NPC Archive Item: Cholesterol measurements in the first few years of statin treatment may mislead

NOTE – This is an archive post from the NPC and has not been updated since first publication. Therefore, some hyperlinks may no longer be working.
MeReC Rapid Review NPC Logo

Glasziou P et al. Monitoring cholesterol levels: measurement error or true change?  Ann Intern Med 2008; 148: 656–61

A single cholesterol level reading may well under- or over-estimate a person’s true average cholesterol level by up to 14%. In a treatment-adherent patient in whom initiation of statin therapy lowers cholesterol below the audit standard of 5 mmol/L, subsequent test results in the first few years of treatment that suggest the patient’s cholesterol is greater than 5 mmol/L are more likely to be incorrect than correct.

Health professionals should be wary of increasing a patient’s lipid-lowering treatment on the basis of a single cholesterol test if they are reasonably confident that the patient is taking the medication as prescribed, at least in the first few years of the patient’s treatment.

What is the background to this?
The Quality and Outcomes Framework (QOF), which determines how most general practices in the UK are paid, rewards practices for measuring the cholesterol of people with coronary heart disease, stroke or transient ischaemic attack (TIA), or diabetes, every year.  It also rewards practices if the recorded total cholesterol of a large proportion of these people is less than 5 mmol/L.  The recently-published NICE guidance on lipid modification also recommends 5 mmol/L as an audit standard for total cholesterol in people with established cardiovascular disease (CVD) but does not say how often this should be measured.  No lipid level target or audit standard is given for people without established CVD (see previous blog).  The NICE guidance on management of type 2 diabetes recommends annual cholesterol monitoring and a target total cholesterol of less than 4 mmol/L (see previous blog).

What is this study about?
The authors looked at data from the Long-term Intervention with Pravastatin in Ischaemic Disease (LIPID) study to see how lipid levels vary between patients and over time. The LIPID study was an RCT of 9,014 patients with established coronary heart disease randomised to receive pravastatin 40mg/day or placebo for an average of 6 years. Participants in both arms had their cholesterol levels measured twice during an eight week placebo run-in period, then at baseline, at 6 months, at 12 months and then annually for five years.  All measurements were made at the same laboratory.

What does this study claim?
We would expect that over time the average cholesterol level of a group of patients might change.  In fact, in LIPID, both groups showed a small, average increase in total cholesterol: from 4.49 mmol/l at 6 months to 4.63 mmol/L at 5 years in the pravastatin group and from 5.67 mmol/L to 5.73 mmol/L in the placebo group.  However, this varied from one individual to another.

In addition, for individual patients there will be short term variations around a stable, but gradually moving, average – due to week-to-week biological variation and analytical variability.  This study found that the 95% confidence intervals on a single cholesterol level measurement were ±14% of the person’s true average cholesterol. In practical terms, this means that for 95 people out of 100 with a single cholesterol measurement of 5.0 mmol/L, their true average cholesterol level will be between 4.3 mmol/L and 5.7 mmol/L.

So, over time, we must consider not only a person’s true, gradually shifting, average cholesterol, but also short term variations in their level.  The authors looked at how good a cholesterol test done at different time points would be at detecting when person’s average cholesterol truly had exceeded 5 mmol/L.  This is expressed as the ratio of false positives (false alarms) to true positives in treatment adherent patients

Initial true cholesterol level
4.5 mmol/L
Number of false alarms for every correct level >5 mmol/L
Year 1 16
Year 3 1.6
Year 5 1

Initial true cholesterol level
4.0 mmol/L
Number of false alarms for every correct level >5 mmol/L
Year 1 >1000
Year 3 10
Year 5 3

So what

We can see that even with quite modest cholesterol levels, in the first few years of treatment, cholesterol measurements are more likely to mislead with regard to progress against audit standards than to identify those who truly have not attained the target.  This is particularly so if the initial cholesterol level on treatment is low.  Even after five years of treatment in patients with an initial cholesterol of 4.5 mmol/L, a test result greater than 5 mmol/L is just as likely to be incorrect as correct.

The authors suggest that after a few years of treatment, a single high or borderline high level might warrant re-testing within a few weeks to obtain a better estimate of a person’s true cholesterol level.

Of course there are limitations to these findings.  Firstly, the true initial level can only be attained by repeated testing, which seems impracticable in practice.  In addition, adherence to treatment in LIPID was high – much higher than is likely in UK general practice.  However, testing cholesterol levels to assess adherence to treatment is not likely to be a useful strategy:  an observed level greater than 5 mmol/L could be due either to poor adherence or sampling variation.

NICE lipid guidance to give patients an evidence based dose of an evidence-based statin (simvastatin 40mg/day), and not worry too much about slavish adherence to cholesterol levels, seems sensible in the light of this paper.

Please comment on this blog in the NPC discussion rooms, or using our feedback form