??fasting lipids -- not so fast

There was a recent review summarizing the data and strongly suggesting that we use nonfasting lipid testing (see nonfasting lipids jamaintmed2016 in folder ”cholesterol fasting vs nonfasting” which has a slew of articles on this in the dropbox)​. details for these recommendations:

--there was a recent consensus statement from the European Atherosclerosis Society comparing nonfasting with fasting lipids, finding that there was not more than an 8 mg/dl difference in total cholesterol, calculated LDL, non-HDL cholesterol, and up to a 26 mg/dl difference for triglycerides, with no difference in HDL, apo A or B (see nonfasting lipid differences eurheartj2016​ in folder ”cholesterol fasting vs nonfasting” ​in dropbox, or Nordestgaard BG. Eur Heart J in press with doi:10.1093/eurheartj/ehw152). they go on to suggest we "consider" a fasting sample if the non-fasting triglycerides are >440 mg/dl.

--several studies have found pretty consistently that nonfasting lipids are appropriate for general cardiovascular screening, including the impressive large study by Mora from 2008 mentioned in last blog below. also there was a meta-analysis of 68 studies which found no diminution of the association between lipids and CVD when using nonfasting measurements

--studies that have specifically included both fasting and nonfasting lipid determinations have found similar and sometimes stronger cardiovascular disease associations with nonfasting vs fasting determinations (including LDL and triglycerides)

--although many intervention studies have used fasting lipids, 3 large clinical trials with 43000 patients used nonfasting lipid determinations

--current recommendations are quite mixed: NCEP suggests fasting lipids, USPSTF recommendations are for either fasting or non-fasting cholesterol and HDL (which is the same as the 2015 NICE recommendations). Even the 2013 ACC/AHA cholesterol guidelines, which suggested that initial screening be with fasting lipids, does allow for nonfasting total and HDL cholesterol (in order to calculate the chol/HDL ratio, or the non-HDL levels)

a few comments:

--i personally have been only using nonfasting lipid measurements for the past 20+ years after i saw a study from the Boston VA hospital which showed very small differences (<10%) between fasting and nonfasting values, with the huge benefit of ordering nonfasting blood work being that i could get the lab test when i was seeing the patient (ie, the patient did not have to return in a fasting state, the test would get done more reliably, and with diabetics it seemed a mixed message to say that one needs to regularize their eating patterns as much as possible but come in tomorrow am before breakfast to get your blood test...). since then, there have been a variety of studies confirming the utility and some suggesting the superiority of nonfasting levels (see the dropbox folder).

--even the data on triglycerides, which have the most variation with eating, find that nonfasting triglycerides are a better predictor of cardiovascular events than fasting (there were several articles in a single issue of JAMA in 2007 finding this, all in the dropbox folder, including Bansal S. JAMA 2007; 298: 309). perhaps this should not be surprising, since we do seem to eat pretty regularly/perhaps too often and our bodies are more exposed to nonfasting than fasting lipids. and there are even studies showing that the role of triglycerides is even more important in those with larger areas under the curve after eating: given the same meal, some people have more profound and longer lasting increases in triglycerides, which seems to correlate with more clinical disease).

--since i have written about this before, i will add a blog from 8/4/14, which also refers to the blog from 11/14/12, and also refers to the study by Samia Mora from 2008, who write the article above. ​

Blog from 8/4/14:

over the years, i have sent out several blogs on the utility of checking nonfasting cholesterol (see blog appended below). there was a recent article in Circulation assessing the prognostic value of fasting vs nonfasting LDL levels in the Natl Health and Nutrition Survey III (see NHANES3 nonfasting LDL circ 2014​ in dropbox, or DOI: 10.1161/CIRCULATIONAHA.114.010001). this study reviewed the cross-sectional NHANES-3 database from 1988-1994 and stratified people by fasting status (fasting = at least 8 hours), followed a mean of 14 years and looked at outcomes, using propensity-matched scoring to adjust the subgroups to make them more apparently equivalent (by choosing the fasting and nonfasting cohorts with similar baseline characteristics). results:

    --4299 pairs of fasting and nonfasting people assessed (62% of the whole group had fasting samples).

    --primary outcome (all-cause mortality): no difference between groups

    --secondary outcome (cardiovascular mortality): no difference between groups. 

    --for both outcomes (not shockingly) there was an equivalent increase as LDL levels increased.

    --they also looked at fasting vs nonfasting triglycerides (which can vary by 20-30% by fasting status) and similarly found no difference in either of the outcomes by fasting status

so, further evidence that nonfasting lipid samples are fine. see my comments in the blog below. many of the major therapeutic intervention trials for lipid reduction used fasting lipids (reasons not particularly clear why that was chosen, but just for reference, several studies did not -- eg Heart Protection Study, Anglo-Scandanavian Cardiac Outcomes Trial).

Blog from 11/14/12:

Recent article (see lipids nonfasting commun study archintmed 2012 in dropbox, or doi:10.1001/archinternmed.2012.3708) which reinforces legitimacy of checking nonfasting lipids.  I stopped checking fasting lipids 15+ years ago, when i saw a small study from the VA hospital in Boston showing that there was only a small difference in total and hdl cholesterol in fasting vs nonfasting, and only a 10% or so difference in calculated LDL (added note: there was another larger study in 2012 confirming a <10% variance -- see Sidhu D, Naugler C. Arch of Intern Med 2012; 172: 1707-10).  So,i did it mostly because it was much easier for patients and more likely to actually get done. This new study, which is pretty huge, found essentially the same thing. In brief:

--all blood tests in Calgary, Canada in a 6-month period in 2011 have had the time since last meal recorded.  They looked at 200K pts, 99% outpatient blood tests. Controlled for age and sex.

--variance for total cholesterol and HDL was less than 2% (some recommendations suggest only testing chol and HDL, and these are the lipids input into the Framingham model)

--variance for calculated LDL was <10%

--variance for triglycerides was <20%

--this article did not look at clinical outcomes

There have been a litany of articles in the past several years, several on triglycerides, showing nonfasting is superior to fasting in prediction of stroke, for example. (triglycerides vary the most by fasting status, and one of the historic issues is that there is no standardized meal prior to checking non-fasting triglycerides, so how reproducible is it????). But these studies were pretty impressive. One of the physiologic reasons may be that high triglycerides after a meal may reflect decreased clearance associated with hyperinsulinemia/metabolic syndrome.  So, either the metabolic syndrome is to blame, or perhaps the increased tissue exposure to high triglycerides and low HDL (they go together here) over time (ie, increased area under the curve).  There are also other articles which look at clinical events in fasting and nonfasting measurements. A large study by Mora at the brigham have found that nonfasting chol and HDL as well as triglycerides were at least as predictive of events as fasting (see fasting nonfasting lipids circ 2008 in dropbox, or DOI: 10.1161/CIRCULATIONAHA.108.777334). More recently there has been more disparity in therecommendations from different medical societies, with some suggesting fasting and others nonfasting samples, which i think reflects the above.