Alcohol Consumption and Risk of Stroke

Document Sample
Alcohol Consumption and Risk of Stroke Powered By Docstoc

Alcohol Consumption and Risk of Stroke
A Meta-analysis
Kristi Reynolds, MPH                              Context Observational studies suggest that heavy alcohol consumption may in-
L. Brian Lewis, MPH                               crease the risk of stroke while moderate consumption may decrease the risk.
John David L. Nolen, MD, PhD, MSPH                Objective To examine the association between alcohol consumption and relative
                                                  risk of stroke.
Gregory L. Kinney, MPH
                                                  Data Sources Studies published in English-language journals were retrieved by search-
Bhavani Sathya, MPH                               ing MEDLINE (1966–April 2002) using Medical Subject Headings alcohol drinking,
Jiang He, MD, PhD                                 ethanol, cerebrovascular accident, cerebrovascular disorders, and intracranial embo-
                                                  lism and thrombosis and the key word stroke; Dissertation Abstracts Online using the

         TROKE IS THE THIRD LEADING               keywords stroke and alcohol; and bibliographies of retrieved articles.
          cause of death and a major cause        Study Selection From 122 relevant retrieved reports, 35 observational studies (co-
          of disability in the United             hort or case control) in which total stroke, ischemic stroke, or hemorrhagic (intracerebral
          States.1,2 In 1999, 167366 deaths       or total) stroke was an end point; the relative risk or relative odds and their variance (or
in the United States resulted from stroke.1       data to calculate them) of stroke associated with alcohol consumption were reported;
Approximately 30% of stroke survivors             alcohol consumption was quantified; and abstainers served as the reference group.
are permanently disabled and 20% re-              Data Extraction Information on study design, participant characteristics, level of alco-
quire institutionalized care.1 Stroke is also     hol consumption, stroke outcome, control for potential confounding factors, and risk esti-
a huge financial burden for patients, their       mates was abstracted independently by 3 investigators using a standardized protocol.
families, and the health care system. The         Data Synthesis A random-effects model and meta-regression analysis were used
cost of stroke in the United States in 2002       to pool data from individual studies. Compared with abstainers, consumption of more
is estimated to be $49.4 billion, which           than 60 g of alcohol per day was associated with an increased relative risk of total
includes direct health expenditures and           stroke, 1.64 (95% confidence interval [CI], 1.39-1.93); ischemic stroke, 1.69 (95%
                                                  CI, 1.34-2.15); and hemorrhagic stroke, 2.18 (95% CI, 1.48-3.20), while consump-
lost productivity resulting from morbid-          tion of less than 12 g/d was associated with a reduced relative risk of total stroke,
ity and mortality.1                               0.83 (95%, CI, 0.75-0.91) and ischemic stroke, 0.80 (95% CI, 0.67-0.96), and con-
   Alcoholic beverages are consumed               sumption of 12 to 24 g/d was associated with a reduced relative risk of ischemic stroke,
widely throughout the world, and an               0.72 (95%, CI, 0.57-0.91). The meta-regression analysis revealed a significant non-
association between alcohol consump-              linear relationship between alcohol consumption and total and ischemic stroke and a
tion and stroke could have considerable           linear relationship between alcohol consumption and hemorrhagic stroke.
public health and clinical implications.          Conclusions These results indicate that heavy alcohol consumption increases the
Over the past 2 decades, many observa-            relative risk of stroke while light or moderate alcohol consumption may be protective
tional epidemiologic studies3-37 have             against total and ischemic stroke.
examined the role of alcohol as both a            JAMA. 2003;289:579-588                                                     
risk factor and a potential protective fac-
tor for stroke. Heavy alcohol consump-            hol consumption is positively related to       Headings alcohol drinking, ethanol, cere-
tion has been linked to an increased risk         risk of stroke.3,25                            brovascular accident, cerebrovascular dis-
of total stroke,23,32 ischemic stroke,29,33 and      We performed a meta-analysis of epi-        orders, and intracranial embolism and
hemorrhagic stroke.3,7,33,35 However, stud-       demiologic studies to examine the rela-        thrombosis and the keyword stroke was
ies investigating the association between         tive risk of stroke at various levels of       performed. The search was restricted to
moderate alcohol consumption and                  alcohol consumption.
stroke have reported conflicting results.                                                        Author Affiliations: Department of Epidemiology, Tu-
Some studies have reported that mod-              METHODS                                        lane University School of Public Health and Tropical
                                                  Study Selection                                Medicine, New Orleans, La.
erate alcohol consumption is inversely                                                           Corresponding Author and Reprints: Kristi Rey-
related to risk of total stroke,31 ischemic       A literature search of the MEDLINE da-         nolds, MPH, Department of Epidemiology, Tulane Uni-
                                                                                                 versity Health Sciences Center, School of Public Health
stroke,27,31,37 and hemorrhagic stroke,27,31      tabase (from January 1966 through              and Tropical Medicine, 1430 Tulane Ave SL18, New
while others found that moderate alco-            April 2002) using the Medical Subject          Orleans, LA 70112 (e-mail:

©2003 American Medical Association. All rights reserved.                                 (Reprinted) JAMA, February 5, 2003—Vol 289, No. 5 579

studies published in English-language            stroke. Five studies that lacked suffi-       ment units of alcohol consumption (eg,
journals and conducted in human sub-             cient data for calculation of relative risk   grams, milliliters, ounces, or drinks
jects. We also conducted a search of ab-         estimates were excluded.56-60 The remain-     consumed every day, week, or month).
stracts listed in Dissertation Abstracts         ing 6 excluded reports did not use ab-        Therefore, we first converted these dif-
Online using the keywords stroke and             stainers as the reference group.61-66 We      ferent units of alcohol consumption to
alcohol, and we performed a manual               included 19 cohort studies and 16 case-       grams per day. Among the 35 studies
search of references cited in published          control studies in our final analysis.        included in our meta-analysis, 20 re-
original study and relevant review ar-                                                         ported alcohol consumption as grams.
ticles. 38-48 The contents of 122 ab-            Data Abstraction                              We used the following conversion fac-
stracts or full-text manuscripts identi-         All data were independently abstracted        tors for the 4 studies that reported al-
fied during the literature search were           in triplicate by means of a standardized      cohol data as milliliters or ounces: 1 mL,
reviewed independently by 2 investi-             data-collection form. Discrepancies were      0.785 g; 1 fl oz, 28.41 mL (United King-
gators in duplicate to determine                 resolved by discussion and referencing        dom); and 1 fl oz, 29.58 mL (United
whether they met the criteria for inclu-         the original publication. We did not con-     States). Two of the 11 studies that re-
sion. When there were discrepancies be-          tact authors to request additional infor-     ported alcohol data as drinks provided
tween investigators for inclusion or ex-         mation. Study characteristics recorded        conversion factors in their articles. The
clusion, other investigators conducted           were as follows: title, article’s first au-   other 9 used common conversion fac-
additional evaluation of the study and           thor’s name, year, and source of publi-       tors.67 In the latter, a drink was de-
discrepancies were resolved in confer-           cation, country of origin, study design       fined as 12 g in the United States, 10 g
ence. To be included in our meta-                (cohort study or case-control study),         in Australia and Europe, and 21.2 g in
analysis, a published study had to meet          characteristics of the study population       Japan, which is the standard drink vol-
the following criteria: (1) observa-             (sample size; sampling methods; and dis-      ume in Japan.67
tional cohort or case-control study in           tribution of age, sex, and race), mea-           Alcohol consumption was reported
which total stroke, ischemic stroke, or          sures of outcome and exposure, dura-          as categorical data with a range in all
hemorrhagic (intracerebral or total)             tion of follow-up (for prospective cohort     studies. We assigned the mean of the
stroke was an end point; (2) relative risk       studies), confounding factors con-            upper and lower bounds in each cat-
or relative odds and their variance (or          trolled for by matching or adjustment,        egory as the average alcohol consump-
data to calculate them) of stroke asso-          and the relative risk (or relative odds)      tion. An upper bound was not re-
ciated with alcohol consumption were             of stroke associated with alcohol con-        ported in many studies for the category
reported; (3) alcohol consumption was            sumption and the corresponding con-           of highest consumption, so we as-
quantified; and (4) abstainers were used         fidence interval (or SE). Relative risks      sumed it to be the same amplitude as
as the reference group.                          overall and in each subgroup, accord-         the preceding category for calculation
   Fifty-three studies were identified and       ing to sex, subtype of stroke, level of al-   of average alcohol consumption in this
abstracted. Four studies reported total          cohol consumption, and type of alco-          category. In our meta-analysis, alco-
hemorrhagic stroke as the outcome,               holic beverage, were abstracted.              hol consumption was categorized into
which includes intracerebral and sub-                                                          5 groups: none (reference), less than 12,
arachnoid hemorrhage.4,7,10,11 None of the       Statistical Analysis                          12 to 23, 24 to 60, and more than 60
studies reported information on subdu-           Relative risk was used as a measure of        g/d. We assigned the level of alcohol
ral hemorrhagic strokes. We have used            the relation between alcohol consump-         consumption from each study to these
the term hemorrhagic stroke through-             tion and risk of stroke. For case-            groups based on the calculated aver-
out the article. Two reports consisted of        control studies, relative odds were used      age consumption of alcohol. In some
the same case patients but different con-        as a surrogate measure of the corre-          studies, the average alcohol consump-
trols and were treated as 2 separate stud-       sponding relative risk. Because the ab-       tion from more than 1 category fell into
ies.23,24 From the 53 studies, 18 were fur-      solute risk of stroke is low, the rela-       the same group of alcohol consump-
ther excluded for various reasons. Two           tive odds approximate the relative risk.      tion in our meta-analysis. When this oc-
studies were excluded because com-               Relative risks from individual studies        curred, we pooled the relative risks
bined risk estimates were reported for           for each level of alcohol consumption         within each category for each study and
men and women but levels of alcohol              and the corresponding SEs were trans-         then we pooled across all studies.
consumption were not the same for men            formed to their natural logarithms to            Both fixed-effects and DerSimonian
as for women.49,50 We excluded 5 stud-           stabilize the variances and to normal-        and Laird random-effects models68 were
ies that examined only the effect of binge       ize the distributions. The SEs were           used to calculate the pooled relative risk
drinking or acute alcohol consumption            derived from the confidence intervals         across levels of alcohol consumption. Al-
(within 24 hours before stroke)51-55 be-         provided in each study.                       though both models yielded similar find-
cause our study assessed habitual alco-             The studies included in our meta-          ings, results from the random-effects
hol consumption and relative risk of             analysis often differed in the measure-       model are presented herein because
580   JAMA, February 5, 2003—Vol 289, No. 5 (Reprinted)                         ©2003 American Medical Association. All rights reserved.
                                                                                                   ALCOHOL CONSUMPTION AND RISK OF STROKE

Table 1. Characteristics of 19 Cohort Studies of Alcohol Consumption and Risk of Stroke
                        Study           Exposure          Duration of       Follow-up               Outcome                 No. of
     Source          Participants      Assessment         Follow-up, y       Process               Assessment           Stroke Cases        Controlled Variables
Donahue et al,3     7878 Men aged    In-person               12          Clinical              Hospital discharge           290        Age, BMI, cigarette smoking,
   1986                45-69 y in        interview                            examinations        diagnosis, clinical                     hypertension, serum
                       Hawaii                                                 at years 2 and      diagnosis, death                        cholesterol, uric acid,
                                                                              6 and               certificate, or                         glucose level, hematocrit
                                                                              continued           autopsy record
Kono et al,4        5135 Men in      Self-administered       19          Vital status          Death certificate            230        Age, cigarette smoking
   1986                Japan              questionnaire                       ascertained
                                                                              by medical
Gordon and          1910 Men aged    Self-administered       29          Vital statistics      Proxy reports or              33        None
   Doyle,5             38-55 y in         questionnaire                       records,             death certificate
   1987                New York                                               newspapers,
                                                                              or reports
                                                                              from proxies
Stampfer et al,6    87 526 US        Self-administered         4         Biennial              Medical records              120        Age, cigarette smoking,
   1988                 women             questionnaire                      question-                                                    hypertension, DM, serum
                        aged                                                 naires                                                       cholesterol level, obesity,
                        34-59 y                                                                                                           exercise, cholesterol intake,
                                                                                                                                          saturated and
                                                                                                                                          polyunsaturated fat intake,
                                                                                                                                          parental history of MI
                                                                                                                                          before age 60 y,
                                                                                                                                          menopausal status,
                                                                                                                                          hormone use, study period
Klatsky et al,7     107 137 US       Self-administered         6         Surveillance of       Clinical diagnosis           674        Age, sex, race, cigarette
    1989               men and            questionnaire                      hospital                                                     smoking, SBP, coffee
                       women                                                 discharges                                                   consumption, BMI,
                       aged 50 y                                                                                                          baseline disease
Shaper et al,8      7735 UK men      In-person                 8         Death register        Clinical diagnosis or        110        Age, cigarette smoking, SBP
   1991                aged              interview                                                  death certificate
                       40-59 y
Goldberg et al,9    6069 Men aged    In-person               15          Clinical              Hospital discharge            70        Age; cigarette smoking; SBP;
   1994                51-75 y in        interview                            examinations        diagnosis, clinical                     serum cholesterol, serum
                       Hawaii                                                 at years 2 and      diagnosis, or                           triglyceride, and serum uric
                                                                              6 and               death certificate                       acid levels, coffee
                                                                              continued                                                   consumption, total caloric
                                                                              surveillance                                                intake
Hansagi et al,10    15 077 Men and   Self-administered       20          Death register        Death certificate            769        Age, cigarette smoking
   1995                 women             questionnaire
                        aged 40 y
                        in Sweden
Iso et al,11 1995   2890 Men aged    In-person               10.5        Not specified         Clinical diagnosis and       178        Age, cigarette smoking,
                       40-69 y in        interview                                                  CT scan                               hypertension, serum total
                       Japan                                                                                                              cholesterol level, DM
Kiyohara et al,12   1621 Men and     In-person               26          Biennial              Neurological                 304        Age, sex, hypertension
    1995               women             interview                           examinations,        examination, CT
                       aged 40 y                                             mail, or             scan,
                       in Japan                                              telephone            angiography,
                                                                                                  lumbar puncture,
                                                                                                  or autopsy
Palmer et al,13     6369 Men and     In-person               22          Questionnaire         Death certificate            159        Age, sex, cigarette smoking,
    1995               women              interview                         every 1-2 y                                                   SBP
                       aged               (1971-1976)
                       18-90 y in    Self-administered
                       England            questionnaire
                                          (after 1976)
Yuan et al,14       18 244 Men       In-person                 9         Annual contact        Death certificate            269        Age, cigarette smoking,
   1997                 aged             interview                                                                                        educational level
                        45-64 y in
Maskarinec et       27 678 Men and   In-person               20          Passive follow-up     Death certificate            433        Age, BMI, cigarette smoking,
   al,15 1998           women            interview                                                                                        ethnicity, educational level
                        aged 30 y
                        in Hawaii
Hart et al,16       5766 Men aged    In-person               21          NHS death             Death certificate            133        Age, BMI, cigarette smoking,
    1999               35-64 y in        interview                          register                                                      DBP, serum cholesterol
                       Scotland                                                                                                           level, educational level,
                                                                                                                                          social class, father’s social
                                                                                                                                          class, car use, siblings,
                                                                                                                                          deprivation category,
                                                                                                                                          adjusted FEV, angina,
                                                                                                                                          ischemia on ECG,

©2003 American Medical Association. All rights reserved.                                                 (Reprinted) JAMA, February 5, 2003—Vol 289, No. 5 581

Table 1. Characteristics of 19 Cohorts Studies of Alcohol Consumption and Risk of Stroke (cont)
                          Study           Exposure          Duration of       Follow-up              Outcome                  No. of
   Source, y           Participants      Assessment         Follow-up, y       Process              Assessment            Stroke Cases       Controlled Variables
Leppälä et al,17      26 556 Men       Self-administered         6.1       National hospital    Clinical diagnosis or         960        Age, BMI, cigarette smoking,
   1999                   aged              questionnaire                      discharge             death certificate                      serum cholesterol level,
                          50-69 y in                                           register and                                                 DM, educational level,
                          Finland                                              national death                                               leisure time physical
                                                                               register                                                     activity, heart disease,
                                                                                                                                            supplementation with
                                                                                                                                              -tocopherol or beta
Romelsjö et al,18     49 618 Men       Self-administered        25         Inpatient care       Clinical diagnosis or         223        BMI, cigarette smoking, BP,
   1999                   aged              questionnaire                      register and          death certificate                      father’s social class,
                          17-45 y in                                           death register                                               running away from home,
                          Sweden                                                                                                            poor school well-being,
                                                                                                                                            parental divorce, poor
                                                                                                                                            emotional control, few
                                                                                                                                            friends, unemployment 3
                                                                                                                                            mo during life, poor health
Gaziano et al,19      89 299 US men    Self-administered         5.5       National Death       Death certificates            150        Age, BMI, cigarette smoking,
   2000                   aged              questionnaire                      Index search                                                 DM, exercise
                          40-84 y
Jousilahti et al,20   14 874 Men       Self-administered        12         National hospital    Clinical diagnosis or         470        Age, BMI, cigarette smoking,
   2000                   and women         questionnaire                      discharge             death certificates                     serum total cholesterol,
                          aged                                                 register or                                                  SBP, DBP, and study year
                          25-64 y in                                           central
                          Finland                                              statistical
                                                                               office of
Djousse et al,21
      ´               5209             In-person                30         Biennial             Clinical diagnosis and        441        Age, BMI, cigarette smoking,
    2002                 Framing-          interview                           examinations          radiographic                           DM
                         ham, Mass,                                                                  images
                         men and
Abbreviations: BMI, body mass index; CT, computed tomography; DBP, diastolic blood pressure; DM, diabetes mellitus; ECG, electrocardiogram; FEV, forced expiratory
  volume; MI, myocardial infarction; NHS, National Health Service; SBP, systolic blood pressure.

significant heterogeneity was identified                    age due to the lack of such detailed in-                 ported total stroke as the outcome. In
among studies.68 A weighted meta-                           formation in most studies.                               addition, 7 studies reported ischemic
regression analysis with no intercept term                     To assess the potential for publica-                  stroke, and 7 studies reported hemor-
was performed to examine the associa-                       tion bias, we constructed a funnel plot                  rhagic stroke as the outcome. The fol-
tion between alcohol consumption and                        in which the log relative risks were plot-               low-up period ranged from 4 to 30
the natural logarithm of the relative risk                  ted against their SEs.70 In addition, a                  years. The study population in 7 co-
of stroke. We used the “pool-first”                         rank correlation for the association be-                 hort studies consisted of men and
method proposed by Greenland and                            tween standardized log relative risks                    women, 1 consisted entirely of women,
Longnecker.69 This method was chosen                        and their SEs was conducted using the                    and 11 consisted of only men.
because several studies reported find-                      Kendall correlation coefficient. The                        Twelve of the 16 case-control stud-
ing a nonlinear, J- or U-shaped relation-                   correlation between sample size and                      ies were conducted outside the United
ship between alcohol consumption and                        relative risk would be high if small stud-               States (TABLE 2). The number of case
relative risk of stroke. This method is ad-                 ies with null results were less likely to                subjects enrolled in these studies
vantageous because it can easily be ex-                     be published. A significant correlation                  ranged from 89 in the study by Hen-
tended to test nonlinearity and identify                    between sample size and relative risk                    rich and Horwitz26 to 677 in the study
J- or U-shaped curves, or other relation-                   would not exist in the absence of this                   by Sacco et al,34 and the corresponding
ships between exposure levels and rela-                     type of publication bias.70                              number of control subjects ranged
tive risks. For each included study, we                                                                              from 153 in the study by Palomäki et
performed an initial fit of a quadratic                     RESULTS                                                  al29 to 1139 in the study by Sacco et
curve. When a nonsignificant term was                       The characteristics of the study sub-                    al.34 Total stroke was the study out-
found in the initial model, a subse-                        jects and design of the cohort studies                   come in 9 studies, whereas 8 studies
quent fit of a simpler model (linear or                     are presented in TABLE 1. Of the 19 co-                  collected data on ischemic stroke and
solitary square term) was conducted.                        hort studies, 8 were conducted in the                    5 collected data on hemorrhagic
   Prestated subgroup analyses were                         United States. The number of subjects                    stroke. Fourteen of the 16 case-control
conducted by subtype of stroke and sex                      in the cohort studies ranged from 1621                   studies were composed of both men
for the different levels of alcohol con-                    in the study by Kiyohara et al 12 to                     and women, 1 case-control study con-
sumption. Subgroup analyses were not                        107 137 in the study by Klatsky et al.7                  sisted of only women, and 1 case-
performed by type of alcoholic bever-                       Among the 19 cohort studies, 15 re-                      control study consisted of only men.
582     JAMA, February 5, 2003—Vol 289, No. 5 (Reprinted)                                       ©2003 American Medical Association. All rights reserved.
                                                                                                ALCOHOL CONSUMPTION AND RISK OF STROKE

Table 2. Characteristics of 16 Case-Control Studies of Alcohol Consumption and Risk of Stroke
  Source, y              Stroke Cases                  Controls             Case Assessment                 Assessment             Controlled Variables
Herman et al,22      132 Male and female        239 Patients from the      Clinical examination         In-person interview    Age, sex
   1983                 patients with              same hospital
                        incident stroke
                        event in 2 hospitals
                        in the Netherlands
Gill et al,23 1986   230 Male and female        230 Hospital patients      Clinical examination,        In-person interview    Age, sex, race, cigarette
                        patients with                                          CT scan,                                           smoking, treatment of
                        stroke diagnosis in                                    angiography, and                                   hypertension, medication
                        the district hospital                                  postmortem
                        in England                                             examinations, or
                                                                               lumbar puncture
Gill et at,24 1988   230 Male and female        577 Male and female        Clinical examination,        In-person interview    Age, race, cigarette smoking,
                        patients with              industrial workers in       CT scan,                                           treatment of hypertension,
                        stroke diagnosis in        the same                    angiography, and                                   social class, drug therapy
                        the district hospital      community                   postmortem
                        in England                                             examinations, or
                                                                               lumbar puncture
Gorelick et al,25    205 Male and female        410 Outpatient clinic      Clinical diagnosis and       In-person interview    Age, sex, race, cigarette
   1989                 patients with              patients                    CT scan                                            smoking, hypertension,
                        incident ischemic                                                                                         method of hospital
                        stroke in 3 medical                                                                                       payment
                        centers in Chicago
Henrich and          89 Male and female         178 Patients               Clinical examination         Telephone interview    None
   Horwitz,26           hospitalized               discharged from the         and CT scan
   1989                 patients with              same hospital
                        ischemic stroke in
Gill et al,27 1991   621 Male and female        573 Male and female        Clinical examination,        In-person interview    Age, sex, race, cigarette
                        hospitalized               industrial workers in       CT scan,                                           smoking, hypertension,
                        patients with              the same                    angiography and                                    social class, medication
                        stroke diagnosis in        community                   postmortem
                        2 centers in                                           examination, or
                        England                                                lumbar puncture
Ben-Shlomo           115 Male and female        165 Generally matched,     Clinical examination,        Cases, in-person       General and selective controls:
   et al,28 1992        hospitalized               115 selectively             CT scan, or                 taped interview        age, sex, cigarette
                        patients with              matched, and 752            lumbar puncture          Controls,                 smoking, hypertension,
                        incident stroke in 3       community controls                                      self-administered      DM, heart disease
                        hospitals in the                                                                   questionnaire       Community controls: age,
                        United Kingdom                                                                                            sex, cigarette smoking,
                                                                                                                                  hypertension, and social
Palomäki et al,29    156 Male hospitalized      153 Hospital patients      Clinical diagnosis           In-person interview    Age, BMI, cigarette smoking,
    1993                patients with                                                                                             hypertension, DM,
                        ischemic stroke in                                                                                        coronary heart disease,
                        Finland                                                                                                   history of snoring
Shinton et al,30     125 Male and female        198 Community              Clinical examination,        Alcohol diary          Age, sex, history of
    1993                patients with              controls                    CT scan, or                                        cardiovascular disease
                        incident stroke in                                     autopsy
                        11 general practice
                        partnerships in
Jamrozik et al,31    501 Male and female        931 Community              Clinical examination,        In-person interview    Age, sex, cigarette smoking,
   1994                 patients with              controls from the           CT scan, MRI, or                                   hypertension, DM,
                        stroke diagnosis in        electoral roles             autopsy                                            previous stroke or TIA,
                        Australia                                                                                                 previous MI, adding salt to
                                                                                                                                  food, consumption of fish
                                                                                                                                    2 times/mo,
                                                                                                                                  claudication, use of
                                                                                                                                  reduced fat or skim milk,
                                                                                                                                  consumption of meat 4
Beghi et al,32       200 Male and female        170 Patients in the        Clinical examination,        In-person interview    Age, sex
   1995                 hospitalized               same hospital and           CT scan, or
                        patients with              202 community               neurological
                        stroke in Italy            controls                    consultation
Caicoya et al,33     467 Male and female        477 Residents of the       Clinical examination or      In-person interview    Age, sex, cigarette smoking,
   1999                 patients with              same community              CT scan                                            hypertension, DM,
                        incident stroke in                                                                                        hypercholesterolemia,
                        Spain                                                                                                     cardiac disease

©2003 American Medical Association. All rights reserved.                                             (Reprinted) JAMA, February 5, 2003—Vol 289, No. 5 583

Table 2. Characteristics of 16 Case-Control Studies of Alcohol Consumption and Risk of Stroke (cont)
    Source                 Stroke Cases                       Controls               Case Assessment               Assessment                 Controlled Variables
Sacco et al,34       677 Men and women with          1139 Community                Brain imaging and           In-person interview       Age, sex, race, BMI, cigarette
   1999                 incident cerebral               controls                       clinical diagnosis                                   smoking, hypertension,
                        infarction in the                                                                                                   DM, cardiac disease,
                        community in                                                                                                        educational level
                        New York
Thrift et al,35      331 Male and female             331 Residents from the        CT scan, MRI, or            In-person interview       Age, sex, BMI, cigarette
    1999                patients with primary           same neighborhood             autopsy                                               smoking, DM, serum
                        hemorrhagic stroke                                                                                                  cholesterol level, SES,
                        from 13 hospitals in                                                                                                educational level, exercise,
                        Melbourne, Australia                                                                                                cardiovascular disease,
                                                                                                                                            hormone replacement
Zodpey et al,36      166 Male and female             166 Patients from the         CT scan                     In-person interview       Age, sex
   2000                 hospitalized patients           same hospital
                        with incident
                        hemorrhagic stroke
                        in India
Malarcher et al,37   224 Female patients with        392 Female community          Hospital discharge          In-person interview       Age, race, BMI, cigarette
   2001                 incident cerebral               residents                     diagnosis, clinical                                   smoking, hypertension,
                        infarction in 59                                              diagnosis,                                            DM, total cholesterol, HDL
                        hospitals in                                                  neuroimaging                                          cholesterol level,
                        Baltimore-Washington                                          results, or autopsy                                   geographic region of
                        region in the United                                          reports                                               residence, educational
                        States                                                                                                              level, coronary heart
Abbreviations: BMI, Body mass index; CT, computed tomography; DM, diabetes mellitus; MI, myocardial infarction; MRI, magnetic resonance imaging; SES, socioeconomic status;
  TIA, transient ischemic attack.

   The results from the random-                           60 g/d, or more then 5 drinks per day,                    important confounders, or studies that
effects model and the meta-regression                     had the highest relative risk.                            did not exclude prevalent stroke cases
analysis test for trend are presented in                     The association between alcohol con-                   at baseline.
TABLE 3. The overall results indicate a                   sumption and relative risk of total                          There was no evidence of publica-
nonlinear association between alco-                       stroke was similar in men and women                       tion bias in our study as indicated by a
hol consumption and relative risk of                      (Table 3 and FIGURE 2) although the                       funnel plot (FIGURE 3) and the Ken-
total stroke (P = .002 for nonlinear                      relative risk was somewhat lower in                       dall correlation coefficient. The Ken-
trend). Compared with the reference                       women consuming less than 12 g/d, or                      dall correlation coefficient for the SE
group of abstainers, alcohol consump-                     less than 1 drink per day, than in men.                   and the standardized log relative risk
tion of less than 12 g/d, or less than 1                  Likewise, the association was similar in                  was −0.072 (P = .17) for all studies.
drink per day based on US conver-                         case-control studies and cohort stud-                     When the outliers were excluded, the
sions, was significantly associated with                  ies, with alcohol consumption of less                     Kendal correlation coefficient for the
a decreased relative risk of total stroke,                than 12 g/d, or less than 1 drink per day,                SE and the standardized log relative risk
while alcohol consumption of more                         among cohort studies and alcohol con-                     became −0.053 (P=.32).
than 60 g/d, or more than 5 drinks per                    sumption of less than 24 g/d, or less
day, was significantly associated with                    than 2 drinks per day, among case-                        COMMENT
an increased relative risk of total stroke.               control studies associated with a sig-                    Several large epidemiologic studies that
The association between alcohol con-                      nificant reduced relative risk while al-                  have examined the effect of alcohol con-
sumption and relative risk of ischemic                    cohol consumption of more than 60 g/d,                    sumption on the risk of stroke have pro-
stroke was J-shaped with the lowest risk                  or more than 5 drinks per day, was as-                    vided inconsistent findings. In our cur-
among those consuming less than 12                        sociated with an increased relative risk.                 rent meta-analysis, we found a J-shaped
g/d, or less than 1 drink per day, or 12                     The findings from the sensitivity                      association between alcohol consump-
to 24 g/d, or 1 to 2 drinks per day, and                  analyses that excluded studies based on                   tion and the relative risk of total and
the highest risk among those consum-                      different inclusion criteria are pre-                     ischemic stroke and a linear associa-
ing more than 60 g/d, or more than 5                      sented in T ABLE 4. Risk estimates                        tion between alcohol consumption and
drinks per day, (FIGURE 1). Relative risk                 changed very little after the exclusion                   the relative risk of hemorrhagic stroke.
of hemorrhagic stroke increased lin-                      of outliers, studies without computed                     Moderate alcohol consumption was as-
early with increasing alcohol consump-                    tomographic scans or other imaging                        sociated with a reduced relative risk of
tion, and those consuming more than                       measures, studies that did not adjust for                 total and ischemic stroke while heavy
584     JAMA, February 5, 2003—Vol 289, No. 5 (Reprinted)                                        ©2003 American Medical Association. All rights reserved.
                                                                                                                                      ALCOHOL CONSUMPTION AND RISK OF STROKE

alcohol consumption was associated                                                         The relationship between alcohol con-                      alcohol-induced hypertension, cardio-
with an increased relative risk of total,                                                sumption and stroke is believed to in-                       myopathy, coagulation disorders, atrial
ischemic, and hemorrhagic stroke.                                                        volve various mechanisms including                           fibrillation, and reductions in cerebral

Table 3. Overall Relative Risk (95% Confidence Interval) of Stroke Associated With Alcohol Consumption and Test for Trend
                                                                                                                                                                                        P Value
                                                                                                        Alcohol Intake, g/d
                                                  No. of                                                                                                          Test for Linear            Test for Nonlinear
                                                 Studies                     12                  12-24                  24-60                    60                Association*                 Association
Overall                                            35            0.83 (0.75-0.91)            0.91 (0.78-1.06)      1.10 (0.97-1.24)    1.64 (1.39-1.93)                                              .002
Type of stroke
   Ischemic                                           15         0.80 (0.67-0.96)            0.72 (0.57-0.91)      0.96 (0.79-1.18)    1.69 (1.34-2.15)                                             .004
   Hemorrhagic                                        12         0.79 (0.60-1.05)            0.98 (0.77-1.25)      1.19 (0.80-1.79)    2.18 (1.48-3.20)                 .004                        .17
   Men                                                27         0.89 (0.79-1.01)            0.94 (0.84-1.05)      1.08 (0.96-1.21)    1.76 (1.57-1.98)                                             .001
   Women                                              16         0.66 (0.61-0.71)            0.79 (0.56-1.11)      0.80 (0.49-1.30)    4.29 (1.30-14.14)                                            .001
Study design
   Cohort                                             19         0.82 (0.73-0.92)            0.94 (0.84-1.05)      1.06 (0.90-1.23)    1.63 (1.49-1.79)                                             .02
   Case control                                       16         0.80 (0.67-0.97)            0.65 (0.44-0.96)      1.12 (0.92-1.37)    1.98 (1.35-2.92)                                             .03
*Tests for linear associations were performed only when nonlinear associations were not statistically significant.

Figure 1. Scatterplot of Log Relative Risk and Meta-Regression Curve of Stroke Associated With Alcohol Consumption by Subtypes of Stroke

                                                                 Ischemic Stroke                                                                             Hemorrhagic Stroke

Natural Logarithm Relative Risk





                                       0   20    40        60         80         100   120      140       160     180           0      20       40      60        80        100        120    140    160    180
                                                                Alcohol Intake, g/d                                                                           Alcohol Intake, g/d

Most studies provided more than 1 relative risk estimate for multiple levels of alcohol consumption.

Figure 2. Scatterplot of Log Relative Risk and Meta-Regression Curve of Stroke Associated With Alcohol Consumption by Sex

                                                                           Men                                                                                    Women

Natural Logarithm Relative Risk





                                       0    20        40         60          80        100        120       140                 0      10        20          30        40         50         60      70     80
                                                                Alcohol Intake, g/d                                                                           Alcohol Intake, g/d

Most studies provided more than 1 relative risk estimate for multiple levels of alcohol consumption.

©2003 American Medical Association. All rights reserved.                                                                                    (Reprinted) JAMA, February 5, 2003—Vol 289, No. 5 585

Table 4. Overall Relative Risk (95% Confidence Interval) of Stroke Associated With Alcohol Consumption According to Different Exclusion
                                                                                                                             Alcohol Intake, g/d

                     Studies Included in Analysis                                        12                          12-24                     24-60                   60
All studies                                                                      0.83 (0.75-0.91)               0.91 (0.78-1.06)          1.10 (0.97-1.24)     1.64 (1.39-1.93)
All studies except outliers*                                                     0.83 (0.75-0.91)               0.91 (0.78-1.06)          1.11 (0.98-1.26)     1.62 (1.46-1.81)
Studies that used computed tomography scans or                                   0.84 (0.75-0.94)               0.86 (0.71-1.05)          1.14 (1.01-1.35)     1.74 (1.37-2.21)
     other imaging measures as an outcome measure†
Studies that controlled for important stroke risk factors‡                       0.81 (0.71-0.92)               0.80 (0.64-1.00)          1.12 (0.94-1.33)     1.62 (1.19-2.21)
Cohort studies that used incident stroke events§                                 0.83 (0.73-0.95)               0.91 (0.77-1.07)          1.02 (0.83-1.26)     1.58 (1.43-1.73)
*The 24 to 60 and     60 g/d levels in the study by Caicoya et al33 were excluded because they were outliers.
†Studies that solely used death certificates or death registries for the outcome assessment were excluded.4,5,10,13-16,19
‡Studies that did not control for age, cigarette smoking or hypertension were excluded.4,5,10,12,14,15,17-19,21,22,26,30,32,35,36
§Cohort studies that did not exclude prevalent stroke events were excluded.4,5,8,10,13-16

                                                                  cluded in our analysis. Our study may                             among studies. For instance, some stud-
Figure 3. Funnel Plot of Log Relative Risk vs
Variance of Log Relative Risks Among All                          inherit the problems of potential bias                            ies used the lowest consumption level
Studies                                                           and confounding effects associated with                           as the reference group while others used
                                                                  observational studies. However, a ran-                            abstainers. In an effort to avoid com-
                                                                  domized controlled trial of alcohol con-                          bining studies that were not compa-
             80                                                   sumption and stroke has not been per-                             rable, we chose to include only those
                                                                  formed and is unlikely to be conducted                            studies that used abstainers as the ref-

                                                                  in the future. Consequently, we must                              erence group. It has been suggested that
                                                                  rely on data from observational stud-                             the U- or J-shaped association be-
             20                                                   ies to draw conclusions and make rec-                             tween alcohol consumption and mor-
              0                                                   ommendations.                                                     tality from cardiovascular disease may
              –2.5   –1.5     –0.5     0.5      1.5       2.5        Second, computed tomographic                                   be due to the inclusion of ex-drinkers
                     Natural Logarithm Relative Risk              scans and other imaging techniques                                in the reference group of abstainers. Ex-
                                                                  were not available for some early stud-                           drinkers may have stopped alcohol con-
Most studies provided more than 1 relative risk esti-
mate for multiple levels of alcohol consumption.                  ies. Furthermore, several studies only                            sumption due to health problems and
                                                                  used death certificates or death regis-                           they are at increased risk for death from
                                                                  ter data for diagnosis of stroke out-                             cardiovascular disease. 47,75,76 How-
blood flow.37,46,71,72 A plausible explana-                       come. However, our findings were un-                              ever, several studies have examined this
tion of a reduced risk of ischemic stroke                         likely due to misclassification of                                potential bias and concluded that the
with moderate alcohol consumption is                              outcome because the relative risks of                             J- or U-shaped relationship between al-
that alcohol increases high-density li-                           stroke associated with alcohol con-                               cohol consumption and risk of cardio-
poprotein cholesterol levels and de-                              sumption did not change after exclu-                              vascular disease mortality held
creases platelet aggregation and fibri-                           sion of studies that did not use com-                             true.6,13,27,77 Moreover, we conducted a
nolytic activity.6,71,72 Epidemiologic                            puted tomography or other imaging                                 sensitivity analysis in which only pro-
studies also have consistently observed                           techniques for diagnosis. Our find-                               spective cohort studies that excluded
a protective effect of moderate alcohol                           ings were also unlikely due to con-                               prevalent stroke cases at baseline were
consumption on coronary heart dis-                                founding effects because the relative                             included, and we found that the shape
ease.73,74 Alcohol-induced hyperten-                              risks of stroke associated with alcohol                           of association remained unchanged.
sion and coagulation disorders are prob-                          consumption were similar among all                                Second, the health effects of binge
able underlying mechanisms for                                    studies and only those studies that con-                          drinking may be different than those for
hemorrhagic stroke.27,45,71 The antico-                           trolled for important risk factors for                            regular drinkers. The failure to differ-
agulant effects of alcohol, although they                         stroke, such as cigarette smoking and                             entiate between these 2 groups could
appear to be beneficial for decreasing the                        hypertension. Additionally, our re-                               possibly obscure the observation of any
risk of ischemic stroke, may play an im-                          sults were unlikely to result from pub-                           true association. Therefore, we only in-
portant role in increasing the risk of                            lication bias as demonstrated by the fun-                         cluded studies that examined the effect
hemorrhagic stroke.71,74                                          nel plot and rank correlation analysis.                           of usual alcohol consumption rather
   There are several potential limita-                               Several methodological issues re-                              than acute alcohol consumption. Third,
tions in our study. First, our study is a                         garding epidemiologic research on the                             the measurement units, especially the
meta-analysis of observational stud-                              health impact of alcohol consumption                              definition of an alcohol drink, varies
ies. The quality of our study depends                             are worth considering. First, the selec-                          among studies. We attempted to over-
on data from original publications in-                            tion of the reference group may vary                              come this problem by applying a com-

586           JAMA, February 5, 2003—Vol 289, No. 5 (Reprinted)                                                ©2003 American Medical Association. All rights reserved.
                                                                                         ALCOHOL CONSUMPTION AND RISK OF STROKE

monly used and validated method sug-         moderate alcohol consumption re-                           tension Care Computing Project. J Hypertens. 1995;
gested by Turner.67 Finally, assessment      duces risk of ischemic stroke. How-                        14. Yuan JM, Ross RK, Gao YT, Henderson BE, Yu MC.
methods for alcohol consumption may          ever, the implications of these find-                      Follow up study of moderate alcohol intake and mor-
                                                                                                        tality among middle aged men in Shanghai, China
also vary among studies. The assess-         ings should be examined cautiously.                        [comments]. BMJ. 1997;314:18-23.
ment of alcohol consumption is usu-          Any advice regarding the consump-                          15. Maskarinec G, Meng L, Kolonel L. Alcohol in-
ally based on self-reported alcohol hab-     tion of alcohol should be tailored to the                  take, body weight, and mortality in a multiethnic pro-
                                                                                                        spective cohort. Epidemiology. 1998;9:654-661.
its. Such data are subject to errors of      individual patient’s risks and poten-                      16. Hart CL, Smith GD, Hole DJ, Hawthorne VM. Al-
recall. For example, heavy drinkers may      tial benefits.                                             cohol consumption and mortality from all causes, coro-
                                                                                                        nary heart disease, and stroke: results from a prospec-
be more likely to underreport their al-                                                                 tive cohort study of Scottish men with 21 years of
                                             Author Contributions: Study concept and design:
cohol consumption. The majority of           Reynolds, Lewis, Nolen, Kinney, Sathya, He.                follow up [comments]. BMJ. 1999;318:1725-1729.
                                                                                                                   ¨ ¨
                                                                                                        17. Leppala JM, Paunio M, Virtamo J, et al. Alcohol
studies in this meta-analysis used in-       Acquisition of data: Reynolds, Lewis, Nolen, Kinney,
                                                                                                        consumption and stroke incidence in male smokers.
person interviews, while 11 studies used     Analysis and interpretation of data: Reynolds, Lewis,      Circulation. 1999;100:1209-1214.
self-administered questionnaires, 1          Nolen, Kinney, Sathya, He.                                                  ¨
                                                                                                        18. Romelsjo A, Leifman A. Association between al-
                                             Drafting of the manuscript: Reynolds, Nolen, Kinney,       cohol consumption and mortality, myocardial infarc-
study conducted telephone inter-             Sathya.                                                    tion, and stroke in 25 year follow up of 49618 young
views, and 1 study used alcohol con-         Critical revision of the manuscript for important in-      Swedish men. BMJ. 1999;319:821-822.
                                             tellectual content: Reynolds, Lewis, Nolen, Kinney,        19. Gaziano JM, Gaziano TA, Glynn RJ, et al. Light-
sumption diaries.                                                                                       to-moderate alcohol consumption and mortality in the
                                             Sathya, He.
   There are several advantages of our       Statistical expertise: Reynolds, Nolen, Kinney, He.        Physicians’ Health Study enrollment cohort. J Am Coll
study. The discrepancies among stud-         Obtained funding: He.                                      Cardiol. 2000;35:96-105.
                                             Administrative, technical, or material support: Lewis,     20. Jousilahti P, Rastenyte D, Tuomilehto J. Serum
ies regarding the association between        Kinney, Sathya.                                            gamma-glutamyl transferase, self-reported alcohol
alcohol consumption and relative risk        Study supervision: He.                                     drinking, and the risk of stroke. Stroke. 2000;31:1851-
                                             Funding/Support: This study was supported in part          1855.
of stroke also may be attributable to a      by grant R01HL60300 from the National Heart, Lung,                        ´
                                                                                                        21. Djousse L, Ellison R, Beiser A, Scaramucci A,
small sample size in the individual stud-    and Blood Institute.                                       D’Agostino R, Wolf P. Alcohol consumption and risk
                                                                                                        of ischemic stroke: the Framingham study. Stroke.
ies, resulting in insufficient statistical                                                              2002;33:907-912.
power. This meta-analysis included a         REFERENCES                                                 22. Herman B, Schmitz PI, Leyten AC, et al. Multi-
large number of people from different                                                                   variate logistic analysis of risk factors for stroke in Til-
                                             1. American Heart Association. 2002 Heart and Stroke
                                             Statistical Update. Dallas, Tex: American Heart As-        burg, the Netherlands. Am J Epidemiol. 1983;118:
populations throughout the world. Ad-                                                                   514-525.
                                             sociation; 2001.
ditionally, we were able to assess the       2. Warlow CP. Epidemiology of stroke. Lancet. 1998;        23. Gill JS, Zezulka AV, Shipley MJ, Gill SK, Beevers
                                                                                                        DG. Stroke and alcohol consumption. N Engl J Med.
pattern of the association between level     352(suppl 3):1-4.
                                             3. Donahue RP, Abbott RD, Reed DM, Yano K. Al-
of alcohol consumption and relative risk     cohol and hemorrhagic stroke: the Honolulu Heart Pro-      24. Gill JS, Shipley MJ, Hornby RH, Gill SK, Beevers
of stroke with precision due to the large    gram. JAMA. 1986;255:2311-2314.                            DG. A community case-control study of alcohol con-
                                             4. Kono S, Ikeda M, Tokudome S, Nishizumi M, Ku-           sumption in stroke. Int J Epidemiol. 1988;17:542-
sample size. Finally, the association be-    ratsune M. Alcohol and mortality: a cohort study of        547.
tween alcohol consumption and rela-          male Japanese physicians. Int J Epidemiol. 1986;15:        25. Gorelick PB, Rodin MB, Langenberg P, Hier DB,
                                             527-532.                                                   Costigan J. Weekly alcohol consumption, cigarette
tive risk of stroke was consistent among     5. Gordon T, Doyle JT. Drinking and mortality: the         smoking, and the risk of ischemic stroke: results of a
subgroups by study design, sex, and          Albany Study. Am J Epidemiol. 1987;125:263-270.            case-control study at three urban medical centers in
                                             6. Stampfer MJ, Colditz GA, Willett WC, Speizer FE,        Chicago, Illinois. Neurology. 1989;39:339-343.
stroke subtype.                                                                                         26. Henrich JB, Horwitz RI. Evidence against the as-
                                             Hennekens CH. A prospective study of moderate al-
   Our findings have important clini-        cohol consumption and the risk of coronary disease         sociation between alcohol use and ischemic stroke risk.
cal and public health implications. In       and stroke in women. N Engl J Med. 1988;319:267-           Arch Intern Med. 1989;149:1413-1416.
                                             273.                                                       27. Gill JS, Shipley MJ, Tsementzis SA, et al. Alcohol
the United States, 44% of adults, aged       7. Klatsky AL, Armstrong MA, Friedman GD. Alco-            consumption: a risk factor for hemorrhagic and non-
18 years or older, are current drinkers      hol use and subsequent cerebrovascular disease hos-        hemorrhagic stroke. Am J Med. 1991;90:489-497.
                                             pitalizations. Stroke. 1989;20:741-746.                    28. Ben-Shlomo Y, Markowe H, Shipley M, Marmot
who have consumed at least 12 drinks         8. Shaper AG, Phillips AN, Pocock SJ, Walker M, Mac-       MG. Stroke risk from alcohol consumption using dif-
in the preceding year.78 Stroke is a ma-     farlane PW. Risk factors for stroke in middle aged Brit-   ferent control groups. Stroke. 1992;23:1093-1098.
                                             ish men [comments]. BMJ. 1991;302:1111-1115.                                ¨
                                                                                                        29. Paloma ki H, Kaste M. Regular light-to-
jor cause of death and disability in the                                                                moderate intake of alcohol and the risk of ischemic
                                             9. Goldberg RJ, Burchfiel CM, Reed DM, Wer-
United States and other countries.2 In       gowske G, Chiu D. A prospective study of the health        stroke: is there a beneficial effect? Stroke. 1993;24:
the United States, there are approxi-        effects of alcohol consumption in middle-aged and el-      1828-1832.
                                             derly men: the Honolulu Heart Program. Circulation.        30. Shinton R, Sagar G, Beevers G. The relation of al-
mately 600000 new stroke cases each          1994;89:651-659.                                           cohol consumption to cardiovascular risk factors and
year.1 Given the widespread consump-         10. Hansagi H, Romelsjo A, Gerhardsson de Verdier
                                                                        ¨                               stroke: the West Birmingham Stroke Project. J Neu-
                                             M, Andreasson S, Leifman, A. Alcohol consumption
                                                       ´                                                rol Neurosurg Psychiatry. 1993;56:458-462.
tion of alcohol in the general popula-       and stroke mortality: 20-year follow-up of 15077 men       31. Jamrozik K, Broadhurst RJ, Anderson CS, Stewart-
tion and the recognized health and eco-      and women. Stroke. 1995;26:1768-1773.                      Wynne EG. The role of lifestyle factors in the etiol-
nomic burdens of stroke, our findings        11. Iso H, Kitamura A, Shimamoto T, et al. Alcohol         ogy of stroke: a population-based case-control study
                                             intake and the risk of cardiovascular disease in middle-   in Perth, Western Australia. Stroke. 1994;25:51-59.
are both important and timely. Our           aged Japanese men. Stroke. 1995;26:767-773.                32. Beghi E, Boglium G, Cosso P, et al. Stroke and al-
study strongly suggests that reducing        12. Kiyohara Y, Kato I, Iwamoto H, Nakayama K, Fu-         cohol intake in a hospital population: a case-control
                                             jishima M. The impact of alcohol and hypertension on       study. Stroke. 1995;26:1691-1696.
alcohol consumption in heavy drink-          stroke incidence in a general Japanese population: the     33. Caicoya M, Rodriguez T, Corrales C, Cuello R,
ers should be an important approach          Hisayama Study. Stroke. 1995;26:368-372.                   Lasheras C. Alcohol and stroke: a community case-
                                             13. Palmer AJ, Fletcher AE, Bulpitt CJ, et al. Alcohol     control study in Asturias, Spain. J Clin Epidemiol. 1999;
to prevention of stroke in the general       intake and cardiovascular mortality in hypertensive pa-    52:677-684.
population. Our study also suggests that     tients: report from the Department of Health Hyper-        34. Sacco RL, Elkind M, Boden-Albala B, et al. The

©2003 American Medical Association. All rights reserved.                                      (Reprinted) JAMA, February 5, 2003—Vol 289, No. 5 587

protective effect of moderate alcohol consumption on       50. Beghi E, Bogliun G, Cosso P, et al. Cerebrovas-         64. Truelsen T, Gronbaek M, Schnohr P, Boysen G.
ischemic stroke. JAMA. 1999;281:53-60.                     cular disorders and alcohol intake: preliminary results     Intake of beer, wine, and spirits and risk of stroke: the
35. Thrift A, Donnan G, McNeil J. Heavy drinking, but      of a case-control study. Ital J Neurol Sci. 1992;13:        Copenhagen city heart study. Stroke. 1998;29:2467-
not moderate or intermediate drinking, increases the       209-214.                                                    2472.
risk of intracerebral hemorrhage. Epidemiology. 1999;      51. Hillbom M, Haapaniemi H, Juvela S, Palomaki H,   ¨      65. Kissela B, Sauerbeck L, Woo D, et al. Subarach-
10:307-312.                                                Numminen H, Kaste M. Recent alcohol consump-                noid hemorrhage: a preventable disease with a heri-
36. Zodpey SP, Tiwari RR, Kulkarni HR. Risk factors        tion, cigarette smoking, and cerebral infarction in young   table component. Stroke. 2002;33:1321-1326.
for haemorrhagic stroke: a case-control study. Public      adults. Stroke. 1995;26:40-45.                              66. Klatsky A, Armstrong M, Friedman G, Sidney S.
Health. 2000;114:177-182.                                  52. Haapaniemi H, Hillbom M, Juvela S. Lifestyle-           Alcohol drinking and risk of hospitalization for ische-
37. Malarcher AM, Giles WH, Croft JB, et al. Alcohol       associated risk factors for acute brain infarction among    mic stroke. Am J Cardiol. 2001;88:703-706.
intake, type of beverage, and the risk of cerebral in-     persons of working age. Stroke. 1997;28:26-30.              67. Turner C. How much alcohol is in a “standard
farction in young women. Stroke. 2001;32:77-83.            53. Hillbom M, Numminen H, Juvela S. Recent heavy           drink”: an analysis of 125 studies. Br J Addict. 1990;
38. Mazzaglia G, Britton A, Altmann D, Chenet L. Ex-       drinking of alcohol and embolic stroke. Stroke. 1999;       85:1171-1175.
ploring the relationship between alcohol consump-          30:2307-2312.                                               68. DerSimonian R, Laird N. Meta-analysis in clinical
tion and non-fatal or fatal stroke: a systematic re-                                           ¨
                                                           54. Juvela S, Hillbom M, Palomaki H. Risk factors for       trials. Control Clin Trials. 1986;7:177-188.
view. Addiction. 2001;96:1743-1756.                        spontaneous intracerebral hemorrhage. Stroke. 1995;         69. Greenland S, Longnecker MP. Methods for trend
39. Camargo CA Jr. Case-control and cohort studies         26:1558-1564.                                               estimation from summarized dose-response data, with
of moderate alcohol consumption and stroke. Clin           55. Haapaniemi H, Hillbom M, Juvela S. Weekend and          applications to meta-analysis. Am J Epidemiol. 1992;
Chim Acta. 1996;246:107-119.                               holiday increase in the onset of ischemic stroke in young   135:1301-1309.
40. Papadakis JA, Ganotakis ES, Mikhailidis DP. Ben-       women. Stroke. 1996;27:1023-1027.                           70. Begg CB. Publication bias. In: Cooper H, Hedges
eficial effect of moderate alcohol consumption on vas-     56. Blackwelder WC, Yano K, Rhoads GG, et al. Al-           LF, eds. The Handbook of Research Synthesis. New
cular disease: myth or reality? J R Soc Health. 2000;      cohol and mortality: the Honolulu Heart Study. Am J         York, NY: Russell Sage Foundation; 1994:399-409.
120:11-15.                                                 Med. 1980;68:164-169.                                       71. Zakhari S. Alcohol and the cardiovascular sys-
41. Eichner ER. Alcohol, stroke and coronary artery        57. Kagan A, Popper JS, Rhoads GG. Factors related          tem: molecular mechanisms for beneficial and harm-
disease. Am Fam Physician. 1988;37:217-221.                to stroke incidence in Hawaii Japanese men: the Ho-         ful action. Alcohol Health Research World. 1997;21:
42. Hillbom M, Kaste M. Alcohol abuse and brain in-        nolulu Heart Study. Stroke. 1980;11:14-21.                  21-29.
farction. Ann Med. 1990;22:347-352.                        58. Rodgers H, Aitken PD, French JM, Curless RH,            72. Gorelick P, Sacco R, Smith D, et al. Prevention of
43. Ashley MJ, Ferrence R, Room R, Bondy S, Rehm           Bates D, James FW. Alcohol and stroke: a case-              a first stroke: a review of guidelines and a multidisci-
J, Single E. Moderate drinking and health: implica-        control study of drinking habits past and present.          plinary consensus statement from the National Stroke
tions of recent evidence. Can Fam Physician. 1997;         Stroke. 1993;24:1473-1477.                                  Association. JAMA. 1999;281:1112-1120.
43:687-694.                                                59. Klatsky AL, Friedman GD, Siegelaub AB. Alcohol          73. Marmot MG. Alcohol and coronary heart dis-
44. Klatsky AL. Alcohol and cardiovascular diseases:       and mortality: a ten-year Kaiser-Permanente experi-         ease. Int J Epidemiol. 2001;30:724-729.
a historical overview. Novartis Found Symp. 1998;          ence. Ann Intern Med. 1981;95:139-145.                      74. Rimm EB, Klatsky A, Grobbee D, Stampfer MJ.
216:2-12.                                                  60. Klatsky AL, Armstrong MA, Friedman GD. Risk             Review of moderate alcohol consumption and re-
45. Gorelick PB, Kelly MA. Alcohol as a risk factor for    of cardiovascular mortality in alcohol drinkers, ex-        duced risk of coronary heart disease: is the effect due
stroke. Heart Dis Stroke. 1992;1:255-258.                  drinkers and nondrinkers. Am J Cardiol. 1990;66:            to beer, wine, or spirits. BMJ. 1996;312:731-736.
46. Hillbom M. Alcohol consumption and stroke: ben-        1237-1242.                                                  75. Wannamethee G, Shaper AG. Men who do not
efits and risks. Alcohol Clin Exp Res. 1998;22(suppl       61. Berger K, Ajani UA, Kase CS, et al. Light-to-           drink: a report from the British Regional Heart Study.
7):352S-358S.                                              moderate alcohol consumption and risk of stroke             Int J Epidemiol. 1988;17:307-316.
47. Wannamethee SG, Shaper AG. Alcohol, coronary           among US male physicians. N Engl J Med. 1999;341:           76. Shaper AG, Wannamethee G, Walker M. Alco-
heart disease and stroke: an examination of the J-shaped   1557-1564.                                                  hol and mortality in British men: explaining the U-
curve. Neuroepidemiology. 1998;17:288-295.                 62. Wannamethee SG, Shaper AG. Patterns of alco-            shaped curve. Lancet. 1988;2:1267-1273.
48. Beilin LJ, Puddey IB, Burke V. Alcohol and hyper-      hol intake and risk of stroke in middle-aged British men.   77. Maclure M. Demonstration of deductive meta-
tension: kill or cure? J Hum Hypertens. 1996;10            Stroke. 1996;27:1033-1039.                                  analysis: ethanol intake and risk of myocardial infarc-
(suppl 2):S1-S5.                                           63. Camargo CA Jr, Hennekens CH, Gaziano JM,                tion. Epidemiol Rev. 1993;15:328-351.
49. Monforte R, Estruch R, Graus F, Nicolas JM, Ur-        Glynn RJ, Manson JE, Stampfer MJ. Prospective               78. National Institute on Alcohol Abuse and Alcohol-
bano-Marquez A. High ethanol consumption as risk           study of moderate alcohol consumption and mortal-           ism. 10th Special Report to the US Congress on Al-
factor for intracerebral hemorrhage in young and           ity in US male physicians. Arch Intern Med. 1997;157:       cohol and Health. Rockville, Md: US Dept of Health
middle-aged people. Stroke. 1990;21:1529-1532.             79-85.                                                      and Human Services; 2000.

588    JAMA, February 5, 2003—Vol 289, No. 5 (Reprinted)                                            ©2003 American Medical Association. All rights reserved.

widely, with particularly high rates of use by internists and
physicians in the Northeast and the South.
Michael A. Steinman, MD
                                                                                     Name Omitted: In the Original Contribution entitled “Combination Therapy With
C. Seth Landefeld, MD                                                                Hormone Replacement and Alendronate for Prevention of Bone Loss in Elderly
San Francisco VA Medical Center                                                      Women: A Randomized Controlled Trial” published in the May 21, 2003, issue of
San Francisco, Calif                                                                 THE JOURNAL (2003;289:2525-2533), Michael McClurg, MD, should be added to
                                                                                     the list of members of the Data and Safety Monitoring Board on page 2532 after
Ralph Gonzales, MD, MSPH                                                             Peggy A. Norton, MD.
University of California, San Francisco
1. Gonzales R, Steiner JF, Lum A, Barrett PH Jr. Decreasing antibiotic use in am-
bulatory practice: impact of a multidimensional intervention on the treatment of     Error in Author’s Name: In the Review article entitled “Alcohol Consumption and
uncomplicated acute bronchitis in adults. JAMA. 1999;281:1512-1519.                  Risk of Stroke: A Meta-analysis” published in the February 5, 2003, issue of THE
2. Gonzales R, Malone DC, Maselli JH, Sande MA. Excessive antibiotic use for acute   JOURNAL (2003;289:579-588) in the byline, the initial letter “L.” was incorrectly
respiratory infections in the United States. Clin Infect Dis. 2001;33:757-762.       placed in front of the name of author Brian Lewis, MPH.

                                  CME ANNOUNCEMENT
                                  Online CME to Begin in Mid-2003
                                  In mid-2003, online CME will be available for JAMA/Archives journals
                                  and will offer many enhancements:
                                  •   Article-specific questions
                                  •   Hypertext links from questions to the relevant content
                                  •   Online CME questionnaire
                                  •   Printable CME certificates and ability to access total CME credits
                                  We apologize for the interruption in CME and hope that you will
                                  enjoy the improved online features that will be available in mid-2003.

2798    JAMA, June 4, 2003—Vol 289, No. 21 (Reprinted)                                        ©2003 American Medical Association. All rights reserved.

hkksew3563rd hkksew3563rd http://