Coffee Consumption and Serum Lipids: A Randomized, Crossover Clinical Trial PHILLIPC. ROSMARIN, M.D., WILLIAM B. APPLEGATE, M.D., M.P.H., GRANTW. SOMES, Ph.D., Memphis,
Tennessee
PURPOSE: The purpose of this study was to determine the effect of consuming three or more cups of filter-brewed coffee per day on levels of serum lipids. SUEtJECTSANDMETHODS: Aprospective,randomized crossover clinical trial was performed. Twenty-one healthy white male subjects completed the trial, and consumed an average of 3.6 cups of coffee a day. Data were evaluated by analysis of variance for repeated measures and t-test for paired means. RESULTS: No effect of coffee consumption on serum total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, or apolipoprotein B was found. Diet, creamer use, and cigarette use as well as group assignment and time factors were controlled for in the analysis. CONCLUSION: We found no effect of moderate consumption of filter-brewed coffee on serum levels of atherogenic lipids. This study supports previous work that filtered coffee has no adverse effect on serum lipids. This has far-reaching implications given the widespread use of coffee and the current concern over coronary risk factors.
F
or more than two decades, there has been debate in the medical literature about possible deleterious effects of coffee consumption on health. Conflicting data have been reported from cross-sectional, longitudinal, and clinical studies. Several cross-sectional and longitudinal studies have reported that coffee consumption is associated with increased coronary heart disease incidence and mortality [l-4] and elevated levels of total serum cholesterol [5-221. However, other studies have not shown an association between coffee consumption and total mortality [23,24], coronary heart disease incidence [23,25], or serum cholesterol levels [26-311. Recently, Williams et al [14] reported a positive correlation between coffee consumption and serum levels of apolipoprotein B, low-density lipoprotein (LDL) cholesterol, and total cholesterol in a cross-sectional study (see Table I). The results from clinical trials have also appeared to be in conflict. An early trial showed that coffee consumption was related to an increase in serum cholesterol [32], whereas subsequent trials have suggested that this adverse effect is only seen with boiled but not filter-brewed [33-351 or instant coffee [36,37] (see Table II). Because of these conflicting results and because filter brewing is the most common form of coffee preparation in the United States, we designed a randomized, crossover trial to examine the effects of consumption of three or more cups of filter-brewed coffee on serum lipids. Although this study cannot address the issue of the impact of coffee consumption on total mortality or coronary heart disease incidence, the prospective nature of the trial allows a more definitive answer to the relationship of moderate consumption of filterbrewed coffee to serum lipoprotein levels.
SUBJECTS AND METHODS
From the Departments of Medicine and Preventive Medicine (PCR, WBA) and the Department of Btostatlstics and Epidemiology (GWS), Umversity of Tennessee, Memphis. The Health Science Center, Memphis, Tennessee. This work was suooorted in Dart bv a errant from the Unlversltv Phvwans Foundation of thd i]nlverslty’of Tennessee. Memphis. Request; for ieprmts should be addressed to Willlam B Aooleeate. M.D.. M.P.H.. Uwersltv of Tennessee, Memphis, Department ‘df iedicine. 800 Madison Avenue, Memphis, Tennessee 38163. Manuscript submitted December 21. 1988. and accepted ,n revised form February 12, 1990.
The design of this study was a prospective, randomized crossover clinical trial. The protocol was approved by the Institutional Review Board of the University of Tennessee and all subjects were given an explanation of the purpose and procedures of the trial. Healthy, young white men were recruited with advertisements placed in the major metropolitan daily newspaper (circulation-412,000) and notices posted in the newsletters of several of the city’s large corporations that employ substantial numbers of persons (combined employment-16,750). Thirty-five people responded in person to the advertisement and were screened in the clinic. Initial screening consisted of intake of demographic data, past and current medical history, coffee use, and the subject’s willingness to comply with the study protocol. If no exclusion criteria were uncovered in the screening and the subjects were willing to comply with protocol, written informed consent was obtained and more baseline history was obApril
1990
The American
Journal
of Medicine
Volume
88
349
COFFEE
CONSUMPTION
AND SERUM
LIPIDS / ROSMARIN
ET AL
TABLE I Cross-Sectional Observational Studies of Coffee’s Effect on Serum Cholesterol Relationship Design
Reference
n=N/A Age N/A n = 444 Age 41-76 n = 1,992 Age 35-69 n= 116 Age 16-62 n = 361 Age N/A n = 2,455 Age 20-64 n = 14,667 Age 20-54 n = 4,342 Age 20-54 n = 5,098 Age 20-49 n = 1,900 Age 40-55 n = 770 Age 18-24, 65-74 n = 4,757 Age 42 (mean) n = 42,627 Age N/A n = 701 Age 20-89 n = 5,089 Age 5-30 n = 77 Age 30-55 n = 1,596 Age 35-64 n = 5,858 Age 65-84 n = 658 Age 20-69 n = 14,667 Age 20-54 n = 9,347 Age 25-64 n = 9,239 Age 25-64
[51
PI i231 I71
1261 PI PI ~71
K'81 PI [lOI [Ill [I21 [I31 [301 [I41
[I51
[I61 1171
[I81 [311
[I91 PI
n= Age n = Age n= Age
WI PI
to Lipids
Tc
18,012 20-59 9,043 30-69 18,070 18-65
LDLc
Comments
Pos.
N/A
Pos.
N/A
0
N/A
Pos.
N/A
Pos. in smokers; 0 in nonsmokers Pos. in older males; 0 in others Pos.
Pos. in smokers; 0 in nonsmokers 0
0
N/A
0
N/A
0
N/A
Pos. in young; 0 in older
N/A
Pos. In males
N/A
Pos.
N/A
Pos. in females; 0 in males 0
Pos. in females
Pos.
Pos.
Strongest
Pos.
Pos.
0 correlation
Pos.
N/A
Pos.
Pos.
Only males studred; 0 correlation or cola Neg. correlation with tea
Pos.
N/A
0 correlation
0
N/A
Pos.
N/A
Pos.
N/A
Pos.
N/A
Pos. in males; 0 in females
N/A
N/A
No adjustment for smokrng; specrfrc demographics reported in a previous study Secondary finding in study of dretary fat and colon cancer Only males studied 5,209 matched controls; secondary frnding in study of Tc in vegetarians Posstble synergism between smoking and coffee on lipids No correlation with HDLc Inverse assoclatron females
wrth HDLc in
1976-1980 NHANES data
HDLc and apoB not correlated with coffee; Tc & apoB neg. correlation with tea in elderly Tc-0 correlation with coffee but pos. correlation with total caffeine in females 0 correlation with tea; 0 correlation of HDLc to coffee or tea
N/A pos. correlation
with apoB
wrth tea wrth tea
with HDLc
Measured effect small (0.3 mmol/L) but statistically significant; 0 correlatron with HDLc Pos. with boiled; 0 with filtered or instant 0 correlation with other caffeine Measured effect small (4 mg/dL) but statistically significant; 0 correlation with HDLc
1s. = positive; 0 = none; Neg. = negative; N/A = informatlon not available; Tc = total cholesterol; LDLc = LDL cholesterol; HDLc = HDL cholesterol: apoB = apollpoproteln E iANES = National Health and Nutrltlon Examination Survey.
tained on coffee, creamer, and cigarette use; regular exercise habits; and past history of serious underlying disease. Baseline examination and laboratory evaluations were performed to eliminate those with occult serious underlying diseases. Exclusion criteria included the following: non-coffee drinkers, heavy coffee drinkers (greater than six cups/day), New York Heart Association class 2 or greater heart disease, hyperlipidemia (total cholesterol greater than 6.85 mmol/L [265 mg/dL]), diabetes mellitus, peptic ulcer disease, obstructive airways disease requiring or potentially requiring theophylline therapy, and diastolic hypertension (diastolic blood pressure greater than 95 mm Hg). Heavy coffee drinkers were excluded because it was thought such persons 350
April
1990
The American
Journal
of Medicine
Volume
88
might have difficulty in abstaining from coffee. Baseline laboratory evaluation included complete blood count, 23-channel chemistry panel, urinalysis, electrocardiogram, and fasting lipid profile including apolipoprotein B. Twenty-four subjects met inclusion criteria and were randomized. Subjects were enrolled between March 31,1986, and January 6,1987, and all had completed the protocol by May 15, 1987. Using previous literature [33], we assumed a standard deviation for total cholesterol for a paired t-test to be 0.76 mmol/L (30 mg/dL). We wanted to be able to detect a 20% change in total cholesterol. We assumed that during the non-coffee phase the total cholesterol value would be 3.88 mmol/L (150 mg/dL), and we wanted to be able to detect a change of 0.76 mmoll
COFFEE
CONSUMPTION
AND SERUM
LIPIDS / ROSMARIN
ET AL 1
TABLEII RandomizedClinicalTrialsof Coffee’sEffecton SerumCholesterol Relationship Reference
[401 [411 ~321
[361 [331
Design
to Lipids
Tc
X-over n= 12 Age 18-63 X-over n = 20 Age 21-49 X-over n= 17 Age N/A X-over n= 12 Age 33-45 4-arm Parallel
LDLc
0
N/A
Short-term
Neg. with instant
N/A
Non-randomized design; all drank decaffeinated first, then caffeinated
Pos. with boiled
N/A
0 with instant
0 with instant
Randomized X-over comparing coffee, tea, and rosehips tea (caffeine-free)
Pos. with boiled
N/A
No significant effect with filtered: all hypercholesterolemic subjects
0 with instant
N/A
Pos. with boiled: 0 with filtered
Pos. with boiled; 0 with filtered
Pos. effect on apoB with boiled, not with filtered; 0 effect with tea
Pas. with boiled; 0 with filtered
Pos. with boiled; 0 with filtered
HDLc and apoB not affected by either boiled or filtered coffee
(3.hour)
experiment
n = 33 [371 [341 [351
Age 35-54 X-over n= 15 Age N/A X-over n = 42 Age 31-60 3-arm Parallel n= 107 Age 25.7 (mean)
wer = crossover: Pos. = oositive: 0 = none: Net?. = neeative: N/A = InformatIon not avaIlable; Tc = total cholesterol; LDLc = LDL cholesterol; HDLc = HDL cholesterol; apoB = apolipoproteln Ei. -
I
”
L, i.e., one standard deviation. For a two-tailed t-test at a level of significance of 0.05 and with a power of 0.80, this would require 10 subjects. Randomization was accomplished using a computer-generated list of numbers that were stored in double-sealed envelopes and held by a secretary not involved with the study. At randomization, the next envelope in sequence was opened to ascertain the group assignment. Randomization was to one of two groups. The first group (A) drank three or more cups of coffee a day for 2 months, then crossed over to abstaining from coffee for 2 months; the second group (B) abstained the first 2 months, and consumed three or more cups of coffee a day the second 2-month period. We decided to have subjects drink at least three cups/day of coffee because of the purported threshold effect noted by Williams et al [14] and because we believed this represented a moderate level of coffee intake. The 2-month length of the two phases of the study was chosen because previous work in this area has shown a plateauing of the effect of coffee on serum cholesterol levels at between 6 and 8 weeks following change in coffee intake [32-341. For this same reason, we did not believe an abstinence phase prior to entry or a washout phase between periods was necessary, since previous studies have demonstrated changes in serum cholesterol levels in similar time frames in both crossover [32,34] and concurrent [33] study designs. Subjects were supplied with one-pound cans of dripgrind coffee and a single-cup drip coffee filter funnel and filters and instructed to use these supplies as much as possible to meet their coffee intake requirements. They were further instructed that, when it was not feasible to use this equipment to meet their intake requirements, it was permissible to use other sources of filter-brewed coffee, but that instant, percolated, and vending machine coffee were not to be used. All
other forms of caffeine, including soft drinks, decaffeinated beverages, tea, and chocolate, were excluded from the diet for the entire 4 months of the study. Subjects were instructed to maintain their usual eating, exercise, alcohol, and cigarette habits with the exception of caffeine use. Subjects were seen at monthly intervals after randomization. At each visit, subjects were interviewed as to their adherence to protocol, changes in diet, and new medications, and determination of vital signs, weight, and new symptoms was performed. Subjects kept logs of their diets, coffee consumption (including type and amount of creamer used), and cigarette use. The diet log consisted of weekly semi-quantitative tabulations of the number of portions of specified foods over the month-long interval between visits. A semi-quantitative instrument was designed to identify the number of weekly servings of high-saturated-fat foods, as well as to identify the number of servings of “distracter” items containing low-fat content. This is a crude estimate of fat intake, but we believe it provided a relative index to determine if any given individual’s fat intake changed during the course of the study. At the crossover point and at completion of the study, fasting blood samples were taken for subsequent lipid profile and apolipoprotein B determinations. Values for total cholesterol, high-density lipoprotein (HDL) cholesterol, and triglycerides were determined by conventional fluorescent polarization enzymatic assay (TDx@ REA@) standardized by calibration with Centers for Disease Control controls. This assay uses dextran sulfate and magnesium to precipitate LDL and very-low-density lipoprotein (VLDL) cholesterol for the HDL cholesterol determination. All lipid profiles were performed on fresh serum samples obtained after at least 12 hours of fasting. Values for LDL and VLDL cholesterol were calculated April
1990 The American
Journal
of Medicine
Volume
88
351
COFFEECONSUMPTION AND SERUM LIPIDS / ROSMARIN ET AL
TABLEIII Baseline Characteristics of Subjects Completing Protocol (mean f SD) Group A (n = 10) Age (years) Range Weight (kg) Body mass index (kg/m2) Daily coffee consumption By Recall (cups/day) Weekly alcohol consumption By Recall (drinks/week) Regular aerobic exercise habits* (number) Systolic blood pressure (mm Hg) Diastolic blood pressure (mm Hg) Total cholesterol mmol/L mg/dL HDL cholesterol mmol/L
w/dL
LDL cholesterol mmol/L w/dL Apolipoprotein B pmol/L mg/dL VLDL cholesterol mmol/L w/dL Triglycerides mmol/L
m/dL
Group I3 (n = 11) 35.4 f 5.9 26-45
35.6 f 6.8 22-44 82.9f 11.2 27.0 f 3.6
28.5 f 7.3
3.85f
4.00f
1.67
1.3 f 1.4 6 114.8f 71.6f
14.3
11.1
5.00+0.76 193.4f 29.4
91.9*27.3 1.72
2.0f4.1 3 116.8f11.8 73.5f 13.1 4.86f1.00 188.1+38.7
0.99f0.21
0.95f0.15
38.4 f 8.0
36.9 f 5.8
3.41 f 0.67 131.7f26.1
118.0 f 33.3
1.72*0.20 88.0 f 10.5
1.67f0.31 85.6% 16.1
0.49f0.18 18.8 f 6.9
0.63 f 0.32 24.45 12.2
1.07f0.40
94.5f35.1
3.05f0.86
1.39f0.69 123.0f61.3
ote: Group A drank coffee first, then abstalned; Group B abstained first, then dran coffee (see text). Stabstlcal analysts (two-tailed t-test, except * = chl-square) revealeN no signtficant differences between the two groups. t Based on molecular weight apollpoprotein BIOO = 513.000.
according to the following formulas: VLDL ‘= triglycerides/5; LDL = total cholesterol - HDL cholesterol (triglyceride&). Serum samples for apolipoprotein B levels were preserved with 0.02% thimersol/sodium azide solution and frozen at -20°C. Upon completion of the study the entire batch was express-shipped frozen to a referral laboratory where apolipoprotein B levels were determined by electroimmunoassay according to methods previously described [38]. This method of preservation and freezing has been shown to provide good stability of the apolipoprotein B fraction with a less than 8% decrease in apolipoprotein B levels over a 6-month period of storage (personal communication, J. Fesmire). Data were collected on standardized forms and entered into our VAX 8600 computer for analysis. Data were evaluated by analysis of variance for repeated measures, which is especially suited for the crossover design to determine any effect of treatment as well as effect of group assignment or time. In addition, separate analysis of changes in lipid values was performed with a t-test for paired means. Analysis with adjustments for change in dietary fat intake, type and amount of creamer used, and cigarette use was also performed.
RESULTS Of the 24 subjects randomized, three failed to return after randomization (Subjects 11, 16, and 19). Therefore, the data presented here represent only the 21 who completed the protocol. The demographics of our subject population are listed in Table III. Throughout 352
April
1990 The American
Journal
of Medicine
Volume
88
the study period, weight did not change significantly (mean change = -0.5 kg). Nine subjects (43%) engaged in regular aerobic exercise, i.e., exercised 20 or more minutes/day at least three times a week. Mean coffee intake during the coffee phase was 3.6 cups/day (range, 2.1 to 6.7 cups/day). Ninety percent of the subjects used no form of creamer in their coffee during the study phase. The two who did use creamer used non-dairy (Subject 2) and a combination of milk or non-dairy (Subject 22). Only two smokers completed the protocol (9.5%); 19 were nonsmokers (90.5%). The mean cigarette use of the two smokers during the study period was 30.9 cigarettes/day (range, 22 to 41 cigarettes/day). Adherence to protocol based on selfreporting of coffee intake was good. Ninety percent of subjects drank three or more cups of coffee during the coffee phase and abstained during the no-coffee phase. In Group A, all subjects were compliant with the protocol. In Group B, there were two non-compliant subjects. One subject drank nine cups of coffee in the second week of the first month of the abstention phase, for a daily average of 0.4 cup/day for the month. He was counseled at the next visit and had no further coffee intake during the abstention phase. One subject failed to abstain during the second month of the nocoffee phase, averaging 2.4 cups/day. He also failed to meet his quota of three cups/day both months of the coffee phase, averaging 2.1 and 2.4 cups/day in the first and second months, respectively. Data were analyzed both including and excluding the smokers and this latter non-compliant subject. We used analysis of variance for repeated measures to search for any effects due to group assignment, coffee consumption, or period. There was no effect of group assignment or coffee consumption on serum lipid results. The values for the various lipid fractions during the coffee and no-coffee phases are given in Table IV. Also given in this table are the standard error of the mean (for the paired t-test), the associated confidence intervals for the difference in the lipid fractions, and the detectable difference with this sample size and observed variance at a level of significance of 0.05 and with a power of 0.80. There was a significant period effect noted for the total study cohort on total cholesterol (p
Tennessee
PURPOSE: The purpose of this study was to determine the effect of consuming three or more cups of filter-brewed coffee per day on levels of serum lipids. SUEtJECTSANDMETHODS: Aprospective,randomized crossover clinical trial was performed. Twenty-one healthy white male subjects completed the trial, and consumed an average of 3.6 cups of coffee a day. Data were evaluated by analysis of variance for repeated measures and t-test for paired means. RESULTS: No effect of coffee consumption on serum total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, or apolipoprotein B was found. Diet, creamer use, and cigarette use as well as group assignment and time factors were controlled for in the analysis. CONCLUSION: We found no effect of moderate consumption of filter-brewed coffee on serum levels of atherogenic lipids. This study supports previous work that filtered coffee has no adverse effect on serum lipids. This has far-reaching implications given the widespread use of coffee and the current concern over coronary risk factors.
F
or more than two decades, there has been debate in the medical literature about possible deleterious effects of coffee consumption on health. Conflicting data have been reported from cross-sectional, longitudinal, and clinical studies. Several cross-sectional and longitudinal studies have reported that coffee consumption is associated with increased coronary heart disease incidence and mortality [l-4] and elevated levels of total serum cholesterol [5-221. However, other studies have not shown an association between coffee consumption and total mortality [23,24], coronary heart disease incidence [23,25], or serum cholesterol levels [26-311. Recently, Williams et al [14] reported a positive correlation between coffee consumption and serum levels of apolipoprotein B, low-density lipoprotein (LDL) cholesterol, and total cholesterol in a cross-sectional study (see Table I). The results from clinical trials have also appeared to be in conflict. An early trial showed that coffee consumption was related to an increase in serum cholesterol [32], whereas subsequent trials have suggested that this adverse effect is only seen with boiled but not filter-brewed [33-351 or instant coffee [36,37] (see Table II). Because of these conflicting results and because filter brewing is the most common form of coffee preparation in the United States, we designed a randomized, crossover trial to examine the effects of consumption of three or more cups of filter-brewed coffee on serum lipids. Although this study cannot address the issue of the impact of coffee consumption on total mortality or coronary heart disease incidence, the prospective nature of the trial allows a more definitive answer to the relationship of moderate consumption of filterbrewed coffee to serum lipoprotein levels.
SUBJECTS AND METHODS
From the Departments of Medicine and Preventive Medicine (PCR, WBA) and the Department of Btostatlstics and Epidemiology (GWS), Umversity of Tennessee, Memphis. The Health Science Center, Memphis, Tennessee. This work was suooorted in Dart bv a errant from the Unlversltv Phvwans Foundation of thd i]nlverslty’of Tennessee. Memphis. Request; for ieprmts should be addressed to Willlam B Aooleeate. M.D.. M.P.H.. Uwersltv of Tennessee, Memphis, Department ‘df iedicine. 800 Madison Avenue, Memphis, Tennessee 38163. Manuscript submitted December 21. 1988. and accepted ,n revised form February 12, 1990.
The design of this study was a prospective, randomized crossover clinical trial. The protocol was approved by the Institutional Review Board of the University of Tennessee and all subjects were given an explanation of the purpose and procedures of the trial. Healthy, young white men were recruited with advertisements placed in the major metropolitan daily newspaper (circulation-412,000) and notices posted in the newsletters of several of the city’s large corporations that employ substantial numbers of persons (combined employment-16,750). Thirty-five people responded in person to the advertisement and were screened in the clinic. Initial screening consisted of intake of demographic data, past and current medical history, coffee use, and the subject’s willingness to comply with the study protocol. If no exclusion criteria were uncovered in the screening and the subjects were willing to comply with protocol, written informed consent was obtained and more baseline history was obApril
1990
The American
Journal
of Medicine
Volume
88
349
COFFEE
CONSUMPTION
AND SERUM
LIPIDS / ROSMARIN
ET AL
TABLE I Cross-Sectional Observational Studies of Coffee’s Effect on Serum Cholesterol Relationship Design
Reference
n=N/A Age N/A n = 444 Age 41-76 n = 1,992 Age 35-69 n= 116 Age 16-62 n = 361 Age N/A n = 2,455 Age 20-64 n = 14,667 Age 20-54 n = 4,342 Age 20-54 n = 5,098 Age 20-49 n = 1,900 Age 40-55 n = 770 Age 18-24, 65-74 n = 4,757 Age 42 (mean) n = 42,627 Age N/A n = 701 Age 20-89 n = 5,089 Age 5-30 n = 77 Age 30-55 n = 1,596 Age 35-64 n = 5,858 Age 65-84 n = 658 Age 20-69 n = 14,667 Age 20-54 n = 9,347 Age 25-64 n = 9,239 Age 25-64
[51
PI i231 I71
1261 PI PI ~71
K'81 PI [lOI [Ill [I21 [I31 [301 [I41
[I51
[I61 1171
[I81 [311
[I91 PI
n= Age n = Age n= Age
WI PI
to Lipids
Tc
18,012 20-59 9,043 30-69 18,070 18-65
LDLc
Comments
Pos.
N/A
Pos.
N/A
0
N/A
Pos.
N/A
Pos. in smokers; 0 in nonsmokers Pos. in older males; 0 in others Pos.
Pos. in smokers; 0 in nonsmokers 0
0
N/A
0
N/A
0
N/A
Pos. in young; 0 in older
N/A
Pos. In males
N/A
Pos.
N/A
Pos. in females; 0 in males 0
Pos. in females
Pos.
Pos.
Strongest
Pos.
Pos.
0 correlation
Pos.
N/A
Pos.
Pos.
Only males studred; 0 correlation or cola Neg. correlation with tea
Pos.
N/A
0 correlation
0
N/A
Pos.
N/A
Pos.
N/A
Pos.
N/A
Pos. in males; 0 in females
N/A
N/A
No adjustment for smokrng; specrfrc demographics reported in a previous study Secondary finding in study of dretary fat and colon cancer Only males studied 5,209 matched controls; secondary frnding in study of Tc in vegetarians Posstble synergism between smoking and coffee on lipids No correlation with HDLc Inverse assoclatron females
wrth HDLc in
1976-1980 NHANES data
HDLc and apoB not correlated with coffee; Tc & apoB neg. correlation with tea in elderly Tc-0 correlation with coffee but pos. correlation with total caffeine in females 0 correlation with tea; 0 correlation of HDLc to coffee or tea
N/A pos. correlation
with apoB
wrth tea wrth tea
with HDLc
Measured effect small (0.3 mmol/L) but statistically significant; 0 correlatron with HDLc Pos. with boiled; 0 with filtered or instant 0 correlation with other caffeine Measured effect small (4 mg/dL) but statistically significant; 0 correlation with HDLc
1s. = positive; 0 = none; Neg. = negative; N/A = informatlon not available; Tc = total cholesterol; LDLc = LDL cholesterol; HDLc = HDL cholesterol: apoB = apollpoproteln E iANES = National Health and Nutrltlon Examination Survey.
tained on coffee, creamer, and cigarette use; regular exercise habits; and past history of serious underlying disease. Baseline examination and laboratory evaluations were performed to eliminate those with occult serious underlying diseases. Exclusion criteria included the following: non-coffee drinkers, heavy coffee drinkers (greater than six cups/day), New York Heart Association class 2 or greater heart disease, hyperlipidemia (total cholesterol greater than 6.85 mmol/L [265 mg/dL]), diabetes mellitus, peptic ulcer disease, obstructive airways disease requiring or potentially requiring theophylline therapy, and diastolic hypertension (diastolic blood pressure greater than 95 mm Hg). Heavy coffee drinkers were excluded because it was thought such persons 350
April
1990
The American
Journal
of Medicine
Volume
88
might have difficulty in abstaining from coffee. Baseline laboratory evaluation included complete blood count, 23-channel chemistry panel, urinalysis, electrocardiogram, and fasting lipid profile including apolipoprotein B. Twenty-four subjects met inclusion criteria and were randomized. Subjects were enrolled between March 31,1986, and January 6,1987, and all had completed the protocol by May 15, 1987. Using previous literature [33], we assumed a standard deviation for total cholesterol for a paired t-test to be 0.76 mmol/L (30 mg/dL). We wanted to be able to detect a 20% change in total cholesterol. We assumed that during the non-coffee phase the total cholesterol value would be 3.88 mmol/L (150 mg/dL), and we wanted to be able to detect a change of 0.76 mmoll
COFFEE
CONSUMPTION
AND SERUM
LIPIDS / ROSMARIN
ET AL 1
TABLEII RandomizedClinicalTrialsof Coffee’sEffecton SerumCholesterol Relationship Reference
[401 [411 ~321
[361 [331
Design
to Lipids
Tc
X-over n= 12 Age 18-63 X-over n = 20 Age 21-49 X-over n= 17 Age N/A X-over n= 12 Age 33-45 4-arm Parallel
LDLc
0
N/A
Short-term
Neg. with instant
N/A
Non-randomized design; all drank decaffeinated first, then caffeinated
Pos. with boiled
N/A
0 with instant
0 with instant
Randomized X-over comparing coffee, tea, and rosehips tea (caffeine-free)
Pos. with boiled
N/A
No significant effect with filtered: all hypercholesterolemic subjects
0 with instant
N/A
Pos. with boiled: 0 with filtered
Pos. with boiled; 0 with filtered
Pos. effect on apoB with boiled, not with filtered; 0 effect with tea
Pas. with boiled; 0 with filtered
Pos. with boiled; 0 with filtered
HDLc and apoB not affected by either boiled or filtered coffee
(3.hour)
experiment
n = 33 [371 [341 [351
Age 35-54 X-over n= 15 Age N/A X-over n = 42 Age 31-60 3-arm Parallel n= 107 Age 25.7 (mean)
wer = crossover: Pos. = oositive: 0 = none: Net?. = neeative: N/A = InformatIon not avaIlable; Tc = total cholesterol; LDLc = LDL cholesterol; HDLc = HDL cholesterol; apoB = apolipoproteln Ei. -
I
”
L, i.e., one standard deviation. For a two-tailed t-test at a level of significance of 0.05 and with a power of 0.80, this would require 10 subjects. Randomization was accomplished using a computer-generated list of numbers that were stored in double-sealed envelopes and held by a secretary not involved with the study. At randomization, the next envelope in sequence was opened to ascertain the group assignment. Randomization was to one of two groups. The first group (A) drank three or more cups of coffee a day for 2 months, then crossed over to abstaining from coffee for 2 months; the second group (B) abstained the first 2 months, and consumed three or more cups of coffee a day the second 2-month period. We decided to have subjects drink at least three cups/day of coffee because of the purported threshold effect noted by Williams et al [14] and because we believed this represented a moderate level of coffee intake. The 2-month length of the two phases of the study was chosen because previous work in this area has shown a plateauing of the effect of coffee on serum cholesterol levels at between 6 and 8 weeks following change in coffee intake [32-341. For this same reason, we did not believe an abstinence phase prior to entry or a washout phase between periods was necessary, since previous studies have demonstrated changes in serum cholesterol levels in similar time frames in both crossover [32,34] and concurrent [33] study designs. Subjects were supplied with one-pound cans of dripgrind coffee and a single-cup drip coffee filter funnel and filters and instructed to use these supplies as much as possible to meet their coffee intake requirements. They were further instructed that, when it was not feasible to use this equipment to meet their intake requirements, it was permissible to use other sources of filter-brewed coffee, but that instant, percolated, and vending machine coffee were not to be used. All
other forms of caffeine, including soft drinks, decaffeinated beverages, tea, and chocolate, were excluded from the diet for the entire 4 months of the study. Subjects were instructed to maintain their usual eating, exercise, alcohol, and cigarette habits with the exception of caffeine use. Subjects were seen at monthly intervals after randomization. At each visit, subjects were interviewed as to their adherence to protocol, changes in diet, and new medications, and determination of vital signs, weight, and new symptoms was performed. Subjects kept logs of their diets, coffee consumption (including type and amount of creamer used), and cigarette use. The diet log consisted of weekly semi-quantitative tabulations of the number of portions of specified foods over the month-long interval between visits. A semi-quantitative instrument was designed to identify the number of weekly servings of high-saturated-fat foods, as well as to identify the number of servings of “distracter” items containing low-fat content. This is a crude estimate of fat intake, but we believe it provided a relative index to determine if any given individual’s fat intake changed during the course of the study. At the crossover point and at completion of the study, fasting blood samples were taken for subsequent lipid profile and apolipoprotein B determinations. Values for total cholesterol, high-density lipoprotein (HDL) cholesterol, and triglycerides were determined by conventional fluorescent polarization enzymatic assay (TDx@ REA@) standardized by calibration with Centers for Disease Control controls. This assay uses dextran sulfate and magnesium to precipitate LDL and very-low-density lipoprotein (VLDL) cholesterol for the HDL cholesterol determination. All lipid profiles were performed on fresh serum samples obtained after at least 12 hours of fasting. Values for LDL and VLDL cholesterol were calculated April
1990 The American
Journal
of Medicine
Volume
88
351
COFFEECONSUMPTION AND SERUM LIPIDS / ROSMARIN ET AL
TABLEIII Baseline Characteristics of Subjects Completing Protocol (mean f SD) Group A (n = 10) Age (years) Range Weight (kg) Body mass index (kg/m2) Daily coffee consumption By Recall (cups/day) Weekly alcohol consumption By Recall (drinks/week) Regular aerobic exercise habits* (number) Systolic blood pressure (mm Hg) Diastolic blood pressure (mm Hg) Total cholesterol mmol/L mg/dL HDL cholesterol mmol/L
w/dL
LDL cholesterol mmol/L w/dL Apolipoprotein B pmol/L mg/dL VLDL cholesterol mmol/L w/dL Triglycerides mmol/L
m/dL
Group I3 (n = 11) 35.4 f 5.9 26-45
35.6 f 6.8 22-44 82.9f 11.2 27.0 f 3.6
28.5 f 7.3
3.85f
4.00f
1.67
1.3 f 1.4 6 114.8f 71.6f
14.3
11.1
5.00+0.76 193.4f 29.4
91.9*27.3 1.72
2.0f4.1 3 116.8f11.8 73.5f 13.1 4.86f1.00 188.1+38.7
0.99f0.21
0.95f0.15
38.4 f 8.0
36.9 f 5.8
3.41 f 0.67 131.7f26.1
118.0 f 33.3
1.72*0.20 88.0 f 10.5
1.67f0.31 85.6% 16.1
0.49f0.18 18.8 f 6.9
0.63 f 0.32 24.45 12.2
1.07f0.40
94.5f35.1
3.05f0.86
1.39f0.69 123.0f61.3
ote: Group A drank coffee first, then abstalned; Group B abstained first, then dran coffee (see text). Stabstlcal analysts (two-tailed t-test, except * = chl-square) revealeN no signtficant differences between the two groups. t Based on molecular weight apollpoprotein BIOO = 513.000.
according to the following formulas: VLDL ‘= triglycerides/5; LDL = total cholesterol - HDL cholesterol (triglyceride&). Serum samples for apolipoprotein B levels were preserved with 0.02% thimersol/sodium azide solution and frozen at -20°C. Upon completion of the study the entire batch was express-shipped frozen to a referral laboratory where apolipoprotein B levels were determined by electroimmunoassay according to methods previously described [38]. This method of preservation and freezing has been shown to provide good stability of the apolipoprotein B fraction with a less than 8% decrease in apolipoprotein B levels over a 6-month period of storage (personal communication, J. Fesmire). Data were collected on standardized forms and entered into our VAX 8600 computer for analysis. Data were evaluated by analysis of variance for repeated measures, which is especially suited for the crossover design to determine any effect of treatment as well as effect of group assignment or time. In addition, separate analysis of changes in lipid values was performed with a t-test for paired means. Analysis with adjustments for change in dietary fat intake, type and amount of creamer used, and cigarette use was also performed.
RESULTS Of the 24 subjects randomized, three failed to return after randomization (Subjects 11, 16, and 19). Therefore, the data presented here represent only the 21 who completed the protocol. The demographics of our subject population are listed in Table III. Throughout 352
April
1990 The American
Journal
of Medicine
Volume
88
the study period, weight did not change significantly (mean change = -0.5 kg). Nine subjects (43%) engaged in regular aerobic exercise, i.e., exercised 20 or more minutes/day at least three times a week. Mean coffee intake during the coffee phase was 3.6 cups/day (range, 2.1 to 6.7 cups/day). Ninety percent of the subjects used no form of creamer in their coffee during the study phase. The two who did use creamer used non-dairy (Subject 2) and a combination of milk or non-dairy (Subject 22). Only two smokers completed the protocol (9.5%); 19 were nonsmokers (90.5%). The mean cigarette use of the two smokers during the study period was 30.9 cigarettes/day (range, 22 to 41 cigarettes/day). Adherence to protocol based on selfreporting of coffee intake was good. Ninety percent of subjects drank three or more cups of coffee during the coffee phase and abstained during the no-coffee phase. In Group A, all subjects were compliant with the protocol. In Group B, there were two non-compliant subjects. One subject drank nine cups of coffee in the second week of the first month of the abstention phase, for a daily average of 0.4 cup/day for the month. He was counseled at the next visit and had no further coffee intake during the abstention phase. One subject failed to abstain during the second month of the nocoffee phase, averaging 2.4 cups/day. He also failed to meet his quota of three cups/day both months of the coffee phase, averaging 2.1 and 2.4 cups/day in the first and second months, respectively. Data were analyzed both including and excluding the smokers and this latter non-compliant subject. We used analysis of variance for repeated measures to search for any effects due to group assignment, coffee consumption, or period. There was no effect of group assignment or coffee consumption on serum lipid results. The values for the various lipid fractions during the coffee and no-coffee phases are given in Table IV. Also given in this table are the standard error of the mean (for the paired t-test), the associated confidence intervals for the difference in the lipid fractions, and the detectable difference with this sample size and observed variance at a level of significance of 0.05 and with a power of 0.80. There was a significant period effect noted for the total study cohort on total cholesterol (p
