Industrial Health 2000, 38, 99–102
Effects of Coffee Consumption against the Development
of Liver Dysfunction: A 4-Year Follow-Up Study of
Middle-Aged Japanese Male Office Workers
Noriyuki NAKANISHI1*, Koji NAKAMURA2, Kenji SUZUKI3 and Kozo TATARA1
1 Department of Social and Environmental Medicine, Course of Social Medicine, Osaka University Graduate
School of Medicine F2, 2–2 Yamada-oka, Suita-shi, Osaka 565-0871, Japan
2 Medical Office, Osaka Main Office, Takenaka Corporation, 4–1–13 Honmachi, Chuo-ku, Osaka 541-0053, Japan
3 Japan Labor and Welfare Association, 1–24–4 Ebisu, Shibuya-ku, Tokyo 150-0013, Japan
Received September 7, 1999 and accepted November 17, 1999
Abstract: The association of coffee consumption with the development of increased serum aspartate
aminotransferase (AST) and/or alanine aminotransferase (ALT) activities over 4 years was studied
in 1221 liver dysfunction-free (serum AST and ALT ?39 IU/l and no medical care for or no past
history of liver disease) Japanese male office workers aged 35 to 56 years. From the analysis using
the Kaplan-Meier method, the estimated incidence of serum AST and/or ALT ?40 IU/l, ?50 IU/l, and
?60 IU/l decreased with an increase in coffee consumption. From the Cox proportional hazards
model, coffee drinking was independently inversely associated with the development of serum AST
and/or ALT ?40 IU/l (p=0.019 by test for tendency), ?50 IU/l (p=0.002), and ?60 IU/l (p=0.007),
controlling for age, body mass index, alcohol intake, and cigarette smoking. These results suggest
that coffee may be protectively against the liver dysfunction in middle-aged Japanese men.
Key words: Coffee, Follow-up study, Liver dysfunction, Aspartate aminotransferase, Alanine
Recent epidemiological studies have suggested possibly
male office workers aged 35 to 56 years in May 1994 in T
beneficial effects of coffee consumption on the occurrence
Corporation, Osaka. During the initial and serial annual
of alcoholic cirrhosis1) and on serum liver enzymes,
examinations, fasting blood samples were drawn from an
particularly serum gamma-glutamyltransferase activity2–9).
antecubital vein for the determination of serum liver enzyme
Many cross-sectional studies investigated the association
levels. Serum AST and ALT concentrations were assayed
between coffee consumption and serum liver enzyme levels
at the Nihon Clinical Laboratories Inc. (Tokyo, Japan), based
in various populations, but few longitudinal studies have
on the methods recommended by the Japan Society of Clinical
been completed3). In this report on a longitudinal population
Chemistry10), with an Olympus AU-5000 in 1994 and an
study based on annual health examinations at the workplace,
Olympus AU-5200 in 1995-1998 (Olympus Japan Co., Ltd.,
we investigated the association between coffee consumption
Tokyo, Japan). The cut-off values for serum AST and ALT
and the development of increased serum aspartate
were set at ?39 IU/l. Quality control of the laboratory was
aminotransferase (AST) and alanine aminotransferase (ALT)
maintained by internal method, and the inter- and intraassay
activities in middle-aged Japanese men.
coefficients of variation for serum AST and ALT were within
The surveillance population consisted of 1464 Japanese
3% from 1994 to 1998. Of a total of 1464 subjects, 182
(12.4%) were identified to have increased levels of serum
*To whom correspondence should be addressed.
AST and/or ALT ?40 IU/l at the initial examinations. We
N NAKANISHI et al.
excluded these 182 subjects to avoid a possible bias,
kg/m2). With respect to age, BMI, alcohol intake, smoking
suggested by Klatsky and Armstrong1), that men with liver
habits, and coffee drinking at entry, the 97 cases which could
disease or liver dysfunction might have reduced coffee
not be followed up until May 1998 did not differ significantly
consumption because of impaired caffeine clearance11). The
from the 1124 cases which could.
32 subjects who had medical care for or past history of liver
As for analytic procedures, the observation times were
disease showed normal AST and ALT values. Among the
calculated by using the date of the initial examination and
1250 liver dysfunction-free (serum AST and ALT (39 IU/l
the date of the incidence of liver dysfunction or the date of
and no medical care for or no past history of liver disease)
follow-up (the fifth examination), or the date of last
subjects, 29 were taking medication for digestive disease.
registration in T Corporation, Osaka. Those who had been
We excluded these 29 subjects, since those with digestive
transferred to another locality or had resigned during the
disease might have reduced coffee consumption. The
follow-up period have censored observation times as do those
remaining 1221 constituted the study cohort and were
members of the cohort who were still in T Corporation, Osaka,
followed up for four years with annual examinations until
at the end of the follow-up. The mean observation period
May 1998. Subjects who were found to have increased levels
of this cohort was 3.7 years with SD of 0.02 years, and the
of AST and/or ALT ?40 IU/l during repeat surveys were
mean (SD) number of times of the measurement of serum
defined to be incidental cases of liver dysfunction, but no
AST and ALT, including the initial examination, was 4.8
one started medication for liver dysfunction or disease during
(0.02). The Kaplan-Meier method was used to estimate the
the observation period. Follow-up until May 1998 could
cumulative incidence of increased serum AST and/or ALT
be completed for only 1124 subjects (92.1%).
levels according to coffee drinking, and the log-rank test
The health examinations at entry included a questionnaire
was used to assess the significance of the unadjusted
on lifestyle and physical examinations. Data on lifestyle
differences among the incidence curves. The Cox
were obtained by interview and selected items of lifestyle
proportional hazards model was used to evaluate the
for this study were alcohol intake, cigarette smoking, and
multivariate relation between coffee drinking and the
coffee drinking. As for alcohol intake, the subjects were
development of increased serum AST and/or ALT levels,
asked to quantify their consumption of alcohol in respect of
controlling for age (exact values), BMI (exact values), alcohol
how much sake (Japanese rice wine) they drank per day.
intake, and cigarette smoking. As for alcohol intake, cigarette
Alcohol intake was then converted into the equivalent number
smoking, and coffee drinking, hazard ratio (HR) estimates
of go, a traditional Japanese unit of volume for sake (1 go=180
compared to the reference level of each subclass were
ml and contains 23 g of ethanol). A question queried about
calculated by creating two dummy variables for each variable
smoking habits (never, past or current smokers), with
as follows: x1=0, x2=0 for the reference level; x1=1, x2=0
subsequent inquiry to past or current smokers about the
for the second level; and x1=0, x2=1 for the third level.
number of cigarettes smoked per day. In this study, past
Tests for heterogeneity tendency of the HR estimates were
and never smokers were combined, and the current amount
also performed. All reported p-values are two-tailed and
of cigarettes smoked was used in the analysis. As for coffee
the level of significance is p<0.05.
drinking, examinees were asked their usual daily intake in
Table 1 shows the cumulative rates of and multivariate
cups. Information on the brewing method of coffee was
hazard ratios for the incidence of liver dysfunction over 4
not collected. Data on alcohol intake, cigarette smoking,
years according to coffee drinking. From the univariate
and coffee drinking were subdivided into terciles and
analysis using the Kaplan-Meier method, the estimated
distributions of these variables were as follows: alcohol intake
incidence of serum AST and/or ALT ?40 IU/l, ?50 IU/l,
(448 for none, 569 for ?45.9 g/day of ethanol, and 204 for
and ?60 IU/l decreased with an increase in coffee
?46.0 g/day of ethanol), cigarette smoking (610 for none,
consumption. The incidence curves of serum AST and/or
301 for ?29 cigarettes/day, and 310 for ?30 cigarettes/day),
ALT ?40 IU/l did not differ significantly among the three
and coffee drinking (160 for none, 677 for 1-2 cups/day,
subclasses of coffee drinking on the basis of the log-rank
and 384 for ?3 cups/day). Body mass index (BMI) was
test (p=0.147), but the incidence curves of serum AST and/
used as a measure of overall obesity. Weight and height
or ALT ?50 IU/l and ?60 IU/l attained statistical significance
were measured with the subjects wearing typical indoor
(p=0.020 and p=0.014, respectively). From the multivariate
clothing but with shoes off. BMI was calculated as weight/
analysis using the Cox proportional hazards model, coffee
height2 (kg/m2), and its mean value was 23.2 kg/m2 (SD 2.6
drinking had an independently inverse parameter-response
Industrial Health 2000, 38, 99–102
COFFEE CONSUMPTION AND LIVER DYSFUNCTION
Table 1. Cumulative rates of and multivariate hazard ratios for the incidence of liver dysfunction over 4 years according to coffee drinking
assessed by Kaplan-Meier method and Cox Proportional Hazards model in 1221 liver dysfunction-free Japanese male office workers aged 35
to 56 years*
Serum AST and/or ALT
*Subjects with normal level of serum AST and ALT ?39 IU/l and without medical care for or past history of liver disease. Hazard ratios (HRs) and
95% confidence intervals (CIs) for drinking 1–2 cups/day and ?3 cups/day of coffee, compared with non-drinking of coffee, were calculated, controlling
for age, body mass index, alcohol intake, and cigarette smoking.
relationship with the development of serum AST and/or ALT
The negative association between coffee and the
?40 IU/l (p=0.019 by test for tendency), ?50 IU/l (p=0.002),
development of increased serum AST and/or ALT levels
and ?60 IU/l (p=0.007), controlling for age, body mass index,
might also be to some extent to caffeine, which is one of the
alcohol intake, and cigarette smoking. Adjusted hazard ratios
major ingredients contained in coffee with various biological
for both 1–2 and ?3 cups/day of coffee drinking relative to
actions and possibly plays a crucial role in the observed
non-drinking decreased as the cutoff point of serum AST
associations of coffee intake. However, green tea, another
and/or ALT increased from 40 to 60 IU/l.
popular source of caffeine in Japan, has not been found to
In this study, coffee drinking was independently inversely
be inversely related to serum AST and/or ALT levels4, 7).
associated with the incidence of increased levels of serum
Sugar and cream in coffee might have the potential effect
AST and/or ALT. The inverse relation between coffee
on serum AST and/or ALT, but they did not attain statistical
drinking and the incidence of increased levels of serum AST
significance for the development of increased serum AST
and/or ALT was progressively stronger as the cutoff point
and/or ALT levels. As for the relationship between coffee
of serum AST and/or ALT increased from 40 to 60 IU/l.
drinking and other lifestyle factors, coffee drinking was
These results suggest that coffee may inhibit the elevation
significantly positively related to cigarette smoking and snack
of serum AST and/or ALT levels and that its effect is more
intake between meals15). Influences of coffee drinking on
pronounced in the protection of the development of higher
serum AST and/or ALT might be indirectly mediated through
serum AST and/or ALT levels.
coping mechanisms of these lifestyle factors. Further studies
The mechanism of the possible decreasing effect of coffee
are warranted to elucidate whether coffee is responsible for
on serum AST and/or ALT are still uncertain. The diterpenes
decreased serum AST and/or ALT levels or the protection
cafestol and kahweol (non-triglyceride lipids present in
of raised serum AST and/or ALT in the liver.
coffee) are shown to be responsible for the hyper-
In connection with our findings, it is interesting to note
cholesterolemic effects of boiled coffee9, 12), and cafestol has
that a large-scale prospective study in the USA has observed
been shown to decrease serum liver enzymes9). We did not
a decreased risk of alcoholic cirrhosis among coffee drinkers1).
obtain information regarding brewing methods of coffee,
The potential beneficial effect of coffee on the liver
but instant coffee is most popular in Japan, followed by
dysfunction or disease among the healthy populations
brewed (mostly filtered) coffee, whereas use of unfiltered
deserves further investigation.
coffee is virtually absent13). The negative association between
filtered or instant coffee and serum AST and/or ALT levels
has been found in several different cross-sectional studies7, 8).
Cafestol is as minimally contained in instant coffee as in
1) Klatsky AL, Armstrong MA (1992) Alcohol, smoking,
filtered coffee14), but we consider that cafestol does not fully
coffee, and cirrhosis. Am J Epidemiol 136, 1248–57.
explain the inverse association between coffee and the
2) Arnesen E, Huseby NE, Brenn T, Try K (1986) The
development of increased serum AST and/or ALT levels.
Tromsø heart study: distribution of, and determinants
N NAKANISHI et al.
for gamma-glutamyltransferase in a free-living
Identify of the cholesterol-raising factor from boiled
population. Scand J Clin Lab Invest 46, 63–70.
coffee and its effects on liver function enzymes. J Lipid
3) Nilssen O, Førde OH (1994) Seven-year longitudinal
Res 35, 721–33.
population study of change in gamma-glutamyl-
10) Japan Society of Clinical Chemistry (1989)
transferase: the Tromsø study. Am J Epidemiol 139,
Recommendation for measuring enzyme activity in
human serum. Jpn J Clin Chem 18, 211–62 (in Japanese
4) Kono S, Shinchi K, Imanishi K, Todoroki I, Hatsuse K
with English abstract).
(1994) Coffee and serum gamma-glutamyltransferase:
11) Cheng WS, Murphy TL, Smith MT, Cooksley WG,
a study of self-defense officials in Japan. Am J
Halliday JW, Powell LW (1990) Dose-dependent
Epidemiol 139, 723–7.
pharmacokinetics of caffeine in humans: relevance as
5) Sharp DS, Benowitz NL (1995) Re: “Alcohol, smoking,
a test of quantitative liver function. Clin Pharmacol
coffee, and cirrhosis” and “Coffee and serum gamma-
Ther 47, 516–24.
glutamyltransferase: a study of self-defense officials in
12) Joeres R, Klinker H, Heusler H, Epping J, Zilly W,
Japan.” [letter; comment] Am J Epidemiol 141, 480–2.
Richter E (1988) Influence of smoking on cafeine
6) Poikolainen K, Vartiainen E (1997) Determinants of
elimination in healthy volunteers and in patients with
?-glutamyltransferase: positive interaction with alcohol
alcoholic liver cirrhosis. Hepatology 8, 575–9.
and body mass index, negative association with coffee.
13) All Japan Coffee Association (1995) A basic survey
Am J Epidemiol 146, 1019–24.
for monitoring trends in the demand for coffee. Tokyo:
7) Tanaka K, Tokunaga S, Kono S, Tokudome S, Akamatsu
All Japan Coffee Association (in Japanese with English
T, Moriyama T, Zakouji H (1998) Coffee consumption
and decreased gamma-glutamyltransferase and
14) Urgert R, van der Weg G, Kosmeijer-Schuil TG, van
aminotransferase activities among male alcohol
de Bovenkamp P, Hovenier R, Katan MB (1995) Levels
drinkers. Int J Epidemiol 27, 438–43.
of the cholesterol-elevating diterpenes cafestol and
8) Casiglia E, Spolaore P, Ginocchio G, Ambrosio GB
kahweol in various coffee brews. J Agric Food Chem
(1993) Unexpected effects of coffee consumption on
liver enzymes. Eur J Epidemiol 9, 293–7.
15) Nakanishi N, Nakamura K, Suzuki K, Tatara K (in Press)
9) Weusten-Van der Wouw MP, Katan MB, Viani R,
Lifestyle and serum gamma-glutamyltransferase: a
Huggett AC, Liardon R, Lund-Larsen PG, Thelle DS,
study of middle-aged Japanese men. Occup Med (Lond).
Ahola I, Aro A, Meyboom S, Beynen AC (1994)
Industrial Health 2000, 38, 99–102