely a lot will depend on (a) whether the osteoarthritis is bicompartmental and (b) the values of the individual patient in weighing up the choice between a safer immediate procedure and the chance of having to go back in for a second and more dangerous operation. So this kind of outcomes research can’t provide all the answers, particularly as so much depends on propensity matching. If you need a reminder of what that means, here it is: “Propensity score matching (PSM) is a statistical matching technique that attempts to estimate the effect of a treatment, policy, or other intervention by accounting for the covariates that predict receiving the treatment.” Doing that for 25 334 UKRs and 75 996 TKRs must have been a colossal job, and you have to wonder how accurately all the covariates could have been recorded and balanced out.

1446  If you are about to undergo surgery and your anaesthetist wishes to use nitrous oxide, let not your heart be troubled. In a randomised, assessor-blinded trial in patients aged at least 45 years with known or suspected coronary artery disease having major non-cardiac surgery, nitrous oxide did not increase the risk of death or cardiovascular complications. In fact, you might want to try some at sub-anaesthetic doses just for pleasure. The first person to do this was the intrepid young Cornishman Humphry Davy in 1799, when he worked with Dr Beddoes at the Pneumatic Institute in Bath. You may experience “sublime emotion connected with highly vivid ideas,” according to Davy’s electrifying account of his extensive experiments on himself and others, which you can read page by page.

The BMJ 18 October 2014 Vol 349
I’m struggling to find much to report from The BMJ this week, having already raided the website on previous occasions. I reported on the failure of a diabetes decision aid in a Dutch trial, and now you can read an editorial about that by two leaders in decision aid design from the Mayo Clinic. This is worth reading for its account of the principles of shared decision making as a conversation, but the authors are too polite to mention the really awful quality of the decision aid used by the Dutch team. This trial just proves we need better ones, and Victor Montori was too modest to mention his own, which is the best I know of.

As a keen follower of David Sackett in the 1990s, I got very interested in the nature of diagnostic decision making and the evaluation of diagnostic tests. But the more I dug in, the more disillusioned I became. In theory—and in the JAMA “Rational Clinical Examination” series—you can look at all the tests and combinations of tests that are used to reach a particular diagnosis in particular sets of patients, and then work out a decision rule to help clinicians use them in the right sequence or combination. In reality, it is hard to find any clinicians who actually do this in real life. A two-by-two table and a Bayesian pre-and post-test probability chart ought to be in every clinician’s pocket, but most of us use our biased, rusty old brains instead. Here is a massive study that was designed to develop a risk prediction model to preoperatively discriminate between benign, borderline, stage I invasive, stage II-IV invasive, and secondary metastatic ovarian tumours. “The Assessment of Different NEoplasias in the adneXa (ADNEX) model contains three clinical and six ultrasound predictors: age, serum CA-125 level, type of centre (oncology centres v other hospitals), maximum diameter of lesion, proportion of solid tissue, more than 10 cyst locules, number of papillary projections, acoustic shadows, and ascites. The area under the receiver operating characteristic curve (AUC) for the classic discrimination between benign and malignant tumours was 0.94 (0.93 to 0.95) on temporal validation.” This is supposed to help surgeons prepare for the right kind of operation for these tumours. Maybe it will be used: it was certainly conscientiously developed.

Plant of the Week: Schizostylis coccinea “Fenland Daybreak”

We once used to have lots of Kaffir lilies in our garden. They are said to exhaust the soil so you are supposed to keep moving them on, but when we did that they began to die out rather than multiply. It may have been owing to a dry summer, because these South African plants like a moist soil with their helping of sunshine.

We have just bought this nice cultivar with coral pink flowers. I am worried though that the “Fenland” in its name may bode ill for its survival in our most unfennish dry clay soil. We’ll try to put it in some better, free draining soil and water it all summer, since that is what it likes. Then in October we may be able to revel in its unexpectedly optimistic looking blooms until the dark claggy frost of November cuts them down and pronounces the garden dead until late February.