MPH Master pieces

Too old at 50

When you read a study like this you want it to be wrong. That is, if you are over 50 yourself.

The attention grabbing result is that surgeons aged over 50 got worse results than those aged 35 to 50. How can we pick a hole in that? The first temptation is to note that the study was based in France and blame it all on garlic and red wine.

Resisting that temptation, we can move on to methodology. Actually, the paper seems quite strong on that. A research question focussed on a well defined operation, with prospective data collection on surgical complications, and good compliance.

One possible explanation  for older surgeons doing less well might be failure to adopt newer techniques but the authors tell us that thyroid surgery really hasn't changed much over the past few decades.

The relationship between age of surgeon and complication rate is assessed through logistic regression to deal with confounding factors such as case mix. Often papers using these sophisticated techniques give you no sight of the raw data, which makes it very difficult to get a feel for what is going on. This paper does not have that failing - there is a useful set of graphs with the raw data. These allow us to see that the whole 'worse over 50' bit depends on just three or four individuals, and in fact the oldest of them is better than his slightly younger colleagues.

What does it all mean? I'm not sure.  But at a minimum it should prompt further studies which hopefully will be as well executed and reported as this one.

 

 

..and now breast cancer screening

Following last week's post on prostate cancer, here is one on breast cancer screening.

The basics are clear: breast cancer screening saves lives but also causes harm because of the stress and side effects of unnecessary operations in women who (a) didn't have breast cancer - false positives and (b) did have cancer but one which was never going to kill them - overdiagnosis.

The key problem is to put numbers on the benefits and the harms, so as to see which is greater.A trial is not feasible so you have to weigh it up using whatever data you can find. Table 1 in the paper helpfully shows the various data sources and estimates used in the calculations. One of the key variables is loss of "quality of life" while waiting for confirmatory tests after being told you have screened positive.

 

Prostate cancer screening

I like this commentary. It is a nuanced, and balanced, piece about a contentious area.

Sleep is a risk factor (well, lack of it is...)

Ten years ago I wrote a piece lamenting the lack of research into the epidemiology of sleep. So of course I now have to commend this paper.  The main limitation is that the study group were volunteers followed by fairly high loss / wastage: given that this is a sample of police officers, the fact that almost 80% were overweight or obese suggests substantial selection bias (or maybe things are a lot worse in the USA than I thought...).

the inverse care law

The inverse care law, coined by the great Julian Tudor Hart, states that those most in need of care receive it least. Here is a glimpse into one mechanism by which this law operates.

Tudor Hart served a deprived population in south Wales. Here is a paper which sums up his life's work.

period and cohort effects

After the flummery of the last two weeks, back to some solid epidemiology.

Age-period-cohort models are used to sort out whether an increase in disease incidence is due to a period effect or a cohort effect. A period effect will show itself as a sudden increase in incidence in all age bands; a cohort effect will only affect a particular age band, but throughout life, so as that cohort ages the peak in incidence will affect ever older age bands - 5 yr olds in 2005, 10 yr olds in 2010, 15 yr olds in 2015 and so on. Actually of course it is never as clear as that an period and cohort effects overlap.

The most famous example of a period effect was the increase in asthma mortality seen in the 1960s - mortality increased in all age bands over only a year or two. This was later seen as due to the introduction of a new treatment which though not obvious at the time increased mortality.

Cohort effects are less easy to exemplify but perhaps smoking will do - it became fashionable among the younger generation, who gradually aged.

If a disease is on the increase, knowing whether you are looking at a period effect or a cohort effect will help to rule in or rule out some potential causes of the increase.

So here is the worked example.

As always, note the data source: a service register. This affects ascertainment: people have to be on the register to receive special services for autism. The authors discuss the biasses which this may cause (e.g. pushier parents get on the register; poorer ones don't.)

 

 

 

Guideline development

Guideline development is part of the world we live in. Here is an interesting account from the American Cancer Society of how they will in future develop guidelines. A guideline for guidelines?

But joking apart we need to know that guidelines have been developed objectively, especially in contentious areas with clear commerical interests.

Outcome measures

Something a bit different today as there doesn't seem to be much of interest in the journals.

Suppose you were asked to come up with some indicators of how well your local health service was performing. It has already been agreed that the five domains of interest are:

1. Preventing premature death
2. Good quality of life in chronic disease
3. Speedy recovery from acute illness
4. Good experience of care
5. Patient safety

What indicators would you suggest? Here are the ones your colleagues in the Department of Health came up with.

 

A triplet of weight loss

This week not one, not two, but three trials of programmes for weight loss.

It's difficult to get one's head around so much information, but the PICO approach can help: for each paper get clear in your mind (ok, so put it on a piece of paper) the

Target Population (e.g. Body Mass Index > 30 / BMI >30 plus other risk factors)

Exact Intervention (group sessions / telephone support etc )

Comparator and

Outcome (weight loss at 3 months / 24 months etc).

Ideally you would wait for a systematic review to put it all together but we don't always have that luxury: we need to make up our own minds now.

A question raised by all of these trials is whether a weight loss of, say, 5% is worth having. In public health we tend to take the view that even small changes can add up to huge benefit at population level. The concept of '"minimal important difference" is taken very seriously in planning clinical trials: the smallest difference which is worth having from the patient's point of view.  

Mobile phones and brain cancer

Well, the first thing to say is that 'brain cancer' is for journalists. We need to be a lot more precise. For a start, most 'brain' cancers are metastases from primary cancers in other organs, and no-one is suggesting that mobile phones cause them.

So we are focussed on primary tumours of the brain, but even so we need to specify which type - gliomas? meningiomas? acoustic neuromas (more accurately called schwannomas)? One of the problems of cancer epidemiology is that cancers are named both by cell type (e.g. adenocarcinoma or squamous) and by location (breast, lung). A mobile phone is held next to the ear so if it is going to cause a cancer the most likely place is the auditory or acoustic nerve, or perhaps the temporal lobe of the brain. As to which cell type is most likely to become malignant, that's for the biologists to tell us.

There is a further complication - I just said 'become malignant' but in the brain even a benign tumour which keeps on growing will squash everything else and kill you. Hence 'brain tumours' not 'brain cancer'.

Unsurprisingly there have been quite a lot of studies looking at this. Mostly they are reassuring, as is this one. There is also a useful editorial. It's an expert opinion not a formal systematic review, and we wouldn't have invented systematic review if expert review were not so unreliable. Nevertheless these experts make some useful comments.

I also note that one research group consistently finds that mobile phone use is hazardous. How come they never get a negative finding? You be surprised how often this happens. I don't think it is fraud or worse; perhaps researchers are just good at picking studies which are likely to confirm their beliefs.