If, as Michael Baum states, you have to screen 1,000 women with mammography for ten years to save one death from breast cancer, is computer-aided detection going to transform the scene? Look at the ROC curve on p.1047: the computer shifts the area under it by 5%. You would still have to screen 1,000 women for 9.5 years. Can it be worthwhile?

The true scientist rejoices when her/his hypothesis is refuted, and this paper provides very clear refutation of the idea that invariant natural killer T cells play an important role in asthma and chronic obstructive pulmonary disease. This idea was put forward by Akbari et al. in the NEJM last year, and I expect some learned correspondence will follow. The authors emphasise how difficult it can be to quantify cells in a mess of bronchial gunge. For myself, I shall be backing the class II MHC-restricted T cells instead, as the authors advise, once I have discovered what they are.
The biggest risk for venous thromboembolism is going into hospital. I was surprised to discover from Albert Speer’s memoirs that this was common knowledge even in 1944, when following an injury he was confined to strict bed rest by Himmler’s physician in an attempt bump him of with a pulmonary embolus. There are several known risk factors which mandate prophylaxis for hospitalised medical patients, quite apart from having a physician from the SS: in the US guidelines, these include heart failure, severe respiratory disease, stroke and immobility, but oddly do not include the various cancers which greatly increase the risk of VTE.
Over the last year, I have been involved in lengthy discussions with a group of mainly academic friends about building a diagnosis from the ground up, the title of this article in the series “Clinical Problem-Solving