Risk calculators in clinical practice

I haven’t had my blood pressure checked for a couple of years. I think the last time was for a life insurance medical so I wasn’t given much option. As with all investigations, it’s always worth having in mind what you plan to do with the result, and I’m pretty sure I wouldn’t be in any hurry to take blood pressure medication. We use QRISK in general practice (it is embedded in some of the GP computer systems) and I’ve been playing around with QRISK online. Here’s my QRISK score based on my current risk factors:


So, over the next 10 years that’s a risk of 0.8% (1 in 125) of a heart attack or stroke. Most of my risk is related to my age and my gender. Can’t do much about them.

Risk calculators can be a bit weird. You’ve got to be careful how you use them. Let’s assume that in fact I’ve got a slightly elevated BP – perhaps a systolic of 160mmHg, which according to NICE guidelines should be treated at that level. According to QRISK-2 (2011), my risk of a heart attack or stroke over the next 10 years is now 1.3%. That is, in my opinion, remains a fairly low risk (and just a 0.5% increase in absolute terms from having a normal blood pressure).

But, say I have a change of heart, my family insist, and I start on medication. I may then revisit the QRISK calculator a couple of months later. I pop in the same details but also check the box ‘on blood pressure treatment’. Even with a systolic BP of 120mmHg my risk is now 2% over 10 years.

It has gone up! That’s not very encouraging and is a good reminder of the limitations of these tools.

The QRISK information page does suggest the calculator can be used this way:

 Where patients are on antihypertensive treatment, should a pre-treatment blood pressure be used when calculating their risk?

No. QRISK®2 has been designed such that if a patient is taking antihypertensive medication then their current blood pressure on treatment can be used rather than a pre-treatment value.

I don’t have details of the algorithm used in QRISK. Presumably what is happening here is that either there is more harm than benefit from anti-hypertensives, or more likely, the sub-group ‘on blood pressure treatment’ has a lot of confounders which haven’t all been teased out. That means that being on blood pressure treatment is simply a marker of other medical issues that worsen risk when looked at on a population basis.

Risk calculators like this don’t offer some magical glimpse of the future for individuals – they are just playing with statistics and, as QRISK are at pains to point out, are best used with careful real world interpretation.

Read More

PRONUT – probiotics for starving kids

PRONUT – probiotics for starving kids

Posted on Wednesday, July 15, 2009 at northerndoctor.com

If I consider the evidence relevant to my daily practice there is no real requirement for me to appraise this paper. Positive or negative it will have no impact on my daily practice. However, the lack of negative results in the literature is often highlighted and that is one reason this trial deserves attention. I could argue (in a rather lofty and supercilious fashion) that 13 million children worldwide with severe malnutrition is an issue that merits an hour or two of anyone’s time. But mostly I just thought it looked interesting.

Probiotics irritate the hell out of me. Whenever I walk sourly down the yoghurt aisle of the supermarket I can feel the nebulous manafacturer claims wafting past. Of course, there is some evidence for probiotics but for a useful probiotic primer for the more skeptically-minded I would recommend (again) Mark Crislip’s Quackcast on the subject. I understand that the evidence of benefit doesn’t extend a huge amount beyond the confines of antibiotic-associated diarrhoea. I particularly like the Crislip interpretation of the ‘immune modulating’ effects of probiotics. He suggests that they are basically inducing chronic inflammation – not too surprising when one considers the constant stream of bacteria being dumped in the system. He also highlights the link between the known effects of chronic inflammation in the mouth and an associated increase in cardiovascular risk.

The trial’s lead author, Marko Kerac, was on the Lancet’s podcast this week (though it stops working just after 8 minutes). He seems an entirely sensible fellow though he gives an unnervingly positive summary of probiotics that I would be chary of accepting. I should point out that if you are tempted to listen to this podcast the author has an horrendous dose of high rising intonation. In addition, the podcast is abruptly truncated at just over 8 minutes.

He also demonstrates the human qualities of us all. In its perfect manifestation the randomised clinical trial can take out the unreliable subjective element of human behaviour. However, humans still have to interpret the results. In the podcast the author initially expresses disappointment at the lack of positive results and I wouldn’t criticise him for wanting to save the lives of thousands of children.

The abstract finishes with:

Subgroup analyses showed possible trends towards reduced outpatient mortality in the Synbiotic group (p=0.06). Interpretation In Malawi, Synbiotic2000 Forte did not improve severe acute malnutrition outcomes. The observation of reduced outpatient mortality might be caused by bias, confounding, or chance, but is biologically plausible, has potential for public health impact, and should be explored in future studies.

After reading the paper I am less convinced about this final conclusion. I get the impression from the paper (though it’s difficult to be certain) that the outpatient phase analysis was entirely post-hoc. The authors have honestly highlighted the potential for any post-hoc findings to be a product of pure chance. I suspect the deep disappointment in the overall negative result has trickled down into the interpretation. The final hook about outpatient mortality adds a tantalising edge to this paper. Getting a paper in the Lancet is tough – would an unrelentingly negative paper with no emphasis on this statistical wrinkle have still been published?

After all, it wasn’t the only significant difference that was found. The Synbiotic group had significantly more severe diarrhoea as an inpatient(at p=0.01 this was the strongest statistical finding in the study), more vomiting as an inpatient (p=0.05) and more cough as an inpatient (p=0.05). These don’t get a mention in the abstract. Given the caution needed when giving probiotics to an immuno-compromised population it would be entirely legitimate to have a quite different emphasis in this report.

At randomisation, groups seem to be well balanced. Minor differences at point of entry to outpatient care (lower HIV and less malnourished according to weight-for-height Z score in the Synbiotic group) raise the possibility of confounding or bias at this point.

In the Synbiotic group 42.6% were HIV positive and in the control group there were 48.5% HIV positive. The p-value for this difference between HIV rates in the two groups is P =0.08. This is not a statistically significant difference but in a total population of 795 children there is only a 4 child swing between significance and non-significance. It is feasible that the outpatient mortality differences could be simply related to this sampling bias and chance. I would not infer any deliberate deception here at all – just the simple desire to do so some good. I’m sure the research will now be repeated in an outpatient population but I’m not convinced the evidence from this trial justified it. On the other hand, maybe their dogged optimism and persistence will triumph and many lives will be saved.

In the podcast the lead author tells the tale of how the senior author, while working for UNICEF in Korea back in ’98, had made the observation that children fed at the local yoghurt factory apparently had better outcomes. He formulated the hypothesis that all those friendly bacteria were having an additive beneficial effect. It would have made a charming modern medical parable – chance favouring the prepared mind and the elegant RCT demonstrating benefit leading to millions of lives saved. No such luck this time.

Kerac, M., Bunn, J., Seal, A., Thindwa, M., Tomkins, A., Sadler, K., Bahwere, P., & Collins, S. (2009). Probiotics and prebiotics for severe acute malnutrition (PRONUT study): a double-blind efficacy randomised controlled trial in Malawi The Lancet, 374 (9684), 136-144 DOI: 10.1016/S0140-6736(09)60884-9

Read More