Richard Lehman’s journal review—18 December 2017

Richard Lehman reviews the latest research in the top medical journals

richard_lehmanNEJM Dec 2017 Vol 377

Gene therapy for haemophilia B

Haemophilia B is the rarer of the two sorts. It is caused by mutations in the factor IX clotting factor gene, rather than mutations in the factor VIII gene. Aren’t Roman numerals annoying? You can imagine Roman merchants getting in the sort of rage that we only know from Windows 10. Many would suffer from abacus foot, a penetrating injury caused by stamping on abaci with sandalled feet. Beads and shards of wood would fly in every direction, especially after attempted long division. I’m going to call factor IX factor 9, in the hope that it’s the start of a trend. Because in this tiny study may lie a medical breakthrough that we’ve been waiting for since the 1980s: a true cure for factor 9 absence, signalling a major triumph in gene replacement. The excellent editorial poses the key question: “What would an ideal therapy for patients with haemophilia, therefore, look like? It would be delivered once and be effective for a lifetime. It would have minimal toxicity. And it would be manufactured and delivered at a cost such that all patients anywhere in the world could receive it.” Yes! I love it, especially the last sentence. Currently, annual treatment with recombinant factor 9 costs $200,000, but after gene transfer therapy, VIII out of the X men didn’t need any at all, and the other II had very little. Theoretically, this should be a lifelong cure. The detail is fascinating: essentially the process consisted of infusing a single-stranded adeno-associated viral (AAV) vector consisting of a bioengineered capsid, liver-specific promoter and factor 9 Padua (factor 9–R338L) transgene at a dose of 5×1011 vector genomes per kilogram of body weight.

Women taking pill get slightly more breast cancer

The worst reading of this Danish whole-population study is that using hormonal contraception causes one extra breast cancer for every 7690 women per year. However, because this is an observational study, it is wrong to use the word “causes,” however good the adjustment for known confounders. Oddly, the authors don’t concede this, instead asserting that: “A quantitative bias analysis showed that a hypothetical unmeasured confounder would need to have a 50% prevalence in the population, increase the risk of breast cancer by a factor of 3, and increase the chance of using hormonal contraception by 2.5 times in order to eliminate the observed relative risk with current or recent use of hormonal contraception.”   

The accompanying editorial fights shy of contesting this and so, for reasons of time and statistical inexpertise, shall I. But we all need to think about where the tipping point lies in observational studies between a presumption of innocence and a presumption of harm. And how to convey this to the people actually using the intervention.

Sucking out proximal DVT disappoints

I may be against Roman numerals, but some Roman medical terms are just too juicy to abandon. Two thousand years ago, the term phlegmasia rubra dolens probably equated with “post-thrombotic syndrome,” since the leg becomes red after a deep vein thrombosis, whereas arterial obstruction renders it white (alba) or blue (cerulea). People with proximal deep venous thrombosis still often end up with painful, swollen red legs, and an obvious preventive strategy is to get rid of the clot as soon as possible. Commonly a catheter device is used to deliver alteplase continually to the clot until most of it has been broken up and retrieved: a costly and uncomfortable procedure which has been widely adopted with little evidence of benefit. Here is a trial in which 692 patients with acute proximal deep-vein thrombosis to receive either anticoagulation alone (control group) or anticoagulation plus pharmacomechanical thrombolysis (catheter-mediated or device-mediated intrathrombus delivery of recombinant tissue plasminogen activator and thrombus aspiration or maceration, with or without stenting). The primary outcome was development of the post-thrombotic syndrome between 6 and 24 months of follow-up. Alas, the only difference was an increase in bleeding complications in the thrombolysis group.

Edoxaban v dalteparin for cancer VTE: bleeds v clots

What were the manufacturers of edoxaban, Daiichi Sankyo, hoping for when they ran this trial comparing their product with dalteparin in patients with venous thromboembolism related to cancer? Probably not that “the rate of recurrent venous thromboembolism was lower but the rate of major bleeding was higher with edoxaban than with dalteparin.” But so it is. Let patients make the choice, and let them know the price.

JAMA Dec 2017  Vol 318

PAD away on the treadmill, forget the fancy stuff

Granulocyte-macrophage colony-stimulating factor (GM-CSF) sounds like a really stimulating choice for colonies of granulocytes and macrophages. But does it help people with peripheral arterial disease to walk further? No.

Machine learning v clinicians

“Mankind cannot bear much reality” is a popular quote from TS Eliot, American avant-garde poet turned English monarchist and High Anglican sage. He may be right, but what mankind finds even worse than reality is uncertainty. Eliot liked reading detective stories, in which pathologists typically declare, “Lord Chilsbury died between the hours of 8 and 9 yesterday evening due to the combined effects of ptomaine poisoning and external asphyxiation.” In truth pathologists are brokers of uncertainty. If you put 11 Dutch pathologists in a room and give them lots of slides of lymph nodes to look at, they will show only moderate agreement about which contain breast cancer metastases. However, the gold standard they were compared with was—wait for it—another pathologist who was allowed more time and better staining. Against this standard, a machine-learning deep algorithm performed better than the Dutch 11. 

And then there are photographs, which we know cannot lie, unless people look at them. A Singaporean study of retinal photographs from diabetic patients showed that once again a deep learning algorithm was at least as good as a panel of experts and trainees in detecting significant pathology. Again, there was a bit of circularity in defining the “gold standard.”

JAMA Intern Med Dec 2017

BP falls in decade before death

From about 40 onwards, most of us start thinking about death. Now that I am 67, I’m almost certainly going to die within a decade, or maybe two, and perhaps I should be looking for the warning signs. They are not quite what you might expect. I am male, and balding. That is always bad. I am a bit fat. That is good. My blood glucose is at the top end of normal. That too is associated with increased life expectancy at my age and in my social group. My blood pressure is down on what it used to be. Now that is really bad, according to this latest study based on repeated individual BP measurements in the UK Clinical Practice Research Datalink. Mean SBP and DBP decreased for more than a decade before death in patients dying at 60 years and older. These BP decreases are not simply attributable to age, treatment of hypertension, or better survival without hypertension.” Ah well.

The Lancet Dec 2017 Vol 390

The most important primary care article in the world?

It’s received wisdom that good primary care is the basis for any successful medical system. Now consider a workforce of 1.7 million primary care doctors serving a population of 1.4 billion people. That sounds quite big and important, doesn’t it? This survey of the Chinese primary care system needs reading by anyone interested in the progress of medicine. It’s a superb work of scholarship and analysis. It tries to be positive, but the picture it paints is grim. Paid on average about £5K per annum, primary care doctors in China have no specific training, no career structure, no standing in the profession, and no trusted position in the community. On the other hand, the Chinese government would like to rebuild the system and train up a new workforce of 400,000 doctors over ten years. Quite a challenge—but if they succeed, it might create a model for the world. Exciting, if you’re an optimist.

Starvation diet lowers glucose in “T2DM”

In crueller times, doctors would say to obese patients struggling with diets “There were no fat people in Belsen.” I heard it with my own ears. In my experience, it is extraordinarily difficult for free-living people to lose weight and keep it off. And there is plenty of evidence that on-and-off weight rebound is not a good thing. But here is a trial in which people with type 2 diabetes lost spectacular amounts of weight and kept it off for at least a year. About half of them became “non-diabetic” by current definitions and were able to give up oral medication. What was the intervention applied in primary care by Professor Lean and his colleagues? A formula starvation diet of 825–853 kcal/day for 3–5 months, followed by phased reintroduction of food and continued support. The conclusion drawn is that “Remission of type 2 diabetes is a practical target for primary care.” Well yes. Preferably using an option grid to help people decide between this single one-year trial and ten-year hard outcome data from bariatric surgery.

Walking in traffic-polluted air

In this study, 40 healthy volunteers, 40 people with chronic obstructive pulmonary disease and 39 with ischaemic heart disease (all aged 60 or older) were monitored before they took a 2-hour walk down Oxford St, or across Hyde Park, and for two days afterwards. Levels of air pollution were also measured. The Hyde Park constitutional did these individuals good, improving their lung function and pulse-wave velocity. Less improvement was obtained from walking down Oxford St, which the authors blame on the buses and taxis belching carbon, NO2 and ultrafine particles along that thoroughfare. But as a pedestrian of the same age I find the main hazard is collision with persons walking blindly with large bags of shopping while scanning their mobile phones. Not to mention being jostled by crowds of obstreperous French schoolchildren shouting at each other. It takes one’s breath away. The moral is always the same: avoid Oxford Street.

  • Stephen John Senn

    Re the dozen Dutch diagnosers (if one includes the gold standard). An old-timer like me can remember the 70s. (At least, this is still true as I write.) The phrase wasn’t machine learning at the time but computer-aided diagnosis. In fact I know about this article https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1789017/ even though as an 18 year-old Economic and Statistics student at Exeter University at the time I was not in the habit of perusing the medical press.. In fact I have to confess that I wasn’t even perusing many statistical journals at the time. and, furthermore, I had yet to encounter a computer. If De Dombal could do it for symptoms in 1972 it should not have been beyond the wit of somebody to do it for pathology.
    It’s astonishing that pathologists still seem to be making a living. Mind you, interesting to see a presumably very precocious Jane Horrocks as a co-author.

  • kidmugsy

    “however good the adjustment for known confounders”: even if adjustment has been honest, competent, and diligent, one can’t be certain of a job well done. That’s just how reality is, bugger it.

    The authors obviously deem unlikely a hypothetical unmeasured confounder having a 50% prevalence in the population, increasing the risk of breast cancer by a factor of 3, and increasing the chance of using hormonal contraception by 2.5 times. Fair enough; that’s an acute remark. But reality is indifferent to what they deem unlikely.

    Still, is the evidence persuasive enough for a woman to use? The typical woman will probably hear the news only through the incompetent and potentially hysterical medium of the newspapers and Twitter.