The effect of fasting status on lipids, lipoproteins, and inflammatory biomarkers assessed after hospitalization for an acute coronary syndrome: Insights from PROVE IT–TIMI 22

Abstract

Background

For decades, fasting for 8 to 12 hours has been recommended for measurement of lipid profiles. The effect of fasting on low‐density lipoprotein cholesterol (LDL‐C) and triglycerides (TG) has been described in healthy cohorts and those with stable disease states. Recently, guidelines suggested that fasting may not be necessary due to its small effect on lipid measures. Little is known, however, regarding whether the impact of fasting is altered in the setting of an acute coronary syndrome (ACS).

Hypothesis

We hypothesized that the post‐ACS period would minimally effect the impact of fasting status on lipid measurements.

Methods

We evaluated the association of fasting on lipid and other biomarkers at the randomization visit, which occurred at a median of 7 days after the onset of an ACS, as well as during follow‐up, in a cohort of 4177 subjects from the Pravastatin or Atorvastatin Evaluation and Infection Therapy–Thrombolysis In Myocardial Infarction 22 (PROVE IT–TIMI 22) trial.

Results

Fasting samples were independently associated with a higher LDL‐C of 4.1 mg/dL and apolipoprotein‐B 100 of 2.6 mg/dL as well as a lower TG of 21.0 mg/dL and high‐sensitivity C‐reactive protein of 0.48 mg/dL. The relative difference was 3.8% for LDL‐C and −11.3% for TG. Fasting did not change total cholesterol, high‐density lipoprotein cholesterol, apolipoprotein A‐I, lipoprotein(a), or apolipoprotein C‐III.

Conclusions

Although fasting does impact lipid measurements, the effect on LDL‐C is small (about 4 mg/dL), both early after ACS and during follow‐up. These data provide support for recent guidelines that no longer advocate for fasting lipid samples, including in the setting of ACS.

1. INTRODUCTION

The measurement of a lipid profile upon presentation with an acute coronary syndrome (ACS) is critical, both for diagnosis of certain conditions (eg, familial hypercholesterolemia) and as a baseline to assess the treatment response during follow‐up to the initiation or titration of guideline‐recommended lipid‐lowering therapy (eg, statins, ezetimibe). It is also well known that an ACS temporarily alters lipid and certain biomarker measurements from those prior to the event, further supporting a repeat measurement upon convalescence.

Guidelines that recommend fasting prior to lipid measurement typically recommend an 8‐ to 12‐hour fast (water intake being allowed). The effect of fasting status on measurement of lipids and other biomarkers has been studied in healthy cohorts and patients with stable disease states. It is currently unknown whether the effect of fasting status is similar immediately after an ACS. In this analysis, we estimate the effect of fasting status on lipids and other biomarkers at a median of 7 days after the onset of an ACS. To assess whether this effect is modified during convalescence, we compared it with an estimation made from measurements taken during follow‐up.

2. METHODS

The study population was derived from Pravastatin or Atorvastatin Evaluation and Infection Therapy–Thrombolysis In Myocardial Infarction 22 (PROVE IT–TIMI 22), a trial performed between November 2000 and February 2004. PROVE IT–TIMI 22 randomized 4162 subjects to atorvastatin 80 mg or pravastatin 40 mg at a median of 7 days after an ACS, with a median follow‐up of 24 months.1 The primary endpoint was a composite of death from any cause, myocardial infarction, stroke, documented unstable angina requiring rehospitalization, or revascularization occurring ≥30 days post‐randomization. Descriptions of the study inclusion and exclusion criteria have been presented previously.

As part of the protocol, all lipids and biomarkers were analyzed in central laboratories. Plasma samples were collected for total cholesterol (TC), low‐density lipoprotein cholesterol (LDL‐C), high‐density lipoprotein cholesterol (HDL‐C), and triglycerides (TG) at randomization, 30 days, 4 months, 8 months, 16 months, and the end of the study. Plasma was collected for apolipoprotein B (apoB), apolipoprotein A‐I (apoA‐I), apolipoprotein C‐III (apoC‐III), lipoprotein(a) (Lp[a]), and lipoprotein‐phospholipase A₂ (Lp‐PLA₂) activity and mass at randomization, 4 months, and the end of study; and for high‐sensitivity C‐reactive protein (hs‐CRP) at randomization, 30 days, 4 months, and the end of the study. Fasting status was captured on the case‐report form as a dichotomous variable: “Had the subject fasted for at least 10 hours prior to the blood sample being collected?” Fasting was strongly recommended at follow‐up visits, but not at the randomization visit.

TC, HDL‐C, and TG were measured immediately on freshly shipped plasma samples, using an enzymatic colorimetric method and Roche Modular system (LabCorp, Raritan, NJ).2 LDL‐C was obtained by calculation (Friedewald formula) or directly measured if TG exceeded 400 mg/dL. ApoB 100 (apoB) and apoA‐I were measured using an immunoturbidimetric assay with a coefficient of variation of <2% on a Hitachi 911 analyzer (Roche Diagnostics, Indianapolis, IN) in a core laboratory (LabCorp).3 Lp‐PLA₂ activity, Lp‐PLA₂ mass, hs‐CRP, and Lp(a) were measured as reported previously.4, 5 Apo C‐III was also measured in a core laboratory.

2.1. Statistical analysis

We compared the characteristics and biomarker levels of subjects by fasting status at the time of randomization. Testing for differences by fasting status was performed using the independent t test, Wilcoxon rank‐sum test, and χ² test, as appropriate. Multivariable adjustment for differences in baseline characteristics was performed using linear regression to evaluate the independent association of fasting on biomarker levels. Age, sex, history of hypertension, history of diabetes mellitus, and history of alcohol consumption were selected a priori for the model. Additional variables were added if there was univariable association with fasting status at a P value <0.10. Finally, for each biomarker, an automated, backward selection algorithm was run to select other variables to stay in the model (significance level to stay = 0.10).

A subset of individuals had lipid profiles measured in both the fasting and nonfasting state during follow‐up. This permitted us to estimate the effect of fasting during a more “stable” period by calculating the intraindividual differences in LDL‐C and TG levels obtained across visits at 4, 8, and 16 months in this subset of individuals (see Supporting Information, Tables S2 and S3, in the online version of this article). We restricted the analysis to subjects recorded as having good compliance (>80% compliant with study drug) at each visit so as to limit the impact of statin treatment on LDL‐C and TG measurements. Using time‐point comparisons (eg, those fasting at month 4 vs nonfasting at month 8), we first calculated the mean of the fasting values obtained at one timepoint and the mean of the nonfasting values obtained at the other timepoint. We then calculated the difference between these means. For example, if the mean LDL‐C was 85mg/dL for all the subjects fasting at month 4 and 80mg/dL for these same subjects when nonfasting at month 8, the mean difference would be 5 mg/dL. We then completed the analysis by calculating a weighted mean of all the mean differences calculated across all the time‐point comparisons for both LDL‐C and TG. The difference due to fasting was reported as absolute and relative change from the nonfasting state.

3. RESULTS

Fasting status was captured at baseline in 4137 subjects (99.4% of the total population). Because fasting was not required, almost one‐half (46.8%) of the subjects were not fasting. Review of the baseline demographics and clinical data revealed a similar distribution of characteristics (Table 1). Unadjusted mean biomarker levels by fasting status are displayed in Table 2. Significant differences were found for LDL‐C, apoB, TG, and hs‐CRP. Compared with nonfasting, mean fasting levels were 4.6 mg/dL (P < 0.001) higher for LDL‐C and 3.1 mg/dL (P < 0.001) higher for apoB, and 19.2 mg/dL (P < 0.001) lower for TG and 0.5 mg/dL (P < 0.001) lower for hs‐CRP. In terms of relative change, fasting levels were 4.3% and 3.1% higher for LDL‐C and apoB, respectively, and 10.3% and 19.2% lower for TG and hs‐CRP, respectively.

Table 1.

Baseline characteristics by fasting status

Characteristics	Total, N = 4137	Nonfasting, n = 1938	Fasting, n = 2199	P Value
Demographics
Age, y	58.2 (11.2)	58.4 (11.0)	58.1 (11.4)	0.398
Male sex	78.2	77.1	79.1	0.113
Race
White	90.8	90.5	91.0	0.356
Black	4.5	4.5	4.5	0.356
Hispanic	3.1	3.5	2.7	0.356
Asian	1.1	0.9	1.3	0.356
Other	0.5	0.6	0.5	0.356
Medical history
Prior MI	18.6	19.4	17.8	0.194
PCI before index event	15.5	14.5	16.4	0.089
CABG before index event	11.0	10.2	11.6	0.156
PAD	5.8	5.0	6.6	0.029
DM	17.7	17.4	17.8	0.745
HTN	50.3	50.0	50.5	0.759
Current smoker	36.7	36.7	36.6	0.930
Family history of CAD	52.8	52.3	53.2	0.542
Moderate/heavy alcohol consumption	39.3	39.2	39.4	0.973
Medications prior to index eventa
Statin	60.7	60.5	60.9	0.796
Insulin	9.3	9.0	9.6	0.493
Sulfonylurea	7.1	7.4	6.8	0.449
Index event and treatment
UA	29.4	29.1	29.6	0.889
MI without ST‐segment elevation	36.2	36.1	36.3	0.889
MI with ST‐segment elevation	34.4	34.8	34.1	0.889
PCI for treatment of index event	68.8	66.5	70.9	0.002
Assessments at randomization
BMI, kg/m2	29.5 (5.7)	29.4 (5.6)	29.7 (5.9)	0.093
SBP, mm Hg	123.8 (18.1)	123.2 (17.9)	124.2 (18.3)	0.067
DBP, mm Hg	72.2 (11.5)	71.2 (11.4)	73.1 (11.5)	<0.001
Heart rate, bpm	69.3 (11.8)	69.8 (11.7)	68.9 (11.9)	0.017
Killip class II–IV	3.8	4.2	3.6	0.324

Abbreviations: BMI, body mass index; CABG, coronary artery bypass grafting; CAD, coronary artery disease; DBP, diastolic blood pressure; DM, diabetes mellitus; HTN, hypertension; MI, myocardial infarction; PAD, peripheral arterial disease; PCI, percutaneous coronary intervention; SBP, systolic blood pressure; SD, standard deviation; UA, unstable angina.

Data are presented as percentages or as mean (SD). Testing for differences by fasting status was performed using the independent t test, Wilcoxon rank‐sum test, and χ² test, as appropriate.

^aDefined as use within 2 weeks prior to the index event or for the management of the index event.

Table 2.

Comparison of baseline biomarker levels by fasting status

Biomarkers	Total, N = 4137	Nonfasting, n = 1938	Fasting, n = 2199	P Value
TC, mg/dL	182.7 (34.5)	182.2 (34.4)	183.1 (34.6)	0.390
LDL‐C, mg/dL	109.2 (29.7)	106.8 (28.6)	111.4 (30.4)	<0.001
Non–HDL‐C, mg/dL	142.6 (34.1)	141.9 (33.8)	143.3 (34.4)	0.192
HDL‐C, mg/dL	40.0 (10.9)	40.3 (11.1)	39.8 (10.7)	0.171
TG, mg/dL	176.2 (103.1)	186.4 (113.2)	167.2 (92.4)	<0.001
ApoB, mg/dL	102.7 (22.4)	101.1 (21.9)	104.2 (22.8)	<0.001
ApoA‐I, mg/dL	123.0 (24.0)	123.5 (23.7)	122.7 (24.2)	0.310
Lp(a), mg/dL	24.9 (24.3)	24.9 (24.9)	24.8 (23.7)	0.939
ApoC‐III, mg/dL	10.3 (4.0)	10.3 (3.8)	10.3 (4.1)	0.967
Lp‐PLA2 activity, nmol/min/mL	40.9 (12.2)	40.8 (12.3)	41.0 (12.1)	0.647
Lp‐PLA2 mass, ng/mL	180.6 (72.1)	180.2 (72.2)	181.0 (72.1)	0.737
hs‐CRP, mg/dL	2.3 (3.0)	2.6 (3.2)	2.1 (2.9)	<0.001

Abbreviations: ApoA‐I, apolipoprotein A‐I; ApoB, apolipoprotein B; ApoC‐III, apolipoprotein C‐III; CI, confidence interval; HDL‐C, high‐density lipoprotein cholesterol; hs‐CRP, high‐sensitivity C‐reactive protein; LDL‐C, low‐density lipoprotein cholesterol; Lp(a), lipoprotein(a); Lp‐PLA₂, lipoprotein‐phospholipase A₂; SD, standard deviation; TC, total cholesterol; TG, triglycerides.

Data are presented as mean (SD). Testing for differences by fasting status was performed using the independent t test.

Review of unadjusted median biomarker levels by fasting status again revealed significant differences for LDL‐C, apoB, TG, and hs‐CRP. Fasting levels were 5 mg/dL (P < 0.0001) higher for LDL‐C and 3 mg/dL (P = 0.0001) higher for apoB, and 16 mg/dL (P < 0.0001) lower for TG and 0.5 mg/dL (P < 0.0001) lower for hs‐CRP (see Supporting Information, Table S1, in the online version of this article). In terms of relative change, fasting levels were 4.8% and 3.0% higher for LDL‐C and apoB, respectively, and 9.8% and 33.3% lower for TG and hs‐CRP, respectively.

Because fasting status might be influenced by a subject's demographic and clinical characteristics, multivariate adjustment using a general linear regression model was used to quantitate the independent association of fasting on mean biomarker levels (Table 3). The model revealed similar results to the unadjusted analyses. Compared with nonfasting, mean fasting levels were 4.13 mg/dL (P < 0.0001) higher for LDL‐C and 2.56 mg/dL (P = 0.0005) higher for apoB, and 21.04 mg/dL (P < 0.0001) lower for TG and 0.48 mg/dL (P < 0.0001) lower for hs‐CRP.

Table 3.

Impact of fasting compared with nonfasting after multivariable adjustment

Biomarker of Interest	β (95% CI)	P Value
Non‐HDL‐C, mg/dL	0.44 (−1.66 to 2.54)	0.6824
Non‐HDL‐C, mg/dL	0.57 (−1.50 to 2.64)	0.5878
LDL‐C, mg/dL	4.13 (2.31 to 5.94)	<0.0001
TG, mg/dL	−21.04 (−27.18 to −14.90)	<0.0001
HDL‐C, mg/dL	−0.13 (−0.77 to 0.50)	0.6823
ApoB, mg/dL	2.56 (1.12 to 4.00)	0.0005
ApoA‐I, mg/dL	−0.18 (−1.63 to 1.28)	0.8133
ApoC‐III, mg/dL	−0.03 (−0.31 to 0.25)	0.8419
Lp‐PLA2 activity, nmol/min/mL	−0.06 (−0.86 to 0.74)	0.8818
Lp‐PLA2 mass, ng/mL	1.92 (−2.90 to 6.74)	0.4342
hs‐CRP, mg/dL	−0.48 (−0.68 to −0.28)	<0.0001
Lp(a), mg/dL	0.12 (−1.51 to 1.76)	0.8837

Abbreviations: ApoA‐I, apolipoprotein A‐I; ApoB, apolipoprotein B; ApoC‐III, apolipoprotein C‐III; BMI, body mass index; CI, confidence interval; DM, diabetes mellitus; HDL‐C, high‐density lipoprotein cholesterol; hs‐CRP, high‐sensitivity C‐reactive protein; LDL‐C, low‐density lipoprotein cholesterol; Lp(a), lipoprotein(a); Lp‐PLA₂, lipoprotein‐phospholipase A₂; MI, myocardial infarction; PAD, peripheral arterial disease; PCI, percutaneous coronary intervention; TC, total cholesterol; TG, triglycerides.

Multivariate adjustment included age, sex, history of DM, current smoking, history of alcohol consumption, BMI, history of prior MI, PCI for treatment of index event, and history of PAD.

Lipid profiles were also measured throughout follow‐up. When we estimated the effect of fasting on LDL‐C and TG assessed during follow‐up visits the impact of fasting status appeared to be relatively stable (Table 4). During follow‐up, fasting mean LDL‐C was 4.3 mg/dL higher and TG was 26.4 mg/dL lower. The relative difference was a 5.4% increase in LDL‐C and a 14.5% decrease in TG.

Table 4.

Impact of fasting status on LDL‐C and TG at baseline vs follow‐up

	Baseline	Follow‐up
Nonfasting LDL‐C, mg/dL	106.8a	80.0b
Fasting impact, mg/dL	4.1c	4.3d
Fasting impact, %	3.8%	5.4%
Nonfasting TG, mg/dL	186.4a	182.5b
Fasting impact, mg/dL	−21.0c	−26.4d
Fasting impact, %	−11.3%	−14.5%

Abbreviations: LDL‐C, low‐density lipoprotein cholesterol; TG, triglycerides.

During follow‐up, 652 intraindividual comparisons were available and used for LDL‐C calculations; 656 were available and used for TG comparisons.

Additional Supporting Information may be found online in the supporting information tab for this article.

Data used for follow‐up measurements are provided in Supporting Tables 2 and 3.

^aMean nonfasting LDL‐C and TG from Table 2.

^bWeighted mean of the nonfasting mean levels for each comparison.

^cEffect of fasting from Table 3.

^dWeighted mean of the difference between fasting vs nonfasting means for each comparison.

4. DISCUSSION

When we analyzed the effect of fasting status on lipid and other biomarkers in those with a recent ACS, fasting was independently associated with a higher level of LDL‐C and apoB compared with the nonfasting state. This small absolute increase in LDL‐C of approximately 4 mg/dL was similar at 1 week after ACS and during further follow‐up, as was the relative increase. TG levels were expectedly lower by approximately 20 mg/dL, the biggest absolute difference for any lipid measurement. We found that hs‐CRP was also lower in the fasting state, whereas TC, HDL‐C, apoA‐I, Lp(a), apoC‐III, Lp‐PLA₂ activity, and Lp‐PLA₂ mass did not change significantly. Most guidelines recommend lipid measurement before starting statin therapy; thus, these results have importance because statin therapy should be initiated (or optimized if already prescribed) at the time of ACS. In addition, assessment of lipid level achievement in statin‐treated patients may be useful on presentation with an ACS to assess whether additional lipid reduction may be needed, either through improved adherence to diet or statins or initiation of other evidence‐based lipid‐lowering medications such as ezetimibe or proprotein convertase subtilisin/kexin type 9 inhibitors.

In stable patients after normal dietary intake, fasting primarily affects TG levels, with only small changes in LDL‐C of approximately 4% to 7%.6, 7, 8, 9 It should be noted that postprandial changes in TG and LDL‐C are dependent upon the quantity of food consumed, liquids consumed, the proportion of each macronutrient (eg, fat) in the meal, and the time since consumption. The postprandial TG increase is especially important, as LDL‐C calculation using the Friedewald Equation (LDL‐C = TC − HDL‐C − [TG/5]) is directly dependent upon the TG measurement (because TC and HDL‐C are relatively stable10, 11). For example, Rifai et al. showed in a well‐controlled study of healthy, young male volunteers that TG can increase up to 150% after a high‐fat meal, with an LDL‐C decrease up to 37%.12 ApoB and apoA‐I often do not change or have a small variable postprandial response. Of note, nonfasting status does not compromise the predictive value of lipids.13, 14 Increases in CRP15 and other inflammatory markers16 in the nonfasting state have been previously reported.

The presence of stable coronary disease has been shown to not affect the postprandial response.17 It has been reported, however, that although lipid levels are relatively stable during the immediate post‐ACS period,18 these levels are altered from the pre‐ACS state19 due to reasons such as the acute phase response, reduction in several HDL‐C regulatory proteins, stress‐induced myocardial injury and necrosis facilitating adrenergic‐mediated adipocyte lipolysis, and lifestyle changes.20 Until now, whether the factors that alter lipids during an ACS also impact the effect of fasting status on these measurements has been unknown.

The data to support a requirement for routine fasting prior to performing a lipid profile are minimal. The rationale for fasting includes (1) the dose‐dependent postprandial changes in lipids due to fat, carbohydrate, and fluid consumption; (2) the clinically significant effects of increased TG (400 mg/dL) on the calculation of LDL‐C when using the Friedewald equation; and (3) the use of fasting samples for lipid measurement in many clinical trials and epidemiological studies. The most obvious advantage of not requiring fasting is that it makes lipid assessment less burdensome for providers and patients. In addition, fasting requirements may stress laboratory resources by causing a large bolus of morning tests. If fasting is not required, it may still be reasonable to provide patients with specific recommendations (eg, avoid large, high‐fat meals), especially if LDL‐C and TG will be measured. Laboratory alerts can help identify patients who may benefit from repeat measurements in the fasting state (eg, TG >400 mg/dL).

Guidelines differ in their recommendations for fasting. The 2013 American College of Cardiology/American Heart Association (ACC/AHA) guidelines prefer a fasting lipid panel before statin initiation to calculate LDL‐C and to improve the diagnosis of certain conditions (eg, familial hypercholesterolemia, hypertriglyceridemia, metabolic syndrome).21 On follow‐up assessment, fasting is also preferred to evaluate treatment response and assess for adherence. In contrast, a 2016 consensus statement from the European Atherosclerosis Society (EAS) and the European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) concluded that the net benefits favor not recommending fasting prior to a lipid profile,22 even though the current European Society of Cardiology (ESC) primary treatment target is LDL‐C.23 Other guidelines that now focus on non–HDL‐C as the primary treatment (as opposed to LDL‐C) do not recommend fasting prior to initiation of therapy or during follow‐up.24, 25, 26 The current study provides reassurance to these recommendations by verifying that TC and HDL‐C do not significantly change by fasting status after the onset of an ACS.

4.1. Study limitations

A potential limitation to our study was that fasting status among subjects was nonrandomized at baseline. We believe, however, that this had a negligible effect for the following reasons: (1) fasting status was not required at baseline, and therefore laboratory collection was driven by factors other than the subjects' baseline characteristics (eg, convenience for the research team); (2) approximately 50% of subjects were nonfasting at baseline, with few differences in baseline characteristics; and (3) multivariate adjustment did not significantly alter the effect of fasting. Another potential limitation was the inability to capture details on time from last meal or the nutritive content of the meal.

5. CONCLUSION

We found that fasting was independently associated with slightly higher LDL‐C and apoB levels at a median of 7 days after onset of an ACS. Fasting was also independently associated with lower TG and hs‐CRP. The effect of fasting on LDL‐C and TG at baseline was similar to that found during follow‐up. TC and HDL‐C were not significantly associated with fasting status. These findings provide reassurance to clinicians measuring these biomarkers in their ACS patients and support those guidelines that do not require routine fasting.

Conflicts of interest

D. Steen receives consulting fees from Sanofi* and Amgen. C. Cannon receives consulting fees from Alnylam, Amgen, Arisaph, AstraZeneca, Boehringer Ingelheim, Bristol‐Myers Squibb, GlaxoSmithKline, Kowa, LipimetiX,* Merck, Pfizer, Regeneron,* Sanofi,* and Takeda (*denotes >$10 000). C. Cannon receives research grants (all >$10 000) from Amgen, Arisaph, Boehringer Ingelheim, Bristol‐Myers Squibb, Daiichi‐Sankyo, Janssen, Merck, and Takeda. The authors declare no other potential conflicts of interest.

Supporting information

Supplementary Table 1: Comparison of Baseline Biomarker Levels by Fasting Status

Supplementary Table 2: Fasting and Nonfasting LDL‐C Measurements during Follow‐up

Supplementary Table 3: Fasting and Nonfasting Triglyceride Measurements during Follow‐up

Steen DL, Umez‐Eronini AA, Guo J, Khan N, Cannon CP. The effect of fasting status on lipids, lipoproteins, and inflammatory biomarkers assessed after hospitalization for an acute coronary syndrome: Insights from PROVE IT–TIMI 22. Clin Cardiol. 2018;41:68–73. 10.1002/clc.22851