Filtered Coffee Raises Serum Cholesterol: Results from a Controlled Study

ORIGINAL COMMUNICATION Filtered coffee raises serum cholesterol: results from a controlled study

E Strandhagen1* and DS Thelle1

1Department of Medicine, Cardiovascular Institute, Sahlgrenska University Hospital/O¨stra, Go¨teborg, Sweden

Objective: Earlier studies and trials have shown a serum cholesterol raising effect of unfiltered coffee, which is reduced by about 80% in filtered coffee. Recent cross-sectional studies and trials, however, have indicated that filtered coffee may have a more pronounced serum cholesterol raising effect than previously anticipated. The objective of this controlled study was to assess the effects of the intake and abstention of filtered brewed coffee on blood lipids. Design: A prospective, controlled study with four consecutive trial periods. The first and third periods were 3 weeks of total coffee abstention. The second and fourth periods consisted of 4 weeks with the subjects consuming 600 ml filter brewed coffee/ day. Setting: Free-living population. Volunteers. Subjects: A total of 121 healthy, nonsmoking men and women aged 29–65 y. Interventions: Not applicable. Main outcome measures: Serum total cholesterol, serum HDL cholesterol, serum triglycerides, serum lipoprotein (a) (Lp(a)), blood pressure and heart rate. Results: The two coffee abstention periods were associated with a decline in serum cholesterol of 0.22 mmol/l (95% CI À0.31, À0.13) and 0.36 mmol/l (95% CI À0.46, À0.26), respectively. Filtered coffee/day 600 ml increased serum cholesterol by 0.25 mmol/l (95% CI 0.15, 0.36) and 0.15 mmol/l (95% CI 0.04, 0.26) during the two coffee drinking periods. Conclusions: Coffee abstention for 3 weeks decreased total serum cholesterol by 0.22–0.36 mmol/l. A volume of 600 ml (about four cups) of filtered coffee/day during 4 weeks raised total serum cholesterol by 0.15–0.25 mmol/l. Sponsorship: None. European Journal of Clinical Nutrition (2003) 57, 1164–1168. doi:10.1038/sj.ejcn.1601668

Keywords: controlled study; filtered coffee; serum cholesterol; serum triglycerides; serum Lp(a)

Introduction reduced (Bak, 1990; Ahola et al, 1991; Van Dusseldorp et al, Almost two decades ago it was shown that the consumption 1991). of coffee was associated with total serum cholesterol (Thelle Bak (1990) showed in her meta-analysis that total et al, 1983). Further studies revealed that the cholesterol- cholesterol increased by an average of 0.008 mmol/l per raising effect was because of the natural content of the cup of filtered coffee, compared to 0.038 mmol/l per cup of diterpenes kahweol and cafestol in the green coffee beans unfiltered or boiled coffee. These figures are seemingly in (Zock et al, 1990; Weusten-Van der Wouw et al, 1994; Urgert contrast to the findings by Fried et al (1992) and to some & Katan, 1997). A number of studies showed that a major extent by Superko et al (1991) where filtered coffee was also part of the diterpenes is retained by a paper filter, whereby associated with a larger increase in serum cholesterol than the cholesterol-raising effect of coffee is substantially previously reported. Urgert and Katan (1997) commented on these discrepancies in their review on the cholesterol-raising effect of coffee and were inclined to ascribe these different *Correspondence: E Strandhagen, Cardiovascular Institute, Sahlgrenska findings to chance. In a recent study, however, we observed a Academy at Go¨teborg University, Sahlgrenska University Hospital/O¨ stra, decrease of 0.28 mmol/l after quitting drinking on average ¨ SE-416 85 Goteborg, Sweden. 4.9 cups of filtered coffee for 6 weeks (Christensen et al, E-mail: [email protected] Received 20 May 2002; revised 25 September 2000; 2001). This led us to assess the effects of the intake and accepted 26 September 2002 abstention of filtered brewed coffee on blood lipids in a Filtered coffee raises serum cholesterol E Strandhagen and DS Thelle 1165 controlled study, originally aiming at the effects of folate and After the first 3 weeks of coffee abstention, the participants pyridoxine supplementation on the homocysteine-raising were instructed to consume 600 ml coffee (four cups/day). effect of coffee (Strandhagen et al, 2002). This period was followed by another 3 weeks of coffee abstention; thereafter, the participants were again asked to take up drinking 600 ml of coffee/day for another 4 weeks. Methods Divergence from the four cups was reported. The partici- Study design pants were allowed to drink tea and other caffeine-contain- The study was organised as a prospective, controlled study ing beverages during the coffee-free periods. with four consecutive trial periods. The first and third periods were 3 weeks of total coffee abstention. The second and fourth periods consisted of 4 weeks with the subjects Effect variables consuming 600 ml filter brewed coffee/day. Nonfasting blood samples were drawn at inclusion and at 3, The main outcome or effect variable was total serum 7, 10 and 14 weeks after start of the study. Prior to analysis, cholesterol and the effect was assessed as the difference prepared serum was stored at À701 C. between the measurements at the beginning and the end of The blood samples were analysed for blood lipids (total the coffee-free periods (coffee abstention) and the difference cholesterol, HDL cholesterol, triglycerides, lipoprotein(a) between measurements at the beginning and at the end of (Lp(a)) and urate in serum. Serum cholesterol and triglycer- the 4 weeks of coffee consumption, see Figure 1. A trial ides were determined by an enzymatic procedure on a duration of 3–4 weeks has previously been shown to be Hitachi 917 analyzer. HDL cholesterol was determined sufficient to get an effect of coffee on serum cholesterol enzymatically after precipitation of VLDL, LDL and chylo- (Grubben et al, 2000; Christensen et al, 2001). micrones by a-cyclodextrinsulphate and dextransulphate. Determination of Lp(a) was done by the method Tint Elize Lp(a) of Biopool International. Serum urate was analysed by Subjects and procedure Hitachi 917 autoanalyser. Body mass index (BMI; kg/m2) was The participants were recruited by advertisement in the recorded once during the study. Blood pressure was recorded largest newspaper in Gothenburg. Inclusion criteria were age by manual device, and EKG and heart rate were recorded at range 30–65 y, free of clinically recognised chronic diseases all five visits. such as cardiovascular diseases, cancer, renal disorders, liver The dietary habits were assessed by dietary frequency disease and diabetes mellitus. They were not on antiepileptic questionnaires at the beginning of the study. A follow-up or cholesterol-lowering drugs, had been using coffee on a survey with special emphasis on changes in food habits regular basis for at least 5 y and were currently nonsmokers during the four different periods was undertaken. The (at least for the last 6 months). dietary questionnaire was based upon a Norwegian version, During the coffee drinking periods, the participants were which has been used in a number of previous studies (Nes instructed to drink about 600 ml filter brewed coffee/day et al, 1992). (four cups), according to standardised measures. The coffee was provided to guarantee that they were all exposed to the same brand and quality of filter brewed coffee. All partici- Statistical methods pants also got the same kind of standardised coffee filter and All analyses were performed using the SASrsoftware. The measuring spoon. statistical significance of the differences was assessed by The coffee filters used were of the brand Euro-Shopper, Student’s t-test except for Lp(a). For this variable, signed rank made by Indupa NV, Zaventem, Belgium. test was used. P-values o0.05 were considered statistically

3 weeks of 4 weeks of coffee 3 weeks of 4 weeks of coffee coffee abstention consumption (600ml/day) coffee abstention consumption (600ml/day)

w0 w1 w2 w3 w4 w5 w6 w7 w8 w9 w10 w11 w12 w13 w141

Blood samples2 Blood samples2 Blood samples2 Blood samples2 Blood samples2 Blood pressure, EKG Blood pressure, EKG Blood pressure, EKG Blood pressure, EKG Blood pressure, EKG

1 w = week 2 Blood samples: serum cholesterol, serum HDL cholesterol, serum triglycerides, serum urate, serum Lp(a) Figure 1 Study design.

European Journal of Clinical Nutrition Filtered coffee raises serum cholesterol E Strandhagen and DS Thelle 1166 significant. The 95% confidence intervals (CI) are presented. The decreases in HDL cholesterol were only 0.02 mmol/l Analysis of variance was performed to test whether the (95% CI À0.06, 0.02) and 0.09 mmol/l (95% CI À0.12, differences of the effect variables differed between the two À0.06), respectively. Serum triglycerides showed an incon- trial periods. sistent pattern with a significant decline during the first coffee abstention period, but a much smaller effect during the second abstention period. Serum urate fell by 11 mmol/l Results (95% CI À19.7, À2.4) and 15.1 mmol/l (95% CI À20.9, À9.3) Coffee intake and blood lipids, blood pressure and during the two abstention periods, both statistically sig- heart rate nificant. Differences in heart rate were small and insignif- A total of 156 people responded to the advertisement and of icant. Both systolic and diastolic blood pressure fell these 124 fulfilled the criteria and were able to take part. significantly during the first abstention period but the Three people decided to withdraw during the study, leaving a results were insignificant for the last abstention period. total of 121 participants. One participant was not able to The effects of consuming 600 ml filtered coffee/day for 4 take part during the first period and five people were not able weeks are shown in Table 2. to take part in the last period, which resulted in 120 Total serum cholesterol increased during both trial periods participants in the first trial period and 116 in the by 0.25 mmol/l (95% CI 0.15, 0.36) and 0.15 mmol/l (95% CI subsequent trial period. 0.04, 0.26), respectively. There was also an incrementing Serum total cholesterol decreased by 0.22 mmol/l (95% CI: effect on serum HDL cholesterol, serum triglycerides, serum À0.31, À0.13) and 0.36 mmol/l (95% CI: À0.46, À0.26) urate, systolic blood pressure and heart rate in both periods, during the two 3 weeks coffee abstention periods (Table 1). but the differences were significant only for serum urate in

Table 1 Effect of 3 weeks of coffee abstention in trial period 1(weeks 0–3), n=120 and trial period 2 (weeks 7–10), n=116.

Period 1 Period 2

week 0 x% week 3 x% Diff. 95% CI P-value week 7 x% week 10 x% Diff. 95% CI P-value (s.d.) (s.d.) (s.d.) (s.d.)

Serum total cholesterol 5.21 (0.92) 4.99 (0.89) À0.22 À0.31, À0.12 o0.0001 5.24 (0.93) 4.88 (0.88) À0.36 À0.46, À0.26 o0.0001 (mmol/l) Serum HDL cholesterol 1.54 (0.38) 1.52 (0.38) À0.02 À0.06, 0.02 0.301 1.59 (0.39) 1.49 (0.39) À0.09 À0.12, À0.06 o0.0001 (mmol/l) Serum triglycerides 1.38 (0.81) 1.24 (0.76) À0.15 À0.25, À0.05 0.004 1.30 (0.73) 1.23 (0.72) À0.07 À0.17, 0.03 0.175 (mmol/l) Serum urate (mmol/l) 269 (68.9) 258 (70.6) À11.0 À19.7, À2.4 0.012 270 (73.4) 257 (70.2) À15.1 À20.9, À9.3 o0.0001 Diastolic blood pressure 78.8 (11.0) 76.7 (9.9) À2.0 À3.5, À0.5 0.012 76.3 (9.0) 77.1 (9.5) 0.8 À0.5, 2.1 0.237 (mmHg) Systolic blood pressure 125.6 (17.3) 121.0 (15.9) À4.5 À6.4, À2.5 o0.0001 122.4 (15.1) 120.6 (16.5) À1.9 À3.9, 0.2 0.070 (mmHg) Heart rate (beat/min) 66.9 (8.9) 66.0 (8.6) À1.0 À2.4, 0.4 0.157 66.8 (9.5) 66.6 (9.0) À0.2 À1.7, 1.4 0.828

Table 2 Effect of four cups (600 ml) of filtered coffee/day for 4 weeks in trial period 1 (weeks 3–7), n=120 and trial period 2 (weeks 10–14), n=116.

Period 1 Period 2

week 3 x% week 7 x% Diff. 95% CI P-value week 10 x% week 14 x% Diff. 95% CI P-value (s.d.) (s.d.) (s.d.) (s.d.)

Serum total cholesterol 4.99 (0.89) 5.24 (0.93) 0.25 0.15, 0.36 o0.0001 4.88 (0.88) 5.03 (0.86) 0.15 0.04, 0.26 0.006 (mmol/l) Serum HDL cholesterol 1.52 (0.38) 1.59 (0.39) 0.07 0.03, 0.11 0.0005 1.49 (0.39) 1.52 (0.39) 0.02 À0.01, 0.05 0.261 (mmol/l) Serum triglycerides 1.24 (0.76) 1.30 (0.73) 0.06 À0.04, 0.17 0.249 1.23 (0.76) 1.32 (0.80) 0.09 À0.02, 0.21 0.122 (mmol/l) Serum urate (mmol/l) 257.4 (70.3) 270.3 (73.3) 12.9 5.7, 20.1 0.0006 257 (70.2) 262.5 (65.7) 5.5 À0.3, 11.3 0.065 Diastolic blood pressure 76.7 (9.9) 76.3 (9.0) À0.5 -2.0, 1.1 0.544 77.1 (9.5) 77.9 (8.7) 0.9 À0.6, 2.4 0.248 (mmHg) Systolic blood pressure 121.0 (15.9) 122.4 (15.1) 1.3 À0.6, 3.3 0.189 120.6 (16.5) 122.5 (15.3) 2.1 0.2, 4.0 0.030 (mmHg) Heart rate (beat/min) 65.9 (8.6) 66.8 (9.5) 0.8 À0.4, 2.1 0.2 66.6 (9.0) 66.8 (9.4) 0.3 À1.2, 1.7 0.723

European Journal of Clinical Nutrition Filtered coffee raises serum cholesterol E Strandhagen and DS Thelle 1167 the first period (12.9 mmol/l, 95% CI 5.7, 20.1) and systolic Table 3 Baseline characteristics of the participants, n=121 blood pressure in the second period (2.1 mmHg, 95% CI). The decline in total cholesterol during coffee abstention was Age (y) (mean (range)) 48.6 (29–65) Women (%) 78 significantly larger during the second trial period, whereas Men (%) 22 the cholesterol increase was larger during the first coffee BMI (kg/m2) (mean 7s.d.) 25.773.4 exposure period. Coffee consumption prior to trial The Lp(a) values were not normally distributed, and only Cups, mean (s.d.) 3.9 (1.83) Filter brewed coffee, n (%) 89 (73) median values and range are presented here. There was no Instant coffee, n (%) 9 (7) difference in the first coffee abstention period (median Unfiltered coffeea, n (%) 15 (12) +2 mmol/l, range 256, À109) but a significant increase in the Combination, n (%) 8 (7) second abstention period (median +15 mmol/l, range 126, a À297). Coffee consumption resulted in significant decreases Boiled, percolator, French pressed. of serum Lp(a) in both trial periods (median À11 mmol/l, range 180, À108 in the first trial period and median and 83 of 116 participants showed a decrease in serum À4 mmol/l, range 338, À208 in the second trial period). cholesterol. Coffee drinking resulted in increasing serum None of the above results were influenced by the folate or cholesterol values in 81 of 120 participants in the first pyridoxine supplementation. period, and in 64 of 116 the participants in the second coffee period.

Dietary monitoring and compliance Of the 121 participants, 89 (74%) reported filter brewed Discussion coffee as their usual coffee. Unfiltered coffee (boiled, The effect of abstaining from filtered coffee for 3 weeks percolator, French pressed) was consumed by 15 people resulted in a serum cholesterol decrease of about 0.22– (12%) and nine people (7%) consumed instant coffee only 0.36 mmol/l, whereas 600 ml (about 4 cups) of filtered (Table 3). The average coffee intake prior to the study was 3.9 coffee/day resulted in an increase by 0.15–0.25 mmol/l. cups/day (s.d. 1.83). Mean levels of BMI and age, and gender These changes are unlikely a consequence of chance or distribution, are shown in Table 3. confounding dietary factors such as changes in other food The dietary survey showed that the intake of buns and items corresponding to coffee intake. Coffee abstention may cakes was unchanged in 100 of the 120 participants in the lead to either a reduction in milk consumption (coffee with first coffee abstention period, and in 107 of the 116 milk) or an increase to compensate a decreased fluid intake. participants for the second period (Table 4). For the coffee In this study, more than 80% of the participants reported no drinking period, the corresponding figures were 104 of 120, change of milk consumption in the coffee abstention and 110 of 116, respectively. Intake of milk was unchanged periods. As a comparison, 600 ml of whole fat milk (3% fat, in 97 of 120 participants and 99 of 116 for the two coffee daily energy intake 2000 kcal) should itself result in a change abstention periods, and in 109 of 120 and 112 of 116 in serum cholesterol by 0.34 mmol/l, according to the participants for the two coffee drinking periods. formula of Keys et al (1965). Coffee consumption or noncompliance was reported by The blood lipid changes in the present study were six people during the first coffee abstention period (mean 1.8 consistent with the results of Superko et al (1991), Fried cups/period), whereas four people reported coffee consump- et al (1992) and Christensen et al (2001). In the study by Fried tion in the second coffee abstention period (mean 0.7 cups/ et al, total cholesterol increased by 0.24 mmol/l after the period). consumption of 7.2 dl (five cups) of filtered coffee for 8 weeks, whereas Christensen et al (2001) showed a decrease of 0.28 mmol/l after quitting drinking on average 4.9 cups of Consistency in the effects filtered coffee for 6 weeks Superko et al (1991) reported that A majority of the participants showed similar changes in the consumption of decaffeinated coffee resulted in an blood lipids. In the two coffee abstention periods, 73 of 120 increment in serum cholesterol compared to regular coffee

Table 4 Compliance (no-change) in intake of milk and buns/cakes during the trial

First period, n=120 Second period, n=116

Coffee-free period Coffee period Coffee-free period Coffee period

Compliance in intake of milk, n (%) 97 (80.8) 109 (90.8) 99 (85.3) 112 (96.6) Compliance in intake of buns and cakes, n (%) 100 (83.3) 104 (86.7) 107 (92.2) 110 (94.8)

European Journal of Clinical Nutrition Filtered coffee raises serum cholesterol E Strandhagen and DS Thelle 1168 or no coffee at all. The differences were in both cases Coffee consumption, therefore, should be discussed in the 0.21 mmol/l. This study differed from the other studies in advising of hyperlipaemic subjects who may do well by design, but the results were of the same magnitude. reducing their coffee consumption. The serum cholesterol decrease in the present study was larger in the second coffee abstention period, most likely because of a more standardised coffee exposure prior to the References coffee abstention. Jee et al (2001) showed in a meta-analysis a Ahola I, Jauhiainen M & Aro A (1991): The hypercholesterolaemic change in serum cholesterol caused by unfiltered coffee by factor of boiled coffee is retained by a paper filter. J. Intern. Med. 0.6 mmol/l when comparing coffee drinkers with noncoffee 230, 293–297. Bak A (1990): Coffee and cardiovascular risk; an epidemiological drinkers. The use of filtered coffee resulted only in a study. Erasmus University, Rotterdam. 0.08 mmol/l change. This meta-analysis contained 18 studies Christensen B, Mosdl A, Retterstl L, Landaas S & Thelle DS (2001): including a total of 885 participants. In the 10 studies of Abstention from filtered coffee reduces the levels of homocysteine F Am. J. Clin. Nutr filtered coffee, 579 people participated. Compared to and cholesterol a randomized, controlled trial. . 74, 302–307. that, the present study as well as that by Fried et al Fried RE, Levine DM, Kwiterovich PO, Diamond EL, Wilder LB, Moy (1992), Christensen et al (2001) and Superko et al (1991) TF & Pearson TA (1992): The effect of filtered-coffee consumption included 701 participants. The 18 studies in the Meta- on plasma lipid levels. Results of a randomized clinical trial. JAMA 267, 811–815. analysis are likely to be more heterogeneous than the Grubben MJ, Boers GH, Blom HJ, Broekhuizen R, de Jong R, van Rijt three larger studies, which have shown an effect of filtered L, de Ruijter E et al. (2000): Unfiltered coffee increases plasma coffee. homocysteine concentrations in healthy volunteers: a rando- Another possible explanation is that the quality of the mized trial. Am. J. Clin. Nutr. 71, 480–484. International Coffee Organization (ICO): ICO document ‘‘Coffee paper filter differs, whereby more or less of the diterpenes statistics No 21’’, London (2001). pass through the filters and into the coffee. A further study of Jee SH, He J, Appel LJ, Whelton PK, Suh I & Klag MJ (2001): Coffee the paper quality and the physical properties of the filters is consumption and serum lipids: a meta-analysis of randomized therefore warranted. controlled clinical trials. Am. J. Epidemiol. 153, 353–362. Keys A, Anderson JT & Grande F (1965): Serum cholesterol response The magnitude of the serum cholesterol changes in to changes in the diet. Metabolism 14, 747–787. the present study corresponds to the effect of unfiltered Nes M, Frost Andersen L, Solvoll K, Sandstad B, Hustvedt BE, Lv A coffee in an earlier meta-analysis (Bak, 1990), the review & Drevon C (1992): Accuracy of a quantitative food frequency of Urgert and Katan (1997) and results from the Swedish questionnaire applied in elderly Norwegian women. Eur. J. Clin. Nutr. 46, 809–821. part of the MONICA study (Strandhagen et al 2000, Pietinen P, Vartiainen E, Seppa¨nen R, Aro A & Puska P (1996): abstract). Changes in diet in Finland from 1972 to 1992: impact on coronary Our results with regard to Lp(a) are inconsistent in the heart disease risk. Prev. Med. 25, 243–250. sense that Lp(a) did not change during the first coffee Strandhagen E, Landaas S & Thelle DS (2002): Folic acid supplement decreases the homocysteine increasing effect of filtered coffee. A abstention period but increased during the next noncoffee randomised placebo controlled study. Eur J Clin Nutr, in press. period. The Lp(a) levels fell, on the other hand during both Strandhagen E, Rosengren A & Thelle DS (2000): Coffee consump- the coffee periods, which corresponds to Urgert et al (1997) tion and blood lipids revisited. 7th Nordic Congress of Nutrition, who showed an Lp(a)-reducing effect of diterpenes. The A˚ land, Finland, Poster (abstract). Superko HR, Bortz Jr W, Williams PT, Albers JJ & Wood PD (1991): standard deviations of Lp(a) were very wide in the present Caffeinated and decaffeinated coffee effects on plasma lipoprotein study, indicating a large interindividual variation. Thus, the cholesterol, apolipoproteins, and lipase activity: a controlled, power to show true effects of low-diterpene-containing randomized trial. Am. J. Clin. Nutr. 54, 599–605. coffee on Lp(a) is likely to be small. Thelle DS, Arnesen E & Frde OH (1983): The Troms heart study. Does coffee raise serum cholesterol. N. Engl. J. Med. 308, 1454– The public health consequences of the present findings 1457. can be assessed by the study of Pietinen et al (1996) of Urgert R & Katan B (1997): The cholesterol-raising factor from coffee dietary changes in the Finnish diet during the years 1972– beans. Annu. Rev. Nutr. 17, 305–324. 1992. They assumed that filter brewed coffee had a Urgert R, Weusten-van der Wouw MPME, Hovenier R, Meyboom S, Beynen AC & Katan MB (1997): Diterpenes from coffee beans lower cholesterol increasing effect than unfiltered coffee, decrease serum levels of lipoprotein(a) in humans: results from whereby 0.3 mmol/l of the serum cholesterol reduction four randomised controlled trials. Eur. J. Clin. Nutr. 51, 431–436. of about 1 mmol/l on the Finnish population could be Van Dusseldorp M, Katan MB, Van Vliet T, Demacker PNM & ascribed to changes in coffee brewing habits. Such differ- Stalenhoef A (1991): Cholesterol-raising factor from boiled coffee does not pass a paper filter. Arterioscler. Thromb. 11, 586–593. ences correspond to 10–15% higher or lower incidence Weusten-Van der Wouw MPME, Katan MB, Viani R, Huggett AC, of coronary heart disease. The possibility that filter Liardon R, Lund-Larsen PG, Thelle DS, Ahola I, Aro A, Meyboom S brewed coffee raises serum cholesterol levels, to the extent & Beynen AC (1994): Identity of the colesterol-raising factor from which is shown in this paper, is particularly of interest in the boiled coffee and its effects on liver function enzymes. J. Lipid. Res. 35, 721–733. Nordic countries having the world’s highest coffee con- Zock PI, Katan ME, Merkus P, van Dusseldorp M & Harryvan JL. sumption (kg/person/y) (International Coffee Organization, (1990): Effect of a lipid-rich fraction from boiled coffee on serum 2001). cholesterol. Lancet 335, 1235–1237.

European Journal of Clinical Nutrition