I From the Center for Human Drug Research, the Department of Cardiology, 2 Supported by SARA LEE/Douwe Egberts, Utrecht, Netherlands.

Transcription

1 A placebocontrolled parallel study of the effect of two types of coffee oil on serum lipids and transaminases: identification of chemical -substances involved in the cholesterol- raisi ng effect of coffee Jeroen van Rooij, Gerrit HD van der Stegen, Rik C Schoemaker, Cees Kroon, Jacobus Burggraaf Leny Hollaar, Ton FFP Vroon, Augustinus HM Smelt, and Adam F Cohen ABSTRACT In a randomized double-blind parallel study in 36 subjects the effect on serum cholesterol of a daily dose of 2 g lipid extracted from green Arabica and Robusta coffee beans was studied. Arabica oil elevated serum total cholesterol by 1.1 mmol/l (95% Cl for the difference from placebo:.41, 1.73 mmol/l); the effect of robusta oil (+.5 mmolfl) was not statistically significant (95% CI: -.1,.92 mmol/l). Arabica oil also raised plasma triglycerides by.8 mmol/l (95% CI:.26, 1.25 mmol/l). The effect of robusta oil on triglycerides was +.14 mmol/l and not significant (95% CI: -.26,.42 mmol/l). In the 12 subjects taking arabica oil an average serum alanine aminotransferase elevation of 18 U/L (95% CI: 9.4, 28.4 U/L) was observed. Because only arabica oil contains kahweol and arabica coffee contains more cafestol than does robusta oil, this is further evidence for the role of diterpenes in the rise of serum cholesterol and alanine aminotransferase after consumption of boiled coffee. Am J C/in Nutr1995;61: KY WORDS Coffee, cholesterol, serum lipids, triglycerides, transaminases, liver biochemistry, kahweol, cafestol, diterpenes INTRODUCTION The association between coffee drinking and coronary heart disease is inconclusive. Several population studies revealed an increased risk for coronary heart disease in heavy coffee drinkers (1), but other studies could not corroborate this association (2, 3). In a recent meta-analysis no clear association between coffee drinking and coronary heart disease was found (4). However, in Scandinavia, where large quantities of boiled unfiltered coffee are consumed (5), a clear relation between coffee consumption and coronary heart disease was demonstrated (6). Population studies concerning the influence of coffee consumption on serum cholesterol also reveal conflicting data. In a review article (7), it was concluded that a positive correlation between coffee intake and serum cholesterol concentration was present in approximately two-thirds of the studies on this subject with an indication of a dose-response relation in some of the studies (8-). Differences in the method of preparation of coffee may explain the difference in outcomes in the different populations that were examined (5, 9-14). Intervention studies revealed a significant rise of serum total cholesterol and low-density-lipoprotein-(ldl) cholesterol concentrations after consumption of boiled coffee for 6 wk (8, 15-17). Because filtered coffee and tea do not affect serum cholesterol concentrations (8, 9, 13, 15, 17), caffeine is not likely to be responsible for the cholesterol-elevating effect of coffee. In one intervention study a rise in the serum concentrations of LDL cholesterol and apolipoprotein B was found if subjects changed from caffeinated coffee to decaffeinated coffee (18). Boiled coffee contains a lipid fraction that does not pass through a paper filter, and after consumption of boiled coffee that is subsequently filtered, no rise in serum cholesterol was detected (16, 19). Administration of a lipid-enriched fraction from boiled coffee to healthy volunteers led to a rise in serum cholesterol (23%), LDL cholesterol (29%), and triglycerides (55%) after 6 wk of administration (2). After supplementation had ended, these concentrations returned to baseline. Coffee oil consists of triglycerides, fatty acid esters of diterpenes and sterols, free sterols, free diterpenes, phospholipids, and other unsaponifiables (2, 21). The main fatty acids in coffee oil in descending order of quantity are linoleic, palmitic, oleic, stearic, and arachidic acids. Typical for coffee oil are the diterpenes (kahweol, cafestol, and 16-o-methylcafestol), which are present mainly as fatty acid esters. In the oil of roasted coffee small amounts of monodehydrated pyrolysis products of the kahweol and cafestol esters occur as well (22). Oil from arabica coffee contains fatty acid esters of kahweol and cafestol, whereas oil from robusta coffee contains lower I From the Center for Human Drug Research, the Department of Cardiology, and the Department of Internal Medicine, University Hospital Leiden, Leiden, Netherlands, and SARA LfDouwe gberts, Department of Research and Quality, Utrecht, Netherlands. 2 Supported by SARA L/Douwe gberts, Utrecht, Netherlands. 3 Address reprint requests to AF Cohen, Center for Human Drug Research, University Hospital Leiden, P Box 96(), 23 RC Leiden, Netherlands. Received May 23, Accepted for publication February 9, Am J C/in Nutr 1995;61: Printed in USA American Society for Clinical Nutrition 1277

2 1278 VAN ROOIJ T AL amounts of cafestol esters, some 16-o-methylcafestol esters, and only traces of kahweol esters. In two recent studies it was demonstrated that cafestol and/or kahweol are probably responsible for the effect of boiled coffee on serum cholesterol concentrations (23, 24). In this study we investigated the effects of a kahweol- and cafestol-containing preparation from arabica coffee and a preparation from robusta coffee containing only cafestol. These preparations were made from green (unroasted) coffee beans to exclude the effect of pyrolysis products formed during the roasting process. Because no reports on administration of green coffee oil to human subjects were published to date we decided to monitor a range of biochemical variables, including liver enzyme activities, and to report observed changes to other study groups using coffee oil in human subjects. SUBJCTS AND MTHODS Subjects Thirty-six healthy normotensive subjects (18 males, 18 females) aged y were recruited. Demographics and pretreatment statistics of the subjects are provided in Table 1. The study was submitted to and approved by the thical Committee of Leiden University Hospital and all subjects gave written acknowledgment of informed consent to participate before the start of the study. Subjects were required to have a fasting serum cholesterol concentration between 4. and 7.3 mmolfl and a body mass index (kg/m2) between the 25th and 75th percentile of the Geigy scientific tables (25). Only subjects who drank filtered coffee, and a maximum of two cups per day of coffee prepared by any other method, were included. All subjects were healthy as determined by a medical history, physical examination, a 12-lead electrocardiograph, and a biochemical and hematological screening. Subjects were exeluded if any abnormalities of clinical significance were found with special attention given to alcohol or drug abuse. Subjects receiving medical treatment were excluded with the exception of women taking oral contraceptive drugs. The clinical phase of the study was conducted between April and August Trial design This trial was designed as a randomized, double-blind, placebo-controlled parallel study with three treatments. The subjects were randomly assigned to 6 wk treatment with peanut oil ( placebo), arabica oil, or robusta coffee oil. ach treatment group of 12 subjects contained equal numbers of male and female participants. Treatments and dosages Coffee oil was extracted with hexane from green beans of arabica and robusta coffee. The concentrations of the diterpene alcohols and esters in the coffee oil used is presented in Table 2. The kahweol and cafestol fatty acid esters consisted mainly of palmitic, linoleic, and stearic acids. Peanut oil was used as placebo. To improve palatability the different types of oil were administered as an emulsion, which contained methylcellulose, saccharose, and raspberry flavor. Quality control of the coffee oil emulsion involved exclusion of significant residuals of hexane and determination of the stability of the diterpene contents over the period of the study. ach subject was instructed to homogenize the emulsion before consuming 7.5 ml emulsion (1 g coffee oil) with the aid of a -ml syringe twice daily for 6 wk. This amount of coffee oil is equivalent to ' L boiled coffee (16, 19). To determine compliance, the subjects received a volume of emulsion sufficient for 1 wk and were requested each week to hand in the used bottle from the preceding week, which was checked for remaining emulsion. Procedures The subjects were asked to maintain their habitual diet. ach week the daily food intake of each subject was estimated by means of a 24-h recall. Dietary information was converted to energy and nutrient intakes with the aid of a computerized version of the Netherlands food table. Subjects were instructed to drink filtered coffee only. No more than two alcoholic drinks were allowed on the day before blood sampling. The subjects were weighed weekly on the same scales with indoor clothing and without shoes; adverse events and concomitant medical treatments were documented. Weekly venous blood samples were taken after an overnight fast by using the briefest possible period of venous occlusion, with each subject being in the same position (sitting or supine) on each occasion. All blood samples were analyzed in the clinical chemistry laboratory and the Laboratory for Cardiobiochemistry; both laboratories are departments of the Leiden University Hospital and participate in a wide range of qualitycontrol schemes. In the blood samples, serum total cholesterol and triglycerides were determined enzymatically on an SMAC2 apparatus (Bayer, Germany). High-density-lipoprotein (HDL) cholesterol was determined enzymatically after precipitation of very-low-density lipoprotein (VLDL) and by tungstic acid on a TABL 1 Demographic data TABL 2 Arabica (n 12) Robusta (n 12) Placebo (n = 12) Diterpene alcohol and ester contents of coffee oils Arabica Robusta BMI (kg/m2) 23.8 ( ) 23.8 ( ) 22.8 ( ) Total cholesterol (mmolil) 5.4 ( ) 4.8 ( ) 5.2 ( ) HDL (mmol/l) 1.5 (.6-2.6) 1.5 (.8-2.) 1.5 (.7-2.5) Triglycerides (mmol/l) 1.5 (.6-2.1).98 (.7-1.4) 1.12 (.6-1.7), Mean baseline values at the start of the study; range in parentheses. Differences between groups were not significant. ND, not detectable. Determinations were performed by means of HPLC with ultraviolet detection. g/kg Kahweol fatty acid esters 85 1 Cafestol fatty acid esters Free kahweol 2 ND Free cafestol o-Methylcafestol esters ND 9

3 CHOLSTROL-RAISING FACTORS IN COFF 1279 Dimension apparatus (DuPont, Wilmington, D), and LDL concentrations were calculated by using the formula of Friedewald et al (26). Additional biochemical indexes measured were urea, creatinine, aspartate aminotrasferase (AST), alanine aminotransferase (ALT), lactate dehydrogenase (LDH), and y-glutamyltransferase (y-gt). All samples were analyzed within 4 h of blood sampling. Apolipoproteins A-I and B were determined by nephelometry in one run in the samples taken before treatment and in the last sample taken during treatment. Twenty milliliters blood was taken for ultracentrifugation at the moment treatment ended from five subjects from both the robusta and placebo group. From six subjects of the arabica group these blood samples were taken 1-6 d (i: 4.5 d) after the intake of oil was stopped; this was because several of the subjects in this group discontinued their intake of coffee oil because of elevations in serum transaminases. Six weeks after treatments were discontinued, all biochemical indexes mentioned above were determined in all subjects and additionally at 3 or 4 d, and 1, 3, and 6 wk after discontinuation of the coffee oil emulsion for the subjects of the arabica group. To rule out the theoretical possibility of interference of the coffee oil emulsion with the assays of AST and ALT, 3, 6, and 12 p.l of the emulsion were added to 1 ml of pooled samples with a known activity for these two enzymes. AST and ALT values were not different from those of the unfortified Statistical sample. analysis For all variables except data obtained by ultracentrifugation, the absolute change from the pretreatment value was calculated. All variables were analyzed by using one-way analysis of variance because only the change from baseline was analyzed. In case of significant overall results, these analyses were followed by unpaired t tests with separate variance estimates. Results are presented with 95% CIs of the difference from placebo. If the CI of the difference does not include zero, the difference is considered to be statistically significant. If no significant effects were detected in the overall analysis, no subsequent contrasts were examined and 95% CIs were not calculated as protection against excessive type I errors resulting from multiple comparisons. All calculations were carried out by using SPSSIPC+ version 4..1 (SPSS, Inc, Chicago). RSULTS Two subjects who started oil intake did not complete the study. One of them discontinued the oil after 1 wk because she felt unwell; this subject was replaced. One subject (arabica group) discontinued the intake of the coffee oil after 4 wk because of nausea after the consumption of the emulsion; the data of this subject were included in the analysis. Transaminase values were not elevated for this subject. Coffee oil was discontinued in eight subjects of the arabica treatment group because of elevations of serum transaminase values. The administration was discontinued after 28 d for four subjects, after 32 d for one, and after 35, 36, and 38 d for three other subjects. No significant clinical events occurred in any of the subjects. All subjects attended weekly appointments; on some occasions there was a shift of some days due to illness or other personal circumstances. On a few occasions subjects had a small breakfast (no more than a glass of milk or some fruit). For these occasions serum lipid values were inspected for extreme values for that subject; no values had to be excluded for this reason. All treatment containers were returned to the center and contained the expected residual amount of oil. Cholesterol and triglycerides In the subjects who received arabica oil the average fasting total serum cholesterol concentration rose by 1.14 ±.97 mmol/l or 2.1% (95% CI of the difference from placebo:.41, 1.73 mmol/l) at the last sampling time before cessation of intake from an average pretreatment value of 5.4 ±.85 mmol/l (Table 3 and Figure 1). Robusta oil tended to induce an increase of.52 ±.62 mmolfl (95% CI: -.1,.92) from 4.8 ±.5 mmolfl, and placebo of.7 ±.46 mmol/l from a baseline value of 5.2 ±.64 mmolfl. The serum LDL-cholesterol concentration calculated with the formula of TABL 3 Average change in biochemical indexes after 6 wk of coffee oil intake 95 % CIs of the difference2 Arabica Robusta Placebo Arabica - Robusta - Arabica - (n 12) (n 12) (n 12) P placebo placebo robusta y-gt (UIL).2 ± ± ± 1.3 AST (U/L) 2.7 ± ± ± (1., 7.1) (.4, 4.) (- 1.18, 4.9) ALT (UIL) 17.9 ± ± ± 6.5 <.1 (9.4, 28.4) (-.9, 8.2) (6.3, 24.2) LDH (U/L) - 6 ± ± ± 16.8 Total cholesterol (mmolfl) 1.14 ± ±.62.7 ±.46.4 (.41, 1.73) (-.1,.92) (-.8, 1.31) Triglycerides (mmolfl).81 ± ±.45.6 ±.34.2 (.26, 1.25) (-.26,.42) (.15, 1.2) HDL cholesterol (mmolfl). ±.48.1 ± ±.21 LDL cholesterol (mmol/l)4.77 ± ± ±.29 Apo Al (gil) -.3 ±.19.5 ± ±.19 Apo B (g/l).35 ± ±.21.5 ±..7 (.11,.48) (-.2,.27) (-.4,.38), All subjects in the robusta and placebo groups completed 6 wk of oil consumption. Coffee oil was discontinued in eight subjects of the arabica treatment group because of elevations of serum transaminase values: after 28 d for four subjects, after 32 d for one, and after 35, 36, and 38 d for three subjects. 2 Calculated if overall analysis revealed significant differences among the three groups. ± SD. 4 Calculated with the Friedewald et al (26) formula.

4 128 VAN ROOIJ T AL 9 5, Weeks B 6 B Weeks Io I 1 1 I - IT I FIGUR 1. A: Average total serum cholesterol concentrations during arabica (l, ii = 12), robusta (, n = 12), and placebo (, n = 12) treatments. The diminishing number of subjects in the arabica group was due to discontinuation of treatment on the basis of serum alanine aminotransferase concentrations exceeding 25 UIL. B: Values for all subjects after discontinuation of administration. The solid line connects values of arabica group subjects (U). Six weeks after discontinuation only one blood sample was taken from all subjects in the robusta ( ) and placebo () groups. Friedewald et al showed a trend to rise only during coffee oil administration:.8 ± 1.1 mmol/l (from 3.4 ±.71 mmol/l at baseline) in the arabica group,.4 ±.65 mmol/l (from 2.9 ±.54 to 3.3 ±.71 mmolfl) in the robusta group and.1 ±.29 (from 3.2 ±.6 mmol/l at baseline) in the placebo group (Table 3). Overall analysis for the change of this variable did not reveal a significant difference between the three groups (P =.11). The mean serum triglyceride concentrations were increased by.81 ±.73 mmol/l (95% CI:.26, 1.25) from 1.5 ±.52 mmolfl after arabica oil. Changes from baseline were not significant for the two remaining treatment groups: after robusta oil consumption +.14 ±.45 mmol/l (95% CI: -.26,.42) from.98 ±.23 mmolfl and +.6 ±.34 mmolfl from ±.38 mmolfl in the placebo group (Table 3 and Figure 2). Arabica oil induced a greater rise than robusta oil (+.67 ±.25 mmol/l; 95 % CI:.15, 1.2). HDL cholesterol values were not affected (P =.79). The effect of the three different oils on apolipoprotein A-I was small and the overall effect was not significant (P =.49). Apolipoprotein B was increased by.35 ±.28 g/l (95% CI:.11,.48) after arabica oil. Changes from baseline were not I I I I FIGUR 2. A: Average serum triglyceride concentrations during arabica (l, n = 12), robusta (, n 12), and placebo (, n = 12) treatment. The diminishing number of subjects in the arabica group was due to discontinuation of treatment on the basis of serum alanine aminotransferase activities exceeding 25 U/L. B: Values for all subjects 6 wk after discontinuation of administration. The solid line connects values of arabica group subjects (l). Six weeks after discontinuation only one blood sample was taken from all subjects in the robusta ( ) and placebo () groups. significant for robusta and placebo: +.17 ±.21 g/l (95% CI: -.2,.27) and +.5 ±. gil, respectively. There were only significant effects of arabica oil on total (P.22) and LDL cholesterol (P =.25). Total cholesterol was 6.29 ±.56 mmol/l after arabica oil (difference from placebo 1.37 mmol/l; 95% CI:.22, 2.52) and 5.23 ±.75 mmol/l after robusta oil (difference from placebo.31 mmol/l; 95% CI: -.94, 1.58). LDL cholesterol was 4.62 ±.79 mmol/l after arabica treatment (difference from placebo 1.25 mmolil; 95% CI:.27, 2.2), and 3.62 ±.62 mmol/l after robusta (difference from placebo.25 mmol/l; 95% CI: -.67, 1.19). Comparison of ultracentrifugation data for HDL cholesterol, VLDL cholesterol, or VLDL triglyceride concentrations did not reveal significant differences. Transaminases and y-glutamyltransferase Arabica oil increased serum ALT values, which were elevated by 17.9 ± 13.9 U/L (95% CI: 9.4, 28.4) from an average of 8.3 ± 3. U/L before treatment (n = 12) at the last blood sampling time before the intake of oil was stopped (Figure 3). This effect was detected during routine monitoring of plasma biochemistry while the study was ongoing. For safety reasons,

5 CHOLSTROL-RAISING FACTORS IN COFF respectively. There were no significant effects of the active treatments on any other biochemical variables monitored. -J Weeks FIGUR 3. A: Average serum alanine aminotrasferase concentrations during arabica (, n 12), robusta (, n 12), and placebo (, n 12) treatment. The diminishing number of subjects in the arabica group was due to cessation of treatment on the basis of serum alanine aminotransferase concentrations exceeding 25 UIL. B: Values for all subjects after discontinuation of administration. The solid line connects values of arabica group (). Six weeks after discontinuation only one blood sample was taken from all subjects in the robusta ( ) and placebo () groups. we decided to stop administration of oil to subjects with serum ALT values exceeding 25 UIL (the upper limit of normal values in our laboratory being 15 U/L). These decisions were made without breaking the study code, but later it was seen that elevated ALT activity occurred in the arabica group only. Consequently, the remaining subjects had lower ALT values and therefore the average values for weeks 5 and 6 are lower and based on a smaller number of subjects (Figure 3). For the robusta and placebo groups the average change in ALT was 2.7 ± 2.4 U/L (95% CI: -.9, 8.2) and - 1. ± 6.5 U/L, respectively. The effect of arabica oil on ALT was greater than that of robusta oil (+15.2 ± 4.1 U/L; 95%CI: 6.3, 24.2). The effects on AST were less pronounced. Arabica changed AST by 2.7 ± 4.4 U/L (95% CI: 1., 7.1), robusta by.8 ± 2. U/L (95 % CI:.4, 4.), and placebo by ± 2.1 UIL. No significant difference between the effects of arabica and robusta oils on AST was detected (+ 1.8 ± 1.4 UIL; 95% CI: -1.18, 4.9) (Table 3). No significant effects on y-glutamyltransferase were detected (P =.87, for overall analysis). The average changes from baseline were.2 ± 2.8 U/L in the arabica group and.6 ± 4. 1 and.8 ± 1.3 U/L in the robusta and placebo groups, Measurements after discontinuation of treatment The results of the biochemical measurements performed 6 wk after cessation of treatment were not significantly different from the pretreatment values. Additional blood samples were n-4 taken in the subjects from the arabica group at 3 and 7 d and 3 wk after treatment ended. Total serum cholesterol in the arabica group was 5.39 ±.85 mmolfl before and 5.35 ±.76 mmol/l 6 wk after treatment ended. A gradual decline was observed over the 6-wk period after discontinuation of treatment (Figure 1B). For the robusta group total cholesterol values before treatment and 6 wk after discontinuation were 4.78 ±.5 and 4.7 ±.51 mmol/l respectively, and for the placebo group were 5.16 ±.64 and 5.28 ±.77 mmol/l (overall analysis: P =.57). Average triglyceride concentrations in the arabica group were 1.5 ±.52 mmol/l before and 1.38 ±.97 mmol/l 6 wk after treatment ended, with a rather steep decline during the first week after discontinuation (Figure 2B). For the robusta group this was.99 ±.23 and 1.4 ±.29 mmol/l, respectively, and for the placebo group was 1.21 ±.38 and 1.38 ±.92 mmol/l (P.58 for overall analysis). For AST, average values of the arabica group before treatment and 6 wk after treatment were 8.2 ± 3. and.2 ± 3.6 U/L, respectively; for the robusta group they were 6.8 ± 1.9 and 8.7 ± 3.6 U/L; and for the placebo group they were 9.6 ± 8.2 and 8.2 ± 3.2 U/L (P.16) (Figure 3B). All other variables also tended to return to baseline in 6 wk. Body weight and dietary factors There were no effects of any of the treatments on body weight. At the start of treatment the total energy intake per 24 h was 664 ± 24 Id in the arabica group, 5397 ± 1813 kj in the robusta group, and 7269 ± 283 Id in the placebo group. The dietary factors calculated from the 24-h recalls (daily intake of alcohol, cholesterol, total energy, saturated fatty acids, monounsaturated and polyunsaturated fatty acids, and total fat) were analyzed for the same time points as were the biochemical variables. None of the dietary factors changed significantly during the three treatments. DISCUSSION This study demonstrated a strong effect of the oil extract from green arabica coffee beans on total serum cholesterol and triglycerides. Compared with peanut oil the absolute effect of arabica oil on cholesterol was more pronounced than that of robusta oil. Neither the absolute effect of robusta oil on total serum cholesterol nor the difference between arabica and robusta treatment was statistically significant. The rise in cholesterol was possibly caused mainly by a rise in LDL cholesterol, because the calculated LDL-cholesterol data showed a trend to rise, particularly in the arabica group. Despite the fact that no baseline values are available, the ultracentrifugation data also suggest a significant effect of arabica oil on LDL. In addition, apolipoprotein B was elevated significantly after arabica oil. In contrast with robusta oil, arabica oil also had a significant effect on serum triglycerides, which almost doubled during the treatment period in the arabica group.

6 1282 VAN ROOIJ T AL The effects on cholesterol and apolipoprotein B during treatment are comparable with those of intervention studies using boiled coffee (9, 15, 16) or using a lipid fraction extracted from it (2, 23). The coffee varieties used in these studies are most likely a blend of the two varieties except for the Scandinavian studies, in which arabica coffee was probably consumed. The effect of the arabica oil used in our study is somewhat stronger than that described earlier. This may be a matter of dosage. The lipid content of boiled coffee ranges from.7 to 2 g/l (16, 19) and the equivalent of the 2 g oil administered daily in our study would be L boiled caffeine-free coffee. An effect on triglycerides was not detected in studies of the effect of boiled coffee, but this finding agrees with studies in which a lipid fraction derived from it was administered (2, 23). The time necessary to reach the maximum effect of coffee oil on cholesterol and triglycerides was 3 wk and it took a similar amount of time for the measures to return to baseline. Arabica coffee oil also had an effect on serum ALT. Although robusta oil induced an increase of ALT as well, this increase was considerably smaller and was not statistically significant. In each treatment group the rise in AST was much smaller than the rise in ALT and was significant only after arabica oil. In a recent study effects on serum ALT activity were observed after administration of coffee oil, boiled coffee, and a kahweol-cafestol mixture (23). ALT elevations were comparable with the effect of arabica oil in this study and did not return to baseline values 4 wk after withdrawal of the treatment. In the present study, no significant differences in serum transaminase activities were detected 6 wk after discontinuation of treatment compared with baseline values. The effects of coffee oil on serum transaminases suggest transient and unspecified effects on hepatocytes. The changes do suggest a more general effect of coffee oil on the liver instead of a specific effect on hepatic lipoprotein metabolism. The rise in ALT observed in this study is of the same magnitude as the elevation of ALT observed with increasing body mass index or alcohol consumption of up to nine units a day (27). During administration of coffee oil rich in diterpenes a decrease in y-glutamyltransferase with a rebound rise after discontinuation of coffee oil administration was reported (23). In our study we did not observe any effect on y-gt, which may be explained by our instructions to the subjects not to consume more than two units of alcohol on the evening preceding every blood sampling. The present study does not confirm the small decrease in serum creatinine concentrations reported (- 3 to -7 mol/l) (23). The clinical significance of small changes in creatinine concentrations are debatable especially if no data on dietary factors are provided and no creatinine clearance was calculated. At present it cannot be determined whether the effect on lipoproteins is mediated by the same factor as the effect on transaminases, as would be suggested by the fact that the pattern of change of transaminases and of serum lipids followed the same time course during and after treatment. Because an elevation of serum lipids is not always paralleled by elevated liver enzymes (28-3), the possibility that induction of liver enzymes may lead to higher serum lipids remains speculative. ffects on serum cholesterol, apolipoprotein B, and transaminases were also described for amiodarone (31, 32) and isotretinoin (33). The mechanism of these effects is also unknown. The data from this study suggest that the factor responsible for the effect on serum lipids is not formed during the roasting process, because the oil we used was extracted from green coffee beans. The clear effect on serum lipids found in this study is comparable with that found in studies of boiled (roasted) coffee. This finding renders it unlikely that the small amounts of pyrolytic products (monodehydrated fatty acid esters of mainly kahweol and cafestol) in roasted coffee oil are of major importance to the cholesterol-raising effect. Oils from roasted arabica and robusta beans differ in their content of the diterpene alcohols and this also applies for green coffee oil (21, 22, 34). Arabica coffee contains fatty acid esters of kahweol and cafestol and robusta coffee contains cafestol esters in smaller amounts than does arabica, and only traces of kahweol esters. The routine roasting of coffee beans for taste optimization affects the amounts of diterpenes only to a limited extent (34). Because we detected a strong effect of the arabica oil on serum lipids and ALT, and a smaller effect of robusta oil that was not significant, it is likely that the kahweol fatty acid esters were at least partly responsible for these effects. However, the data do not exclude the cafestol esters as candidates; the differences in effect between the two types of oil may be a result of the difference in cafestol content. Data that are suggestive of a dose-response relationship for effects of both kahweol and cafestol on cholesterol and ALT were reported ( 23), and considerable amounts of cafestol were detected in coffee varieties not associated with effects on cholesterol (24). For these reasons we consider both kahweol and cafestol to have the potential to raise serum cholesterol and transaminases, in contrast with the suggestion by other authors (23) that cafestol would be the main compound responsible for the effects. Published animal studies involving kahweol and cafestol are sparse and are mainly focused on the anticarcinogenic effects and induction of glutathione-s-transferase by the two compounds. No toxic effects were detected in these experiments (35). The coffee oil content of brewed coffee can be reduced extensively by filtering it through paper (16, 19), a procedure that is likely to reduce the effects of coffee on transaminases and lipids to a clinically insignificant level (17). Our results may lead to further elucidation of the mechanisms involved, which may be studied in animals or liver cell culture. Finally, quantitation of kahweol and cafestol esters may be used to estimate the effect of coffee produced by different brewing methods without having to perform studies with large numbers of subjects. U We are grateful to Arnoud van der Laarse of the Laboratory for Cardiobiochemistry (Department of Cardiology, University Hospital Leiden, Netherlands) and Henry CR Brandenburg (Department of Clinical Pharmacy, University Hospital Leiden) for the preparation of the coffee oil emulsions. RFRNCS 1. LaCroix AZ, Mead LA, Liang KY, Thomas CB, Pearson TA. Coffee consumption and the incidence of coronary heart disease. N ngl I Med 1986;315: Grobbee D, Rimm B, Giovannucci, Colditz G, Stampfer M,

7 CHOLSTROL-RAISING FACTORS IN COFF 1283 Willett W. Coffee, caffeine, and cardiovascular disease in men. N ngi I Med 199;323: Dawber TR, Kannel WB, Gordon T. Coffee and cardiovascular disease. Observations from the Framingham Study. N ngl I Med 1974; Myers MG, Basinski A. Coffee and coronary heart disease. Arch Intern Med 1992;152: Lindahl B, Johansson I, Huhtasaari F, Hallmans G, Asplund K. Coffee drinking and blood cholesterol-effects of brewing method, food intake and life style. I Intern Med 1991 ;23: Tverdal A, Stensvold I, Solvoll K, Foss OP, Larssen PL, Bjartveit K. Coffee consumption and death from coronary heart disease in middle aged Norwegian men and women. Br Med I 199;3: Thelle DS, Heyden 5, Fodor 1G. Coffee and cholesterol in epidemiological and experimental studies. Atherosclerosis 1987;67: Aro A, Tuomilehto I, Kostianen, Uusitalo U, Pietinen P. Boiled coffee increases serum low density lipoprotein concentration. Metabolism 1987;1 1: Kark ID, Friedlander Y, Kaufmann NA, Stein Y. Coffee, tea, and plasma cholesterol: the Jerusalem Lipid Research Clinic prevalence study. Br Med J 1985;291: Stensvold I, Tverdal A, Foss OP. The effect of coffee on blood lipids and blood pressure. Results from a Norwegian cross-sectional study, men and women, 4-42 years. I Clin pidemiol 1989;42: Bonaa K, Amesen, Thelle DS, Forde OH. Coffee and cholesterol: is it all in the brewing? Br Med I 1988;297: Thelle DS, Arnesen, Forde OH. The Tromso heart study. Does coffee raise serum cholesterol? N ngl J Med 1983;38: Tuomilehto I, Tanskanen A, Pietinen P, et al. Coffee consumption is correlated with serum cholesterol in middle-aged Finnish men and women. J pidemiol Community Health 1987;41: Thelle DS. Coffee and cholesterol: what is brewing? I Intern Med 1991;23:289-91(editorial). 15. Forde OH, Knutsen SF, Arnesen, Thelle DS. The Tromso heart study: coffee consumption and serum lipid concentrations in men with hypercholesterolaemia: a randomised intervention study. Br Med J 1985;29: van Dusseldorp M, Katan MB, van Vliet 1, Demacker PNM, Stalenhoef AFH. Cholesterol-raising factor from boiled coffee does not pass a paper filter. Atheroscler Thromb 1991;11: Bak AA, Grobbee D. The effect on serum cholesterol levels of coffee brewed by filtering or boiling. N ngi I Med 1989;321: Superko HR. Bortz W, Williams PT, Albers II, Wood PD. Caffeinated and decaffienated coffee effects on plasma lipoprotein cholesterol, apolipoproteins, and lipase activity: a controlled, randomized trial. Am J Clin Nutr 1991;54: Ahola I, Jauhiainen M, Aro A. The hypercholesterolaemic factor in boiled coffee is retained by a paper filter. I Intern Med 1991;23: Zock PL, Katan MB, Merkus MP, van Dusseldorp M, Harryvan IL. ffect of a lipid-rich fraction from boiled coffee on serum cholesterol. Lancet 199;335: Folstar P. Chemistry. In: Clarke Ri, Macrae R, eds. Coffee. lsevier Applied Science, Speer K, Tewis R, Montag A. A new roasting component in coffee. In: Proceedings of the 14th International Scientific Colloquium on Coffee (San Francisco 1991). Paris: Association Scientifique Internationale du Caf#{233}, 1992: Weusten-van de Wouw MPM, Katan MB, Viano R, et al. Identity of the cholesterol-raising factor from boiled coffee and its effects on liver function enzymes. I Lipid Res 1994;35: Heckers H, G#{246}belU, Kleppel U. nd of the coffee mystery: diterpene alcohols raise serum low-density lipoprotein cholesterol and triglyceride levels. I Intern Med 1994;235: Ciba-Geigy. Geigy scientific tables. 8th ed. Basel, Switzerland: Ciba- Geigy Limited, Friedewald \V I, Levy RI, Frederickson DS. stimation of the concentration of low-density lipoprotein cholesterol in plasma without use of the preparative ultracentrifuge. Clin Chem 1972;18: Robinson D, Whitehead TP. ffect of body mass and other factors on serum liver enzyme levels in men attending for well population screening. Ann Clin Biochem 1989;26: LaPorte R, Valvo-Gerard L, Kuller L, et al. The relationship between alcohol consumption, liver enzymes and high-density lipoprotein cholesterol. Circulation 1981;64(suppl III): Pettersson P, llsinger BM, Sjoberg C, Bjorntorp P. Fat distribution and steroid hormones in women with alcohol abuse. I Intern Med l99;228: Frimpong NA, Lapp IA. ffects of moderate alcohol intake in fixed or variable amounts on concentration of serum lipids and liver enzymes in healthy young men. Am I Clin Nutr 1989;5: Roche IF, Netter P. Gay G. Amiodarone-induced hepatitis: biological, histological, diagnosis, development and role of associated factors. Review of the literature. Report of two cases. Rev Med Interne 1989; : Wiersinga WM, Trip MD, van Beeren MH, Plomp TA, Oosting H. An increase in plasma cholesterol independent of thyroid function during long term amiodarone therapy. Ann Intern Med 1991;l14: Shalita AR, Armstrong RB, Leyden II, Pochi P, Strauss IS. Isotretinoin revisited. Cutis 1988;42: Nackunz B, Maier HG. Diterpenoids in coffee. (Diterpenoide im Kaffee.) Z Lebensm Unters Forsch 1987;184:494-9(in German). 35. Fenwick LKT. Inhibition of carcinogenesis by some minor dietary constituents. Int Symp Princess Takamatsu Cancer Res Fund 1985;16: