MPH Master pieces

Melanoma and sunscreen

The message from this study is what we want to hear, but there is plenty of interest in the detail. The research was done in a small town located, appropriately enough, on Australia's Sunshine Coast.

Melanoma is a slightly tricky cancer - making a diagnosis is not simple, even for pathologists. In this paper a distinction is made between invasive and 'preinvasive'. Over the past 30 years pathologists have been more ready to label lesions as melanoma, which has undoubtedly affected statistics on the number of cases (but this process will not of course affect statistics on the number of deaths).

What should we do with these findings? If you lived in Australia, would you put sunscreen on your kids every day?

Don't drop dead!

Modify your risk of sudden cardiac death - don't smoke (obviously); BMI between 21 and 25; eat a Mediterranean diet (veg, grain,nuts, fish and daily alcohol); and exercise 30 minutes each day at 3 - 6 METs. At least if you are a nurse. 

Here is the study - worth a careful read. I think its main value is to show just how much depends on lifestyle. What would be your guess on how much your chances of a heart attack decrease if you keep to the four rules above? Though admittedly the diet and (even more so) the exercise rules are quite tough. 

Notice how each variable was measured. Also notice that of the 120 000 nurses in the cohort only 80 000 are included in the study, mostly because of non-completion of the diet questionnaire. I'm not sure why they focussed on sudden death rather than all death from cardiovascular disease. 

Reducing violent crime

Is violence a public health issue? I've always thought so. So I am pleased to see a formal write-up of the famous Cardiff project. I guess the project leaders must have been disappointed that they didn't see more dramatic effects, but the intervention was simply to provide up-to-date intelligence to the police. From then on it depends on whether police action can prevent injury. Selling beer in plastic not glass seems sensible; other measures seem less assured of success.

The statistics are, as always with time series analyses, tricky but the graphs speak clearly.

Smoking - the social context

Here is an interesting use of data from a national survey. Must admit I hadn't heard of the British Household Panel Survey before, but the 'Panel' in the title indicates that it is longitudinal - the sample people are interviewed in successive years. Hence behaviour change - in this case smoking quitting and smoking initiation - can be studied.

The questions on social capital are pretty basic - one question on whether you can trust other people, and one on membership of groups and clubs - but that is typical for items in omnibus surveys. Nevertheless the results are interesting.

Not quite sure why it took two authors in Sweden to analyse this British dataset! but it reminds us that often with these large datasets contain a wealth of knowledge waiting to be revealed by creative ideas for analysis.

The BHPS is now part of the new 'Understanding Society' survey, which also incorporates health measurements such as blood pressure, BMI, SF12 and much else.

Screening: Lung 1 Ovary 0

You don't see full trials of screening very often, but here are two mega trials.

First up: cancer of the ovary. It is one of the commonest fatal cancers in women, often presents late, easier to treat if caught early, the screening test (blood sample and ultrasound, albeit transvaginal) is simple: seems like a no-brainer. Gynaecologists, understandably distressed by seeing so many cases present late, argue long and loud for screening. But...

It is a good write-up, worth reading carefully. Note also that 78 000 women were randomised - that's why you don't see full trials of screening very often. Also, once you've done a huge trial like this you probably won't want to fund a subtly different one (e.g. different screening interval, newer technology etc etc).

Next: lung cancer. An even bigger mega trial. Screening trials for lung cancer forty years ago were disappointing. With the technology then available, by the time you could see the cancer it was too late. But nowadays imaging is much better. Useful to use the PICO framework for this paper: Population, Intervention, Comparator, Outcome. Note the target population and the comparator. All cause deaths is an important outcome to check because of course all that radiation from CT scans could perhaps have caused a few extra leukaemias etc.

Unusually but entirely correctly the authors emphasise the importance of cost-effectiveness studies now the main result is in (they have these in hand).