Unadjusted seasonal variation in serum lipid levels by sex: total cholesterol (A), low-density lipoprotein (LDL) cholesterol (B), high-density lipoprotein (HDL) cholesterol (C), and triglycerides (D). To convert total, LDL, and HDL cholesterol from milligrams per deciliter to millimoles per liter, multiply by 0.0259; triglycerides from milligrams per deciliter to millimoles per liter, multiply by 0.0113.
Ira S. Ockene, David E. Chiriboga, Edward J. Stanek III, Morton G. Harmatz, Robert Nicolosi, Gordon Saperia, Arnold D. Well, Patty Freedson, Philip A. Merriam, George Reed, Yunsheng Ma, Charles E. Matthews, James R. Hebert. Seasonal Variation in Serum Cholesterol LevelsTreatment Implications and Possible Mechanisms. Arch Intern Med. 2004;164(8):863–870. doi:10.1001/archinte.164.8.863
A variety of studies have noted seasonal variation in blood lipid levels. Although the mechanism for this phenomenon is not clear, such variation could result in larger numbers of people being diagnosed as having hypercholesterolemia during the winter.
We conducted a longitudinal study of seasonal variation in lipid levels in 517 healthy volunteers from a health maintenance organization serving central Massachusetts. Data collected during a 12-month period for each individual included baseline demographics and quarterly anthropometric, blood lipid, dietary, physical activity, light exposure, and behavioral information. Data were analyzed using sinusoidal regression modeling techniques.
The average total cholesterol level was 222 mg/dL (5.75 mmol/L) in men and 213 mg/dL (5.52 mmol/L) in women. Amplitude of seasonal variation was 3.9 mg/dL (0.10 mmol/L) in men, with a peak in December, and 5.4 mg/dL (0.14 mmol/L) in women, with a peak in January. Seasonal amplitude was greater in hypercholesterolemic participants. Seasonal changes in plasma volume explained a substantial proportion of the observed variation. Overall, 22% more participants had total cholesterol levels of 240 mg/dL or greater (≥6.22 mmol/L) in the winter than in the summer.
This study confirms seasonal variation in blood lipid levels and suggests greater amplitude in seasonal variability in women and hypercholesterolemic individuals, with changes in plasma volume accounting for much of the variation. A relative plasma hypervolemia during the summer seems to be linked to increases in temperature and/or physical activity. These findings have implications for lipid screening guidelines. Further research is needed to better understand the effects of a relative winter hemoconcentration.
Small longitudinal and larger cross-sectional studies1- 6 suggest that cholesterol levels are higher in the fall and winter than in the spring and summer. Studies7 reporting the most striking effects suggest that in areas of extreme seasonal climatic variation, such as Finland, there may be as much as a 100-mg/dL (2.59-mmol/L) seasonal variation in serum cholesterol levels. A longitudinal study8 of plasma lipid levels in the United States reported a mean seasonal change in plasma total cholesterol concentration of 7.4 mg/dL (0.19 mmol/L). However, the study was restricted to hyperlipidemic men. A cross-sectional study6 of seasonal variation in plasma lipid levels, using the US National Cholesterol Education Program guidelines for hyperlipidemia of 240 mg/dL or greater (≥6.22 mmol/L), reported that 25.4% of men were at or above this level in the winter, whereas only 13.5% met this cutoff point in the summer.
Despite accumulating evidence of seasonal variation in blood lipid levels, such variation is not considered in the National Cholesterol Education Program guidelines.9- 11 Seasonal variation is also rarely taken into account in the clinical management of hyperlipidemia, although the seasonal differences described in the previous paragraph would result in large differentials in the frequency of patients being labeled as hypercholesterolemic at different times of the year. The described seasonal variation also would lead to a lower percentage of patients reaching goal cholesterol levels if treatment were initiated in the summer and repeated measurement carried out in the winter.
To resolve these issues, we conducted a longitudinal study of seasonal variation in lipid levels in healthy volunteers. We also measured all of the factors thought likely to be related to such variation.
Participants in the SEASONS (Seasonal Variation in Blood Lipids) Study, described in detail elsewhere,12 were recruited primarily from the Fallon Healthcare System, a health maintenance organization serving central Massachusetts. Additional individuals of Hispanic descent residing in Worcester County, Massachusetts, were recruited in an attempt to increase the diversity of the study population. Individuals who were residents of Worcester County, who were aged 20 to 70 years, and who had telephone service were eligible to participate in the study. Eligible individuals were not taking cholesterol-lowering medications and were not actively participating in lipid-lowering or weight-control diets, did not have possible causes of secondary hyperlipidemia, and were free of chronic life-threatening illness. Participants were recruited between December 1, 1994, and February 28, 1997, and enrollment occurred throughout the calendar year. The institutional review boards of the Fallon Healthcare System and the University of Massachusetts Medical School approved all participant recruitment and data collection procedures. Each participant signed an approved informed consent form before entering the study.
Demographic data and health information were collected by self-administered questionnaire. Meteorologic data were obtained from the National Weather Service. Anthropometric data, including weight, height, and waist and hip circumferences, and fasting blood lipid values were assessed at baseline and then every 3 months, within a 3-week window on either side of the individual's quarterly appointment date, to the 1-year anniversary point (a total of 5 assessments).
Telephone-administered 24-hour recalls were used as the dietary assessment method, with interviews conducted every 3 months. At each quarterly data collection point, three 24-hour recalls were conducted on randomly selected days (including 2 weekdays and 1 weekend day). These recalls included the collection of physical activity and light-exposure data. All dietary data for the 24-hour recalls were entered into and analyzed using nutrient calculation software (Nutrition Data System NDS DOS versions 2.6 to 2.9; Nutrition Coordinating Center, University of Minnesota, St Paul).13
Estimates of physical activity energy expenditure (metabolic equivalent [MET] hours per day) were calculated from the 24-hour recalls for household, occupational, and leisure activities using methods described by Ainsworth and colleagues.14 One MET hour per day is approximately equivalent to 1 kcal/kg per day. The validity of this method has been evaluated using the Baecke Physical Activity Questionnaire and whole-body accelerometers, and it has been found to be comparable to these standard methods.15,16
Light exposure was determined by having participants recall information about outdoor and indoor exposure to natural and artificial lighting. The validity of the light-exposure questionnaire was confirmed in a group of volunteers by comparing the 24-hour recall estimates with those obtained from a subset of 80 participants who wore a light-recording device (Actillume; Ambulatory Monitoring Inc, Ardsley, NY) for the week before the 24-hour recalls.16
Analyses of cross-sectional and longitudinal data were conducted. Cross-sectional comparisons of baseline response by sex were made using 2-sample t tests (for continuous variables) and χ2 tests (for categorical variables). Longitudinal data were analyzed using mixed models (SAS Proc Mixed; SAS Institute Inc, Cary, NC) with restricted maximum likelihood methods.17 The primary outcome variables for the longitudinal analyses were the blood lipid measures.
Two methods were used to evaluate seasonal effects in the mixed models. First, the date the blood sample was obtained was used to classify each lipid measure into a season of the year, using the "light season definition," centered at the equinoxes, to maximize variation in light exposure (winter: November 6 to February 4; spring: February 5 to May 6; summer: May 7 to August 5; and fall: August 6 to November 5). Season was used as a fixed effect in the mixed-model analyses, and estimates of change in cholesterol concentration between seasons were constructed. Second, the date the blood sample was obtained was used to define sine and cosine coefficients for a sine-shaped seasonal model that assumed a period of 365 days.18 Estimates of fixed-effect regression coefficients for the sine and cosine terms in the mixed model were transformed to estimate the amplitude and phase of the seasonal effects. A first-order Taylor series expansion was used to construct estimates of the variance of the amplitude and phase from the variance estimates of the sine and cosine coefficients (for additional information, see http://www-unix.oit.umass.edu/~seasons/se35.pdf..
Sex-specific mixed models were fit solely with seasonal effects, using participants as random effects. Subsequent models were fit that controlled for various time-dependent covariates separately, including body mass index (calculated as weight in kilograms divided by the square of height in meters), percentage of calories from saturated fat, physical activity (total and leisure), total light exposure, hemoglobin level, relative plasma volume19 (calculated from a participant-specific mean hemoglobin value, assuming that the hemoglobin level remains constant during the year in healthy individuals), and age. We evaluated the impact of including the covariates by assessing the extent to which the estimated amplitude of the seasonal effect changed when the covariates were considered; percentage change in seasonal amplitude was used for this purpose and was obtained by comparing amplitude from the model with covariates with that of the model without covariates. Similar analyses also were conducted for participants in the upper or lower quartile distribution of blood lipid levels.
Before blood samples were obtained from participants in the SEASONS Study, up to three 24-hour recall measures were obtained of dietary intake, physical activity, and light exposure in each quarter. Because only 1 lipid measurement was made during the 3-month interval, we first estimated average covariate values for each participant during the interval. Rather than using simple 3-day averages as the estimates, we used best linear unbiased predictors from mixed models to estimate longer-term (28-day) average values,20 assuming that the longer-term average covariate values had stronger relationships with lipids. Thus, the best linear unbiased predictor estimates of dietary intake, physical activity, and light exposure were used in analyses of lipid levels. Finally, the percentage of the study population, by sex, with total cholesterol levels of 240 mg/dL or greater (≥6.22 mmol/L) in winter and summer was obtained using the participant average at each season.
A total of 1254 (25%) of the 5000 individuals initially contacted by telephone met the study entry criteria and agreed to attend an initial clinic visit. Of these, 641 individuals attended the first visit and completed the baseline interview. Sixty-one individuals (10%) dropped out of the study after a single blood lipid measure, 39 (6%) completed only 2 lipid measures, and 24 (4%) completed only 3 such measures. A participant was operationally defined as having completed the protocol if at least 4 of the potential 5 lipid measures were obtained. Using this definition, 517 individuals (81% of the enrolled population) completed the study. Comparison of these 517 participants with those who did not complete the study suggests that younger and nonwhite enrollees were more likely to drop out (data not shown). There were no statistically significant differences in occupation categories, formal education level, or current smoking status between the 2 groups.
The final analysis for this study used a subset of 476 study participants who had at least 4 quarterly cholesterol measures, of which at least 1 was obtained during the summer and at least 1 during the winter, using the light season definition.
Baseline data suggest that white, married, well-educated, and overweight people characterize the group as a whole (Table 1). Men had more years of formal education than women. Women were less likely to be overweight than men. More than 50% of the study participants worked in service industries or in white-collar occupations. Smokers composed approximately 16% of the study population, with no statistically significant sex differences.
The average total cholesterol level was 222 mg/dL (5.75 mmol/L) among men and 213 mg/dL (5.52 mmol/L) among women. Interassay and intra-assay coefficients of variation for total cholesterol were ±1.4% and ±1.5% and for low-density lipoprotein (LDL) cholesterol were ±1.7% and ±2.4%, respectively (in compliance with Centers for Disease Control and Prevention–accepted ranges). Unadjusted averages for each of the 4 light seasons, by sex, are given in Table 2. The results suggest that women are more likely to display seasonal variation in blood lipid levels than men. A winter-summer comparison indicated that in women, total, LDL, and high-density lipoprotein (HDL) cholesterol levels were statistically significantly higher in winter, whereas only HDL cholesterol levels showed a similar pattern in men.
Anthropometric and dietary data did not demonstrate statistically significant seasonal changes, whereas physical activity and light exposure did (Table 2). As a whole, the study group had a relatively low level of physical activity. Seasonal variation in physical activity was determined primarily by changes in nonoccupational (or leisure and household) activities in men and women. In men, there was a secondary peak in leisure activities during the winter, a finding described in previous work with these data.21
Hematologic data, particularly hemoglobin and relative plasma volume, showed statistically significant seasonal variation, suggesting a hemodilution effect during the summer and the converse during the winter. There was a positive correlation between relative blood volume and both physical activity and exposure to ambient temperature (data not shown).
Results of the sinusoidal regression analyses show that the mean seasonal amplitude for the total serum cholesterol level was 3.9 mg/dL (0.10 mmol/L) (95% confidence interval [CI], 1.2-6.6 mg/dL [0.03-0.17 mmol/L]), with a peak on December 5 (95% CI, ±41 days), in men and 5.4 mg/dL (0.14 mmol/L) (95% CI, 2.5-8.3 mg/dL [0.06-0.21 mmol/L]), with a peak on January 3 (95% CI, ±31 days), in women (Table 3 and Figure 1). Seasonal amplitude seemed to be of greater magnitude among participants in the upper quartile of the total cholesterol distribution (6.0 mg/dL [0.16 mmol/L]; 95% CI, 1.3-10.7 mg/dL [0.03-0.28 mmol/L] for men and 12.0 mg/dL [0.31 mmol/L]; 95% CI, 5.6-18.4 mg/dL [0.14-0.48 mmol/L] for women), but it was not statistically significantly different from the group as a whole.
Seasonal variation in LDL cholesterol levels followed a pattern similar to that observed for total cholesterol in men and women but was of lower magnitude and was statistically significant only in women (Table 3). Likewise, HDL cholesterol levels peaked in the winter, and the amplitude was statistically significant in both sexes. Seasonal variation in triglyceride levels was also present, but the amplitude was statistically significant only in women, and the peak occurred during the fall.
Results of the regression analyses suggest that changes in plasma volume seem to explain a substantial proportion of the observed seasonal variation in total and LDL cholesterol levels. After controlling for relative plasma volume, the observed seasonal variation was no longer statistically significant, except for total serum cholesterol level in women. Amplitude of seasonal change in HDL cholesterol levels, however, did not seem to be as affected by changes in plasma volume as was that of total and LDL cholesterol levels.
We observed that during the winter compared with the summer, there was a 7.0% relative increase in the number of men with cholesterol levels of 240 mg/dL or greater (≥6.22 mmol/L) (from 29.9% to 32.0%) (P = .42) and a 47.4% relative increase in women (from 19.0% to 28.0%) (P<.001) (Table 4). Overall, 22.2% more people had total cholesterol levels of 240 mg/dL or greater (≥6.22 mmol/L) in the winter than in the summer (P = .003). The corresponding values using LDL cholesterol levels of 160 mg/dL or greater (≥4.14 mmol/L) showed no increase during the winter compared with the summer in men and a 29.4% increase in women (from 22.4% to 29.0%).
These data demonstrate seasonal variation in blood lipid levels. The analysis of the longitudinal data fits a sinusoidal curve, with a peak in the winter and a trough in the summer. These findings are similar to those reported elsewhere in the literature,2,5,6,8,22- 29 although with lower seasonal amplitude. We also found the amplitude of seasonal variability to be greater in women than in men and in individuals in the upper quartile of the total cholesterol distribution vs those in lower quartile.
Previously, we observed seasonal variation in various covariates, including physical activity,21 light exposure, saturated fat intake, and mood.30 None of these covariates seemed to explain a substantial proportion of the observed seasonal variation in blood lipid levels.
After controlling for relative plasma volume, the amplitude of seasonal variation in blood lipid levels decreased statistically significantly, with a greater effect attributable to relative plasma volume in men. During the summer, the increase in environmental temperature or physical activity, or most likely a combination of the two, could be contributing to a hemodilution effect, with an accompanying apparent decrease in blood lipid levels. The underlying mechanism could be mobilization of fluid from the interstitial to the intravascular compartment due to heat acclimatization.31- 33 Plasma hypervolemia during physical training has been associated with 2 major factors: an elevation in renin activity and vasopressin levels during exercise, which facilitates sodium and water retention, and a progressive, chronic increase in plasma albumin content, which provides increased blood-water binding capacity.34- 36 The fact that the HDL cholesterol concentration was not as affected by changes in plasma volume could be related to the complex interactions between seasonal changes in physical activity. Reductions in HDL cholesterol levels in the warmer months as a consequence of hypervolemia may have been counterbalanced by increased HDL cholesterol concentrations resulting from increased physical activity levels in the summer.
From the clinician's perspective, seasonal variation in blood lipid levels does occur, although such seasonality seems to be determined primarily by hemodilution in the summer and hemoconcentration in the winter. This leads to the following question: Is it the actual blood lipid concentration, at any given time, hemodilution or not, that is the most important factor in determining the rate of cholesterol deposition in the arteries, or is it the absolute amount of cholesterol carried in the blood that determines the pathophysiologic effects?37
If the former interpretation is correct, the measured cholesterol level does not require adjustment. However, patients whose lipid measures were considered borderline for therapy in the summer may require another measurement in the winter.
On the other hand, if the pathophysiologic effect is based on the absolute amount of cholesterol (ie, cholesterol concentration adjusted for the hemodilution effect), then there would be a bias toward inappropriately classifying individuals as hypercholesterolemic during the winter months.6 The same would pertain to patients who initiate drug therapy during the winter.
The prevalence of hypercholesterolemia in the United States is approximately 29% of the adult population, or 52 million people.38 Assuming an equal distribution of new diagnoses during the year, approximately 13 million people should be diagnosed as having hypercholesterolemia during the winter months. Our calculations suggest that during the winter months there may be an excess of 22% of individuals diagnosed as having hypercholesterolemia. This figure would translate into approximately 2.86 million people potentially being labeled as hypercholesterolemic, mainly because of relative winter hemoconcentration, and into 800 000 people likely being treated for hypercholesterolemia (≈$840 million in drug treatment alone [≈$1050/y per person], as calculated from average wholesale prices for statins39) as a result of misclassification of cholesterol status due to seasonal variation.
More patients are probably diagnosed as having high blood lipid levels during the winter as related to the distribution of cardiac events during the calendar year, with more coronary events occurring during the winter months,40 further increasing the diagnostic and treatment selection biases.
In this study, the greatest amplitude of seasonal change in blood lipid levels was observed in people in the upper quartile of the cholesterol distribution, particularly women. In this subgroup, another concern would be apparent treatment failure, evidenced when patients start treatment for hypercholesterolemia during the summer and are reevaluated during the winter, when hemoconcentration might be "artificially" increasing serum lipid levels. Using LDL cholesterol as the reference for treatment of hypercholesterolemia, as per the National Cholesterol Education Program–Adult Treatment Panel III guidelines,11 decreases the level of misclassification because of the lower magnitude of order of LDL cholesterol compared with that of total cholesterol. This observation would further reinforce the need to focus on LDL cholesterol levels rather than on total cholesterol levels for treatment purposes.
The seasonal variability in blood lipid levels observed in this study is of lower magnitude than that reported previously in countries such as Finland.7 However, it is also smaller than that reported in more recent studies in the United States.6 Many factors might be playing a role in the observed decrease in the amplitude of the seasonality of blood lipid levels, including changes in food availability, light exposure, environmental temperature, and physical activity41,42 that have occurred over time. These changes tend to blunt the usual human exposure to seasonal changes in environment, food supply, and physical activity.43 It is also possible that the demanding nature of the study tended to bias our participants toward individuals who have lifestyles that are relatively more constant during the year. A third possibility could be that most studies on seasonal variation in cholesterol levels are based on a cross-sectional study design. Seasonal variation in relative plasma volume may provide a plausible explanation for seasonal differences in cardiac events.44- 46 The clinical relevance of this finding is apparent, as may be an understanding of the mechanism of action. There are reports of statistically significant increases in fibrinogen levels and tissue plasminogen activator activity during the winter months,47- 49 which could be related to plasma volume contraction in response to cold exposure, determining a "clustering of peak values of haemostatic coronary risk factors."50,51 During the summer months, however, the body would benefit from a relative hemodilution, explained in part by a complex interaction of exposure to increased temperature and physical activity, which would determine a "hypocoagulable state" as a result of relative hemodilution. At the same time, the relative decrease in winter cardiac deaths during the past several decades has been associated with the availability of central heating.45 Therefore, seasonal changes in relative plasma volume presumably related to an interaction between exposure to increased temperature and physical activity could help explain seasonal differences in cardiac events. Changes in plasma volume may also represent another mechanism through which regular exercise exerts its cardioprotective effect.
In conclusion, this study demonstrates seasonal variation in blood lipid levels, with a peak in the winter and a trough in the summer. Our findings suggest that there is greater amplitude in seasonal variability in women and in people with hypercholesterolemia. However, changes in relative plasma volume seem to explain a substantial proportion of the observed seasonal difference in blood lipid levels. Changes in temperature and/or physical activity in winter and summer seem to be related to concomitant changes in relative plasma volume.
The information provided by this study could assist in the continuous development of guidelines for the treatment of hypercholesterolemia; however, we do not believe that season-specific guidelines would be justified. Further research is needed to better understand the mechanism through which physical activity and temperature control systems could aid in the prevention of coronary heart disease morbidity and mortality.
Corresponding author: Ira S. Ockene, MD, Division of Cardiovascular Medicine, Department of Medicine, University of Massachusetts Medical Center, Worcester, MA 01655 (e-mail: Ira.Ockene@umassmed.edu).
Accepted for publication May 30, 2003.
The SEASONS Study was supported by grant R01-HL52745 from the National Heart, Lung, and Blood Institute, Bethesda, Md.
We thank Laura Robidoux and Priscilla Cirillo, RN, for assisting with study recruitment and data collection; Kelly Scribner, BS, for coordinating the 24-hour recalls; SEASONS dietitians Susan Nelson, MS, RD, Christine Singelton, RD, Pat Jeans, RD, Karen Lafayette, RD, Deborah Lamb, RD, Stephanie Olson, RD, Eileen Capstraw, RD, and Barbara Olendzki, MPH, RD, for conducting the recalls; and Thomas Hurley, MS, for his organizational and data management expertise.