Coffee consumption and mortality from cardiovascular diseases and total mortality: Does the brewing method matter?

Aim

The aim of this study was to investigate whether the coffee brewing method is associated with any death and cardiovascular mortality, beyond the contribution from major cardiovascular risk factors.

Methods and results

Altogether, 508,747 men and women aged 20–79 participating in Norwegian cardiovascular surveys were followed for an average of 20 years with respect to cause-specific death. The number of deaths was 46,341 for any cause, 12,621 for cardiovascular disease (CVD), 6202 for ischemic heart disease (IHD), and 2894 for stroke. The multivariate adjusted hazard ratios (HRs) for any death for men with no coffee consumption as reference were 0.85 (082–0.90) for filtered brew, 0.84 (0.79–0.89) for both brews, and 0.96 (0.91–1.01) for unfiltered brew. For women, the corresponding figures were 0.85 (0.81–0.90), 0.79 (0.73–0.85), and 0.91 (0.86–0.96) for filtered, both brews, and unfiltered brew, respectively. For CVD, the figures were 0.88 (0.81–0.96), 0.93 (0.83–1.04), and 0.97 (0.89–1.07) in men, and 0.80 (0.71–0.89), 0.72 (0.61–0.85), and 0.83 (0.74–0.93) in women. Stratification by age raised the HRs for ages ≥60 years. The HR for CVD between unfiltered brew and no coffee was 1.19 (1.00–1.41) for men and 0.98 (0.82–1.15) for women in this age group. The HRs for CVD and IHD were raised when omitting total cholesterol from the model, and most pronounced in those drinking ≥9 of unfiltered coffee, per day where they were raised by 9% for IHD mortality.

Conclusion

Unfiltered brew was associated with higher mortality than filtered brew, and filtered brew was associated with lower mortality than no coffee consumption.

Ischemic heart disease, stroke, smoking, unfiltered brew, filtered brew

Introduction

Globally, coffee is the most frequently used central stimulant, and probably the safest.¹ It has been in common use for a little more than half a thousand years, and, during that time, has been claimed as a healthy remedy for any disorder (including the plague) and the cause of all evil. More current research is still divided, and the jury has not, for example, decided whether coffee consumption contributes to increased risk of ischemic heart disease (IHD) or not. Observational epidemiological studies assessing the association between coffee consumption and IHD mortality display an array of results. They range from showing no association^2,,,,–7 to a protective effect,^8,–10 an increased risk,^11,–13 and a J-shaped curve.¹⁴ Relevant meta-analyses on the same issue have resulted in estimates indicating no increased risk, U-shaped risk curves, or an inverse association.^15,,,,–20

The heterogeneity in these results indicates that there are population-specific factors that are affecting the possible association between coffee consumption and IHD, and that these factors are unevenly distributed across study populations. One particular factor might be the brewing method – that is, whether the final beverage has passed through a paper filter or not. Unfiltered coffee contains a substantial amount of the low-density lipoprotein (LDL)-cholesterol, increasing diterpenes kahweol and cafestol that might contribute, at least theoretically, to a higher IHD risk.^21,22 Meanwhile, possible protective substances or chemical components in coffee and the uneven distribution of confounders associated with both coffee and IHD should be considered.

Therefore, we studied the association between amount and type of coffee (filtered, unfiltered) and the risk of total, cardiovascular, IHD, and stroke mortality in a large cohort of middle-aged men and women, taking cigarette smoking and other potential confounders (mediators) into consideration, as well as blood lipids.

Methods

The data stem from the Norwegian Counties Study,²³ the Norwegian Age-40 Program,²⁴ and the CONOR Study, a pooling of compatible cohort studies in Norway.²⁵ We collected data from 1985 to 2003. There was 85% participation in the Norwegian Counties Study, 70% in the Age-40 Program, and 60% in CONOR. Altogether, 635,718 men and women aged 20–79 years participated, of which 600,197 were without a history of cardiovascular disease (CVD), diabetes, or cancer at baseline. Of these, 508,747 reported information on coffee and the other variables that entered into the multivariate analyses. History of CVD was self-reported, whereas we confirmed history of cancer by linkage to the Cancer Registry of Norway.

Exposure variables

The participants filled in a questionnaire at home, which was checked for inconsistencies at the screening site. The relevant questions for the present report concerned the amount and type of coffee consumed, smoking habits, and duration of formal education. Height (cm), weight (kg), and blood pressure were measured at the screening, and a non-fasting blood sample was drawn for total cholesterol and serum triglyceride determinations by an enzymatic method. The blood samples were centrifuged on the spot and sent to the laboratory for analyses. A Dinamap apparatus measured blood pressure three times. The mean of the two last readings was used in the analyses (more details reported in previous publications^23-25). In Norway, coffee is traditionally brewed using a paper filter, resulting in a drip-brewed beverage, or by directly letting the ground coffee beans simmer in close-to-boiling water. We refer to the first method as filtered coffee and the latter as unfiltered.

The questions about coffee read: Version (1) How many cups of coffee do you usually drink daily? (Write 0 if you don’t drink coffee): 1) unfiltered coffee (coarsely ground), number…, 2) coffee other, number…; Version (2) How many cups of coffee do you drink daily? 0 or less than 1, 1–4 cups, 5–8 cups, 9 or more cups. What type of coffee do you usually drink daily? Unfiltered (coarsely ground), filter, instant, decaffeinated coffee, do not drink coffee. Instant and decaffeinated coffee were categorized as filtered coffee. We categorized brewing method as filtered only, filtered and unfiltered, and unfiltered only. The number of coffee cups per day (/d) were 0, 1–4, 5–8, and ≥9.

Confounding variables

The following variables were included in the analysis as potential confounders: age, number of cigarettes/d, total cholesterol, triglycerides, systolic blood pressure, body mass index, education, and year of examination. Educational length comprised nine categories: <7 years, 7–9 years, ….17–18 years, and > 18 years; and physical activity was ≥1 hour vigorous physical activity per week or walking around, cycling, or other activity for at least four hours a week.

Follow-up and endpoint ascertainment

The participants were followed from date of participation to date of emigration, date of death, or December 31, 2013. CVD as the cause of death is defined by International Classification of Diseases (ICD)-8 390-458, ICD-9 390-459, ICD-10 I00-I99, by IHD ICD-8 and ICD-9 410-414, ICD-10 I20-I25, and stroke by ICD-8 and ICD-9 430-438, ICD-10 I60-I69. The number of deaths during follow-up were 46,341, 12,621 ,6202 and 2894 for any death, cardiovascular, ischemic, and stroke death, respectively. Doctors completed the death certificates with no further validation. All death certificates were sent to the Norwegian Cause of Death Registry.

Statistics

We adjusted the mean values and percentages across the coffee categories for age and sex. We used the “adjmean” procedure in STATA: “Adjmean calculates adjusted means and SEs from linear regression estimates for the coffee groups, adjusted for sex and age. The covariates are set at their mean values”.²⁶ To estimate hazard ratios (HRs) for mortality, we ran the Cox proportional hazards model. Both cups/d and brewing type were factor variables. No coffee consumption was the reference when comparing the mortality in this group with the mortality in the three brewing methods. Among coffee consumers, filtered brew and 1–4 cups/d was the reference for comparison with the mortality in the eight other brew–cups combinations. The final model included the cups/d, brewing method, total cholesterol, triglycerides, systolic blood pressure, body mass index, educational length, physical activity, use of antihypertensives, and study as covariates. “Study” comprised the Norwegian Counties Study, Age-40 Program, and CONOR, defined as factors. We estimated HRs with and without total cholesterol as the covariate. We evaluated the proportional hazards assumption by visual inspection of the plot –ln(-ln(Survival probability)) versus ln(t) by each level of a variable.²⁶ The graphs for CVD by brewing method and cups/d are depicted in Supplementary Figures 1 and 2. We assessed interaction between variables by the log likelihood ratio from models with and without term(s). For factorial interactions, we used the binary operator.²⁶ We considered p-values less than 0.05 statistically significant.

Results

The filtered brew was preferred by 59% of participants, whereas 20% preferred unfiltered brew, and 9% used both brews; 12% did not drink coffee (Table 1). The share of men was lower in three of the groups and higher in the mixed group. Total cholesterol and number of cigarettes per day was lowest in the no coffee group and highest in the unfiltered group. The cholesterol level differed by 0.49 mmol/l between the unfiltered and no coffee group. The educational length was highest among the coffee abstainers. Equality across the groups was rejected for all variables (p < 0.0001).

Table 1.

Baseline characteristics by brewing method. Coffee-drinking men and women aged 20–79 years, without a history of cardiovascular disease, diabetes, or cancer.

	No coffee	Filtered	Both brews	Unfiltered
N	59,787	300,168	44,342	104,450
Age	41	43	42	44
Male (%)	45	48	54	47
Sex- and age-adjusted
Means
Cholesterol (mmol/l	5.52	5.69	5.83	6.01
Triglycerides (mmol/l)	1.77	1.67	1.72	1.82
Systolic blood pressure (mmHg)	130	131	131	132
Body mass index (kg/m2)	25.4	25.1	25.2	25.3
N cigarettes/d	2.1	5.3	5.0	6.2
Percentages
Coffee cups/d
0	100	0	0	0
1–4	0	50	39	41
5–8	0	40	46	43
9+	0	10	15	15
Educational length
0–10 years	40	46	45	61
11–14 years	28	27	30	25
15+ years	32	26	25	14
Study participants
Norwegian counties	8	10	9	17
Age-40 Program	63	73	65	68
CONOR	29	17	26	14

Table 2 shows adjusted all-cause and cause-specific mortality according to brewing method. The mortality was lower in the coffee groups than in the no coffee group. However, many of the HRs were comparable with no difference from the no coffee group, especially among men. Filtered brew was associated with lower HRs for CVD in men, whereas all three brews were associated with significantly lower CVD mortality in women. Analyses of non-smokers only gave similar results (see Supplementary Table 3).

Table 2.

Hazard ratios (HRs) with 95% confidence intervals (95% CIs) by coffee brew. Men and women aged 20–79 years, without a history of cardiovascular disease, diabetes, or cancer.

	Men			Women
	Deaths	HRa	95% CI	Deaths	HRa	95% CI
Any death
No coffee	2101	Ref		1629	Ref
Filtered	14,723	0.85	0.82–0.90	10,593	0.85	0.81–0.90
Both brews	2030	0.84	0.79–0.89	1113	0.79	0.73–0.85
Unfiltered	7939	0.96	0.91–1.01	6213	0.91	0.86–0.96
CVD
No coffee	573	Ref		355	Ref
Filtered	4318	0.88	0.81–0.96	2316	0.80	0.71–0.89
Both brews	629	0.93	0.83–1.04	220	0.72	0.61–0.85
Unfiltered	2579	0.97	0.89–1.07	1631	0.83	0.74–0.93
IHD
No coffee	305	Ref		124	Ref
Filtered	2399	0.91	0.81–1.03	858	0.83	0.69–1.01
Both brews	358	0.94	0.81–1.09	80	0.73	0.55–0.97
Unfiltered	1451	1.00	0.89–1.14	627	0.84	0.69–1.03
Stroke
No coffee	127	Ref		114	Ref
Filtered	765	0.72	0.59–0.87	764	0.84	0.69–1.02
Both brews	110	0.84	0.65–1.08	61	0.64	0.47–0.88
Unfiltered	460	0.84	0.68–1.02	493	0.87	0.71–1.07

CVD: cardiovascular disease; IHD: ischemic heart disease.

^aAdjusted for age (years), triglycerides (mmol/l), number of cigarettes/d, systolic blood pressure (mmHg), body mass index (kg/m²), physical activity (yes/no), educational length (<7 years, 7-9, …17-18, >18 years), use of antihypertensives (yes/no), and total cholesterol (mmol/l).

Table 3 shows the HRs for any death and CVD death stratified by age at baseline. For men aged 60 years or older, there is no favorable effect of coffee drinking on cardiovascular mortality. Men drinking the unfiltered brew had a higher mortality than the men not drinking coffee. Setting the cut-off at 70 years raises the HR to 1.40 (1.04–1.88). For women, the HR did not change with 70 years cut-off.

Table 3.

Hazard ratios (HRs) with 95% confidence intervals (95% CIs) by coffee brew and two age groups. Men and women aged 20–79 years, without a history of cardiovascular disease or diabetes.

	<60 years			≥60 years
Age groups	Deaths	HRa	95% CI	Deaths	HRa	95% CI
Men
Any death
No coffee	1536	Ref		565	Ref
Filtered	10,636	0.87	0.82–0.92	4087	0.83	0.76–0.91
Both brews	1666	0.83	0.77–0.89	364	0.93	0.82–1.06
Unfiltered	5346	0.94	0.89–1.00	2593	1.02	0.93–1.12
CVD
No coffee	411	Ref		162	Ref
Filtered	3012	0.88	0.79–0.97	1306	0.91	0.77–1.07
Both brews	505	0.88	0.78–1.01	124	1.05	0.83–1.34
Unfiltered	1606	0.89	0.79–0.99	973	1.19	1.00–1.41
Women
Any death
No coffee	1151	Ref		478	Ref
Filtered	7636	0.88	0.83–0.94	2957	0.80	0.72–0.88
Both brews	837	0.79	0.72–0.87	276	0.83	0.71–0.96
Unfiltered	3916	0.90	0.84–0.96	2297	0.94	0.85–1.04
CVD
No coffee	188	Ref		167	Ref
Filtered	1266	0.81	0.69–0.94	1050	0.80	0.68–0.95
Both brews	112	0.61	0.48–0.77	108	0.94	0.74–1.20
Unfiltered	768	0.75	0.64–0.89	863	0.98	0.82–1.15

CVD: cardiovascular disease.

^aAdjusted for age (years), triglycerides (mmol/l), number of cigarettes/d, systolic blood pressure (mmHg), body mass index (kg/m²), physical activity (yes/no), educational length (<7 years, 7–9, …17–18, >18 years), use of antihypertensives (yes/no), and total cholesterol (mmol/l).

Among coffee consumers, the reference group of 1–4 cups/d of filtered brew had the lowest mortality and ≥9 cups/d of unfiltered brew had the highest mortality (Table 4). However, in women, the mortality in the mixed group was comparable to the mortality in the reference group for all three consumption levels. The HRs were generally higher when total cholesterol was not among the covariates. This was most distinct for unfiltered brew and IHD mortality. The HRs were 7% to 9% higher without cholesterol in men consuming 5–8 cups and ≥9 cups/d of unfiltered coffee, respectively. Corresponding figures for women were of the same magnitude.

Table 4.

Hazard ratios (HRs) with 95% confidence intervals (95% CIs) by brew and number of cups/d, with and without adjustment for total cholesterol. Coffee-drinking men and women aged 20–79 years, without history of cardiovascular disease or diabetes.

		Any death		Cardiovascular		Ischemic heart disease		Stroke
Brew	Cups/d	HRa	HRb (95% CI)	HRa	HRb (95% CI)	HRa	HRb (95% CI)	HRa	HRb (95% CI)
Men			n = 24,692		n = 7526		n = 4208		n = 1335
Filtered	1–4	Ref	Ref		Ref		Ref		Ref
Filtered	5–8	1.00	1.00 (0.92–1.08)	1.18	1.17 (1.02–1.35)	1.21	1.21 (1.00–1.47)	1.13	1.13 (0.80–1.59)
Filtered	9+	1.13	1.12 (1.08–1.18)	1.19	1.17 (1.08–1.27)	1.17	1.14 (1.02–1.27)	1.23	1.24 (1.03–1.48)
Both brews	1–4	1.02	1.02 (0.98–1.06)	1.11	1.10 (1.03–1.17)	1.13	1.12 (1.03–1.23)	1.07	1.07 (0.91–1.25)
Both brews	5–8	0.99	0.99 (0.92–1.06)	1.10	1.08 (0.95–1.22)	1.02	0.99 (0.83–1.18)	1.34	1.35 (1.02–1.80)
Both brews	9+	1.16	1.15 (1.10–1.20)	1.23	1.18 (1.09–1.27)	1.29	1.22 (1.10–1.35)	1.17	1.19 (0.99–1.42)
Unfiltered	1–4	1.08	1.08 (1.02–1.13)	1.13	1.11 (1.01–1.23)	1.19	1.18 (1.03–1.34)	1.09	1.10 (0.86–1.41)
Unfiltered	5–8	1.06	1.04 (0.95–1.15)	1.15	1.10 (0.92–1.31)	1.31	1.24 (1.00–1.54)	1.01	1.03 (0.64–1.66)
Unfiltered	9+	1.17	1.15 (1.09–1.22)	1.25	1.17 (1.05–1.29)	1.32	1.21 (1.06–1.39)	1.20	1.23 (0.95–1.58)
Women			n = 17,919		n = 4167		n = 1565		n = 1318
Filtered	1–4	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
Filtered	5–8	0.96	0.96 (0.87–1.06)	0.97	0.96 (0.78–1.18)	1.03	1.00 (0.71–1.41)	0.68	0.68 (0.45–1.03)
Filtered	9+	1.11	1.11 (1.06–1.16)	1.12	1.10 (1.01–1.20)	1.03	0.99 (0.85–1.15)	1.10	1.09 (0.94–1.28)
Both brews	1–4	1.09	1.09 (1.04–1.14)	1.13	1.12 (1.02–1.23)	1.17	1.15 (0.99–1.33)	1.03	1.03 (0.88–1.21)
Both brews	5–8	0.98	0.98 (0.89–1.07)	0.99	0.97 (0.79–1.19)	0.92	0.87 (0.61–1.24)	0.89	0.89 (0.61–1.29)
Both brews	9+	1.12	1.11 (1.06–1.16)	1.12	1.08 (0.98–1.20)	1.17	1.10 (0.94–1.28)	1.02	1.01 (0.85–1.21)
Unfiltered	1–4	1.16	1.16 (1.08–1.24)	1.22	1.21 (1.03–1.42)	1.11	1.08 (0.83–1.40)	1.14	1.13 (0.84–1.52)
Unfiltered	5–8	1.00	0.99 (0.84–1.17)	0.92	0.89 (0.59–1.33)	0.90	0.82 (0.44–1.55)	0.85	0.84 (0.40–1.78)
Unfiltered	9+	1.20	1.20 (1.10–1.30)	1.25	1.20 (1.01–1.42)	1.27	1.17 (0.90–1.51)	1.04	1.03 (0.74–1.43)

^aAdjusted for age (years), triglycerides (mmol/l), number of cigarettes/d, systolic blood pressure (mmHg), body mass index (kg/m₂), physical activity (yes/no), educational length (<7 years, 7-9, …17-18, >18 years) and use of antihypertensives (yes/no)).

^bAdjusted for age (years), triglycerides (mmol/l), number of cigarettes/d, systolic blood pressure (mmHg), body mass index (kg/m₂), physical activity (yes/no), educational length (<7 years, 7-9, …17-18, >18 years), use of antihypertensives (yes/no), and total cholesterol (mmol/l).

Discussion

We have found that filtered coffee is associated with lower mortality than no coffee or unfiltered coffee only. Furthermore, we have found that the raised IHD mortality for unfiltered brew is mediated partly through its association with total cholesterol.

The studies on the association between coffee intake and mortality have given diverging results, from favorable effect, via no effect, to unfavorable effect. Our study is comparable to all results. Drinking unfiltered coffee is favorable – it is not associated with elevated mortality, except among older men, where unfiltered brew is associated with elevated cardiovascular mortality.

The association between coffee consumption and total serum cholesterol has been known since 1983, and later studies identified the lipid-raising components of coffee, the diterpenes kahweol and cafestol.²¹ The concentration of these diterpenes in the final brew depends strongly on the brewing method, a finding that explained most of the heterogeneous results published regarding coffee intake and blood lipids. A portion of unfiltered brewed coffee contains about 30 times the concentration of the lipid-raising diterpenes kahweol and cafestol compared to filtered coffee.²⁷ However, the porosity of the filter and the particle size of ground roasted coffee are determinants of the cafestol content in filtered coffee.²⁸

The lower mortality associated with filtered coffee as compared with no coffee might arise from coffee being rich in antioxidants, including polyphenols. Polyphenols can inhibit oxidation of LDL, exert anti-thrombotic effects, and improve endothelial dysfunction. The evidence for this is from in vitro studies or animal studies.²⁹ Coffee consumption is also associated with lower risk of diabetes, which is a risk factor for CVD.^30,31 A selection to the no coffee group is another possibility. People may abstain from coffee due to some conditions apart from those we have accounted for.

Total homocysteine (tHcy) is associated with cardiovascular death in observational studies. Coffee drinking is associated with tHcy, especially in high coffee consumption.³² This association applies to both the unfiltered and filtered types.^33,–35 We do not know, however, if the unfiltered type specifically has an effect on tHcy. On the other hand, our findings of a favorable association between filtered coffee consumption and CVD mortality agree with findings of an association between higher consumption of filtered coffee and lower levels of markers of inflammation and endothelial dysfunction.³⁶

The tendency of a higher mortality at higher consumption levels fits with a finding from the Tromsø Heart Study where heavy coffee drinking might be associated with some unknown socioeconomic or regional factors.³⁷ The information about coffee as an exposure variable stems from a questionnaire filled in before and checked by a research nurse at the screening site. We only have information from one single occasion and misclassification may have occurred, especially if the intake varies by season.

We do not know the size of the cups varying from small mocha cups (50–80 ml) in parties to mugs at work (180–220 ml), averaging about 125 ml.³⁸ Furthermore, some people use milk, cream, or sugar in their coffee, and some have the habit of consuming sweets together with their coffee. These and other potentially dietary habits were not accounted for.

Another aspect is that people are likely to have changed their coffee consumption pattern during follow-up. Most notably, there has been a switch from unfiltered to filtered coffee.³⁸ Norsk Kaffeinformasjon reports that 4% of coffee consumers drink traditional boiled coffee and that more than 70% drink traditional filtered coffee.³⁹ Since the 1990s, the French press method has been in the market, but only 10% brew this way today. Espresso coffee is consumed by 15% of the Norwegian population. The pods (capsules) have obtained footing in Norwegian workplaces. Finally, 18% use instant coffee and only 0.7% of the coffee imported to Norway is decaffeinated.⁴⁰ French press, espresso, and pods contain lipid-raising substances. We do not know the pattern of changing coffee habits and to what extent these may have influenced our estimates. However, a change in favorable direction is suggested from the findings of a 12-year follow-up study, where an adverse association between coffee consumption and coronary heart disease mortality during the first six years of follow-up disappeared during the last six years of follow-up.⁴¹ We can only speculate as to why the men aged 60 and above fared worse with unfiltered coffee regarding cardiovascular deaths than the younger men. It may be that the older group were less inclined to change to the filtered brew, and that we are facing a sex- and age-specific selection towards the more favorable brews. This may also explain why women reporting both categories are doing better.

We do not know whether our findings are valid for other populations. However, it is reasonable to assume that the cholesterol raising effect of unfiltered coffee is generalizable. Thus, a high intake of unfiltered coffee could be most unfavorable in high-risk groups. In Norway, there is an excess risk of acute myocardial infarction among immigrants from South Asia and Former Yugoslavia, and, if appropriate, these groups may benefit from moderate consumption of filtered coffee.⁴²

One strength of the study is the large number of participants with complete follow-up with respect to mortality. The well-known potential confounders are registered or measured and taken into account.

We found a higher total and CVD mortality for unfiltered than for filtered coffee consumption. The lowest mortality was among consumers of 1–4 cups of filtered coffee per day.

Author contribution

AT contributed to the conception or design of the study and drafted the manuscript. RS contributed to the acquisition of data, interpretation of data, and critical revision of the article for important intellectual content. JMC and DST contributed to the interpretation of data and critical revision of the article for important intellectual content. All authors gave final approval of the article.

Acknowledgement

We thank the Norwegian Counties Study, the Age-40 Program, the CONOR study, and the Norwegian Cause of Death Registry.

Declaration of conflicting interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: DST has acted as a consultant on issues related to coffee consumption for the Norwegian Coffee Information (Norsk Kaffeinformasjon, European Coffee Brewing Centre). AT, RS and JMC have nothing to disclose.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.