OUP user menu

Factors Predicting Change in Frequency of Heavy Drinking Days among Alcohol-Dependent Participants in the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC)

Khaled Sarsour, Joseph A. Johnston, Denái R. Milton, Amy Duhig, Catherine Melfi, Howard B. Moss
DOI: http://dx.doi.org/10.1093/alcalc/ags036 443-450 First published online: 6 April 2012

Abstract

Aims: To discover the predictors of change in the frequency of heavy drinking (HD) over a 4-year period in alcohol dependent (AD)-individuals identified in the general population, namely, among participants of the US National Epidemiologic Survey on Alcohol and Related Conditions interviewed at Wave 1 (2001–2002) and at Wave 2 (2004–2005). Methods: The study cohort included subjects meeting DSM-IV criteria for AD in the past year at Wave 1 (= 1484), who were present at Wave 2 (= 1172) and had complete data on factors of interest (= 1123). Frequency of HD was defined as the number of HD days (HDD) (≥5 drinks per day for men and ≥4 for women). Change in frequency of HDD from baseline (Wave 1) to ∼3 years later (Wave 2) was determined. An analysis of covariance model (ANCOVA), adjusting for baseline HDD, was used to examine individual factors associated with change in frequency of HDD, while a multivariable regression model was employed to assess factors associated with change in frequency of HDD simultaneously. Results: Overall, there was a decrease in mean (SE) HDD [from 119.4 (1.8) at Wave 1 to 82.5 (2.1) at Wave 2, P < 0.0001]. Compared with smokers, non-smokers had a mean (SE) HDD reduction of 13.4 (6.7), P < 0.05. AD criteria of tolerance was significantly associated (P < 0.05) with less reduction in HDD. Change in depression/dysthymia status was associated with greater reduction in HDD in the ANCOVA model, but not the fully adjusted multivariable model. Conclusion: Findings from this study suggest that smoking and AD criteria of tolerance are important factors for long-term follow-up of AD patients and they should influence the selection of the kinds of interventions required for AD patients to achieve maximal therapeutic benefit.

INTRODUCTION

Alcohol dependence (AD) is a complex disorder, characterized by alcohol craving, loss of control over drinking behavior, physical dependence on alcohol and tolerance to the effects of alcohol. The 1-year prevalence of AD in the US adult population is 3.8%, and the lifetime prevalence is 12.5% (Grant, 1994, 1997; Caetano and Cunradi, 2002; Hasin et al., 2007). The natural history of AD is variable. AD is frequently comorbid with alcohol abuse (AA) and often takes a chronic course, with 25% of affected individuals remaining dependent 5 years after an initial diagnosis (Dawson, 2000b). AD contributes to a host of medical and psychiatric conditions, and imposes a substantial social and economic burden on affected individuals and society as a whole.

Heavy drinking (HD) (also known as hazardous drinking) is defined as drinking five (four) or more drinks in a single day for men (women) in accordance with the low-risk drinking guidelines proposed by the National Institute on Alcohol Abuse and Alcoholism (Alcoholism, 2005) and has been shown to be predictive of the development of AD and AA (Dawson and Archer, 1993). In addition, a growing literature supports the association between HD and a variety of adverse health outcomes (Dawson, 2000a; Rehm et al., 2003, 2009; Mokdad et al., 2004; Patra et al., 2007; Flowers et al., 2008). While controversy exists regarding the use of measures of drinking behavior as surrogate outcomes in AD patients, the use of HD days (HDD) as a treatment target in intervention studies appears to be gaining acceptance (Allen, 2003; Gastfriend et al., 2007; EMEA, 2009; Falk et al., 2010).

The predictors and correlates of HD in the general population have been found to involve a combination of individual and contextual variables across the life-course (Naimi et al., 2003; Balluz et al., 2004; Serdula et al., 2004; Hill and Angel, 2005; Karlamangla et al., 2006; Bernstein et al., 2007; Galea et al., 2007; Joutsenniemi et al., 2007; Ahern et al., 2008; Kestila et al., 2008). In the general population, HD was found to decline with increasing age and was found to be associated with male gender, lower socioeconomic status and smoking. On the other hand, quitting smoking and the life transition to adult roles and responsibilities such as marriage were found to be independently protective against HD (Chilcoat and Breslau, 1996; Bachman et al., 1997; Karlamangla et al., 2006; Bjork et al., 2008). Distal contextual factors such as peer drinking norms and ‘neighborhood disorder’, quantified in terms of conditions that signify the breakdown of social order such as unsafe streets, the presence of abandoned housing and high unemployment, were also associated with HD (Hill and Angel, 2005; Ahern et al., 2008).

The purpose of the present study was to understand the natural history of HD behavior in a cohort of prevalent AD subjects, and to examine factors that may predict change in HDD frequency overtime. In this way, we hope to provide some insight into the complex interplay between behavior and psychopathology, and to provide a context for those considering the use of HDD as a clinical trial endpoint. The availability of data from the second wave of the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC), including DSM-IV-based diagnoses of alcohol-use disorders and detailed information on drinking behavior and adverse consequences, makes possible for the first time, analyses of change in HD behavior which take into account the dynamic nature of the AD and other behaviors and conditions which can influence drinking behavior.

METHODS

Sample

NESARC is a nationwide household survey designed and conducted by the National Institute on Alcohol Abuse and Alcoholism. The NESARC sample (n = 43 093) is representative of the civilian, non-institutionalized adult population of the USA and includes people living in households, military personnel living off base and people residing in selected group quarters. A single adult was randomly selected from each sample household for a face-to-face interview completed by trained US Census Bureau field representatives. The overall response rate of the survey at Wave 1 was 81%. The analyses used in this study were based on the 1484 respondents who met DSM-IV criteria for AD with or without abuse in the past year at Wave 1 of the survey. Of the 1484 respondents with AD, the study cohort included the 1123 (76%) subjects with complete data at both waves of the survey. A duration of 3 years elapsed between Waves 1 and 2.

The NESARC used a multistage stratified sampling design, oversampling adults 18–24 years of age, African Americans and Hispanics to ensure adequate numbers for comparisons involving these subgroups. Sampling weights are included in the NESARC data set so that information from NESARC can be extrapolated to produce population estimates. A more detailed description of the NESARC sampling design and methodology can be found in previously published studies (Grant, 2004; Hasin et al., 2007).

Measures

The frequency of HDD was determined using the survey question that asked how often the participant drank five or more drinks (for men) or four or more drinks (for women) in a single day during the past 12 months. The categorical responses from that question were converted numerically in the following manner: every day = 365 days; nearly every day = 286 days; 3–4 times a week = 182 days; 2 times a week = 104 days; once a week = 52 days; 2–3 times a month = 30 days; once a month = 12 days; 7–11 times in the last year = 9 days; 3–6 times in the last year = 4.5 days; 1 or 2 times in the last year = 1.5 days and never in the last year = 0 days. The outcome measure used in this study was a change in frequency of HDD from Wave 1 to Wave 2 (HDDWave 2–HDDWave 1).

For many variables, a variety of assessments were available for use, administered at either Wave 1 or Wave 2, each with reference to a different time period (e.g. past year, since last wave and lifetime). AA and AD status, based on DSM-IV criteria, was determined at both waves based on past year assessments, as was frequency of HDD. The diagnosis of AA is established based on the presence of any one of the AA criteria, while an AD diagnosis requires at least three of seven AD criteria to be met (see Table 2 for individual criteria).

Sociodemographic factors, including gender, race/ethnicity, education, marital status, employment status, family history of AD (among all first and/or second degree relatives) and family history of depression (among first-degree relatives excluding children and/or second-degree relatives excluding grandchildren) were determined based on Wave 1 data. Smoking since Wave 1 assessed at Wave 2 was used to define current smoking. Past treatment seeking for alcohol-use disorders (which included seeking treatment at alcoholics anonymous, a health care professional or any community agency or professional), and the presence of other psychiatric and substance abuse disorders, as well as individual DSM-IV dependence and abuse criteria were based on a combination of Wave 1 past year and/or Wave 2 since last wave, excluding last 12 months assessments. Finally, change in depression/dysthymia status was defined as the change from the year preceding Wave 1or the 2 years following Wave1 to the assessment reported for the year preceding Wave 2.

Statistical analyses

The primary objective of this study was to determine factors associated with change in the frequency of HDD from Wave 1 to Wave 2 in NESARC. The factors included in the analysis were gender, categorical age (<30, 30–39, 40–49, ≥50), ethnicity/race, education, marital status, employment status, family history of AD, family history of depression, smoking status, past treatment seeking, change in depression/dysthymia status, bipolar disorder, combined anxiety group (generalized anxiety disorder, panic disorder without agoraphobia or social phobia), cannabis use/abuse, other substance use/abuse (cocaine, opioid and amphetamine), individual AA (i.e. role failure, physically hazardous use, drink despite social problems and legal problems) and AD criteria (i.e. tolerance, withdrawal, persistence, drink larger amounts/longer time, time spent drinking or recovering, reduced activities and drink despite physical/psychological problems). Association between each individual factor and change in the frequency of HDD was determined using an analysis of covariance (ANCOVA) model adjusting for baseline HDD. Those individual factors significant in the ANCOVA models were subsequently included in a multivariable regression model, in addition to baseline HDD, to assess the association with change in the frequency of HDD for all independent factors simultaneously. All statistical tests of differences in independent measures were conducted using a two-sided significance level of 0.05. All estimates are presented as least squares means and standard errors (SE). Collinearity between study variables was investigated using a correlation matrix and the use of variance inflation factor (VIF). No correlation between the study variables was >0.5. All model VIF values were <5 consistent with there being no significant collinearity in any of the models. All analyses were computed using specific procedures in SAS version 9.1 (SAS Institute Inc., Cary, NC, USA) designed for the analysis of complex survey data.

RESULTS

Descriptive characteristics

Descriptive characteristics of the subjects in the study cohort are presented in Table 1. Fifty-four percents were younger than age 30, while 7% were 50 years of age or older. The majority of subjects was male (68.0%), white (70.9%), employed (81.9%) and had health insurance (69.0%). Fifty-five percents of the study cohort had a college degree or some college education and less than half were married or living as if married. Most had a family history of AD (74.0%), and about half had a family history of depression, while 40.5% had one or more psychiatric comorbidities. Fifty-two percents were current smokers. Interestingly, only 16.0% had a past history of treatment seeking for alcohol-use disorders.

View this table:
Table 1.

Descriptive characteristics of study sample (n = 1123)

Measuren (%)
Demographics
 Male705 (68.0)
 Age (years)
  <30537 (53.5)
  30–39258 (21.5)
  40–49219 (17.8)
  ≥50109 (7.3)
 Race/ethnicity
  White679 (70.9)
  Black188 (10.7)
  American Indian30 (3.4)
  Asian14 (2.0)
  Hispanic212 (13.0)
 Employed902 (81.9)
 Insured784 (69.0)
 Education
  Less than high school192 (16.6)
  High school/general education diploma (GED)334 (28.4)
  Some college306 (28.6)
  College degree291 (26.5)
 Marital status
  Married357 (34.9)
  Never married526 (49.2)
  Other240 (15.9)
Family history
 Alcohol dependence842 (74.0)
 Depression591 (53.8)
Comorbidities
 Depression/dysthymiaa333 (29.0)
 Bipolar disordera145 (12.6)
 Anxietya228 (22.4)
 Cannabis use/abusea229 (22.5)
 Other substance use/abusea,b130 (12.8)
 At least one psychiatric comorbidity456 (40.5)
Behaviors
 Smoking status570 (51.9)
 Past treatment seeking182 (16.0)
  • aThepresence of these disorders was based on past year prior to Wave 1 or since Wave 1 but 12 months prior to Wave 2.

  • bThis variable includes cocaine, opioid and/or amphetamine use/abuse.

The most commonly endorsed AD criterion at Wave 1 was ‘drinking larger amounts for longer periods of time than intended’ (92.8%), followed by ‘persistent desire or unsuccessful efforts to cut down or control drinking’ (80.8%). The least commonly endorsed criterion was ‘reducing important social, occupational or recreational activities because of alcohol use’ (21.4%). Approximately three-fourths of subjects also met the DSM-IV criteria for AA. Among all of the AA criteria, recurrent alcohol use in situations in which it is physically hazardous was most prevalent (68.0%). A complete listing is presented in Table 2.

View this table:
Table 2.

The prevalence of DSM-IV alcohol dependence and abuse criteriaa

Study sample (n = 1123), n (%)
Alcohol dependence criteria
 Tolerance881 (79.2)
 Withdrawal847 (75.6)
 Drinking larger amounts or over a longer period than was intended1030 (92.8)
 Persistent desire or unsuccessful efforts to cut down or control drinking919 (80.8)
 More time is spent in activities to obtain the substance, use the substance or recover from its effects579 (53.0)
 Drinking continued despite physical/psychological problems701 (63.5)
 Reduced activities225 (21.4)
 At least three dependence criteria1123 (100)
 At least four dependence criteria832 (74.0)
 At least five dependence criteria549 (51.4)
 At least six dependence criteria304 (28.8)
 All seven dependence criteria128 (12.2)
Alcohol abuse criteria
 Role failure240 (23.1)
 Physically hazardous use732 (68.0)
 Drinking despite legal problems150 (13.8)
 Drinking despite social problems417 (38.7)
 At least one abuse criteria825 (75.8)
 At least two abuse criteria439 (41.2)
 At least three abuse criteria210 (20.1)
 At least four abuse criteria65 (6.3)
Alcohol dependence only298 (24.2)
Alcohol dependence and alcohol abuse825 (75.8)
  • aTime-frame includes past year prior to Wave 1 or since Wave 1 but 12 months prior to Wave 2.

The DSM-IV criteria endorsed during the past year at Wave 1 for those AD subjects included in this study was compared with AD subjects not included in the study due to loss to follow-up between Wave 1 and Wave 2. This assessment revealed that a higher percentage of AD subjects lost to follow-up endorsed five of the seven DSM-IV criteria compared with those included in the analysis. In addition, mean (SE) number of HDD was higher for those lost to follow-up (137.1 [2.6]) than for those included in the study (120.1 [1.9]; P = 0.0539).

ANCOVA and multivariable results: factors associated with change in the frequency of HDD

ANCOVA and multivariable results are presented in Table 3. Variables not significantly associated with the study outcome were excluded from the multivariable model. The variables excluded from the final model included age, race/ethnicity, marital status, employment status, family history of depression, past history of anxiety, past treatment seeking, DSM-IV AD criteria of withdrawal and DSM-IV AA criteria of role failure.

View this table:
Table 3.

aANCOVA and multivariable results: factors associated with change in the frequency of heavy drinking days

ANCOVA resultsaMultivariable results
MeasureΔ HDD LS means (SE)P-valueΔ HDD LS means (SE)P-value
HDD at Wave 1Adjustment measure−0.77 (0.03)<0.0001
Demographics
 Gender
  Female−46.2 (5.9)0.0228−24.2 (12.0)0.3902
  Male−28.3 (4.7)4.3 (11.2)
 Age (years)
  <30−42.6 (4.9)0.0880Not included in model
  30–39−31.2 (7.3)
  40–49−14.4 (10.9)
  ≥50−27.7 (12.7)
 Race/ethnicity
  White−36.5 (4.5)0.8563Not included in model
  Black−27.8 (11.9)
  American Indian−23.7 (29.0)
  Asian1.5 (42.7)
  Hispanic−34.3 (8.0)
 Education
  Less than high school−13.3 (9.7)0.0079−9.1 (12.7)0.3103
  High school/GED−25.7 (7.6)−19.2 (11.9)
  Some college−39.1 (6.4)−26.5 (11.9)
  College degree−50.4 (6.4)−29.9 (12.3)
 Marital status
  Married−33.6 (6.6)0.9567Not included in model
  Never married−35.1 (5.1)
  Other−31.9 (10.5)
 Employed
  Yes−36.2 (4.0)0.2766Not included in model
  No−24.5 (9.8)
Family history
 Alcohol dependence
  Yes−30.1 (4.3)0.0306−21.9 (10.8)0.8432
  No−45.4 (5.9)−20.5 (11.4)
 Depression
  Yes−35.3 (4.8)0.7093Not included in model
  No−32.6 (5.6)
Behaviors
 Smoking status
  Yes−20.8 (5.3)0.0002−14.5 (10.8)0.0471
  No−48.3 (4.9)−27.9 (11.2)
 Past treatment seeking
  Yes−13.5 (13.0)0.0878Not included in model
  No−37.9 (4.0)
Comorbidities
 Δ Depression/dysthymia status
  No depression/dysthymia at either wave−36.5 (4.1)0.0037−17.8 (11.3)0.0646
  Depression/dysthymia resolved−54.9 (7.2)−39.8 (10.3)
  New onset of depression/dysthymia4.1 (22.4)−4.8 (22.8)
  Depression/dysthymia at both waves−9.1 (13.4)−22.3 (13.7)
 Bipolar disorder
  Yes−7.8 (11.8)0.0147−12.2 (13.3)0.1616
  No−37.9 (3.7)−30.1 (11.2)
 Anxiety
  Yes−27.8 (8.5)0.4079Not included in model
  No−35.9 (4.2)
 Cannabis use/abuse
  Yes−14.4 (9.0)0.0151−22.5 (13.3)0.7939
  No−39.8 (4.3)−19.9 (9.8)
 Other substance use/abuseb
  Yes2.5 (13.0)0.0028−11.4 (13.9)0.1499
  No−39.4 (3.8)−31.0 (10.9)
Alcohol dependence criteria
 Tolerance
  Yes−29.0 (4.1)0.0009−13.0 (10.7)0.0394
  No−53.1 (6.2)−29.4 (11.8)
 Withdrawal
  Yes−33.7 (4.1)0.8611Not included in model
  No−35.2 (7.6)
 Drink larger amounts/longer time
  Yes−31.2 (3.7)<0.0001−15.2 (9.6)0.1716
  No−70.5 (8.5)−27.1 (12.9)
 Persistence
  Yes−30.5 (4.3)0.0121−15.6 (10.5)0.1189
  No−49.2 (6.1)−26.8 (11.6)
 Time spent drinking or recovering
  Yes−23.3 (5.3)0.0025−21.6 (10.7)0.8971
  No−46.0 (5.0)−20.8 (11.4)
 Drink despite physical/psychological problems
  Yes−24.5 (4.9)0.0002−19.9 (10.3)0.6800
  No−50.5 (4.9)−22.5 (11.6)
 Reduced activities
  Yes−10.1 (10.4)0.0077−15.0 (12.8)0.3020
  No−40.4 (3.9)−27.4 (11.3)
 Δ Alcohol dependence status
  Alcohol dependence resolved−60.9 (3.8)<0.0001−43.6 (10.8)<0.0001
  Alcohol dependence at both waves13.2 (6.7)1.2 (11.7)
Alcohol abuse criteria
 Role failure
  Yes−19.5 (9.1)0.0741Not included in model
  No−38.4 (4.3)
 Physically hazardous use
  Yes−24.4 (4.7)<0.0001−15.9 (12.6)0.3811
  No−54.1 (5.1)−26.4 (11.6)
 Legal problems
  Yes−9.2 (10.6)0.0126−18.4 (12.6)0.6130
  No−38.0 (3.9)−24.0 (11.1)
 Drink despite social problems
  Yes−16.2 (8.0)0.0022−22.3 (12.1)0.8164
  No−45.2 (4.0)−20.0 (11.1)
 Δ Alcohol abuse status
  No alcohol abuse at either wave−53.8 (5.4)<0.0001−27.1 (14.1)<0.0001
  Alcohol abuse resolved−70.8 (5.6)−53.0 (11.3)
  New onset of alcohol abuse−2.5 (14.2)−5.6 (15.7)
  Alcohol abuse at both waves9.5 (6.9)0.9 (12.4)
  • LS, least squares.

  • aAnalysis of covariance (ANCOVA) results using a model with change in the frequency of HDDs as the dependent variable and the individual measure as independent variable adjusting for HDD at Wave 1.

  • bThis variable includes cocaine, opioid and/or amphetamine use/abuse.

There was an overall reduction in the frequency of HDD from Wave 1 to Wave 2 in the study cohort. Mean (SE) HDD decreased from 119.4 (1.8) days at Wave 1 to 82.5 (2.1) days at Wave 2. We observed robust associations with change in the frequency of HDD for the following factors: baseline HDD, smoking status, DSM-IV AD criteria of tolerance, change in AA status and change in AD status. Adjusting for all other factors, a mean (SE) of 1 day increase in baseline HDD was associated with a 0.77 (0.03) day reduction in HDD at Wave 2. Change in AA status was significantly associated with a reduction in HDD (P < 0.0001). Those AD subjects who also met the AA criteria at Wave 1 but not at Wave 2 had a mean (SE) reduction of 53.0 (11.3) days (improved), while those who met the AA criteria at Wave 2 but not at Wave 1 experienced a mean (SE) reduction of 5.6 (15.7) days. Similarly, a change in AD status was significantly associated with a reduction in HD (P < 0.0001). Those who no longer met AD criteria at Wave 2 had a mean (SE) reduction of 43.6 (10.8) days while those who remained alcohol dependent at Wave 2 had a mean (SE) increase of 1.2 (11.7) days (worsened).

Endorsement of the DSM-IV AD criterion of tolerance was associated with significantly less reduction in HDD (P = 0.0394). AD subjects who did not report tolerance experienced a significant mean (SE) reduction of 29.4 (11.8), days while those who reported tolerance experienced a mean (SE) reduction of 13.0 (10.7) days. None of the other individual AD or AA criteria was significantly associated with change in the frequency of HDD in this study cohort.

Controlling for other factors in the multivariable model including baseline differences in HDD, change in AD and AA status and individual DSM-IV criteria, smoking status was associated with change in the frequency of HDD. Non-smokers had a mean (SE) reduction of 27.9 (11.2) days while those who smoked had a mean (SE) reduction of 14.5 (10.8) days. Neither cannabis use/abuse nor other substance use/abuse was significantly associated with change in HDD. Change in depression/dysthymia between Wave 1 and Wave 2 was associated with change in frequency of HDD (P = 0.0037) in the ANCOVA model but not the multivariable (fully adjusted) model. Although two-thirds of the subjects did not experience depression/dysthymia at either wave, those whose depression/dysthymia resolved from Wave 1 to Wave 2 (14.8%) had a mean (SE) reduction of 54.9 (7.2) days while those who had onset of depression/dysthymia between Wave 1 and Wave 2 (3.1%) had a mean increase (SE) of 4.1 (22.4) days in their HD.

Only 16.0% of the study cohort at baseline had reported previous treatment for their alcohol-use disorder, and this was not significantly associated with change in HDD between Wave 1 and Wave 2. Age, gender, education, and family history of AD were not significantly associated with change in the frequency of HDD between Wave 1 and Wave 2.

DISCUSSION

Using data from both waves of NESARC, we identified several factors independently associated with a change in the frequency of HDD among a nationally representative sample of AD subjects. In these exploratory analyses, baseline HDD and change in AA or AD status were associated with a greater reduction in HDD overtime. On the other hand, smoking and the DSM-IV AD criteria of tolerance were found to be significantly associated with less dramatic reductions in HD overtime. In the multivariable model, none of the sociodemographic factors which have been shown to be associated with HD in the general population were found to be associated with the change in HD frequency in this AD sample. Interestingly, resolution of depression or dysthymia was associated with greater reduction in HDD in the ANCOVA model but not the multivariable model.

While the individual sociodemographic variables of gender and education were associated with a change in HDD in the ANCOVA model, none remained so in the multivariable model. These variables have been associated with the onset of HD and AD (Grant, 1997; Naimi et al., 2003; Hasin et al., 2007; Leonard and Homish, 2008; Merrick et al., 2008). However, they may not necessarily be highly associated with its course or natural history. The lack of a significant association between changes in the frequency of HDD and family history should be interpreted with caution due to the broad nature of this variable's definition. In this study, a respondent was considered to have family history of AD if they reported any first- and/or second-degree relatives with AD versus other studies which used first-degree relatives only to define family history. Future studies should consider examining these characteristics in more detail and investigate the possibility of ascertaining family history using objective measures not subject to response bias.

It is interesting to note that age in this sample was not associated with change in HDD. This is inconsistent with studies of the general population. In studies of the general population, the risk of HD is found to decline with age with a particularly prominent decrease in the frequency of HD from 19 to 25 years old cohort to older age cohorts (Karlamangla et al., 2006; Kerr et al., 2009). Understanding the mechanisms, including the behavioral and contextual variables, by which age plays a differential role in determining risk of onset and subsequent reduction of HD in AD populations is important for future research.

Our finding that smoking is associated with less reduction in HDD is consistent with the smoking and AD literature. Current smokers have a greater risk of AD and increased frequency and duration of smoking is associated with greater severity of AD (Gulliver et al., 1995; Daeppen et al., 2000; John et al., 2003a,b). The finding that smokers have a less dramatic decrease in their HD overtime suggests the need to target risk factors holistically, although it is not clear whether smoking is causally associated with AD or if it is simply a marker of an underlying predisposition to addictive behavior.

The evidence surrounding the nature of the link between AD and depression or dysthymia is controversial. A number of studies found that AD is causally associated with the risk for major depression (Hasin and Grant, 2002; Fergusson et al., 2009). On the other hand, the self-medication hypothesis posits that depressed patients use alcohol as a coping mechanism to control or alleviate their symptoms, ultimately leading to AD (Khantzian, 1997). Recent evidence is consistent with the hypothesis that major depression can actually precede AD and increases its risk even after adjusting for confounding by genetic and other factors (Lukassen and Beaudet, 2005; Kuo et al., 2006). Taken together, these studies suggest that the association between AD and depression is complex and bidirectional (for an example, see Gilman and Abraham (2001)) and would depend on the target population or setting. This study found that change in depression/dysthymia was not associated with HDD. This study did not, however, include data on the underlying reasons for change in depression or dysthymia. Further investigation is required to understand the bidirectional interactions between depression and AD and to understand the role of treatment for the underlying depression in the reduction of HDD.

Importantly, when adjusting for other factors, past treatment-seeking was not associated with reduction in HD. However, only a very small percentage of patients sought treatment. Nearly 70% of this nationally representative AD sample has current health insurance. However, only 16.0% reported any history of treatment seeking (and of those, a small percentage probably seeks treatment at a healthcare professional's office (Cunningham and Blomqvist, 2006)). This further highlights the need to better understand this therapeutic gap (Town et al., 2006). Additional research is clearly needed to better understand the multi-dimensional barriers to treatment seeking in this problematic population.

Findings from this study should be interpreted in light of a number of limitations. The study cohort consisted of those respondents who met DSM-IV criteria for AD with or without abuse in the past year at Wave 1 of the survey. This potential selection bias is a limitation since those participants who met the criteria between waves were not included in the analysis. In addition, 4% of the Wave 1 AD subjects were not included in the present analysis due to missing data. Those subjects had more severe AD as indexed by greater percentage of them endorsing DSM-IV criteria. If those participants not included in the study had a systematically different change in frequency of HDD, different conclusions could have been drawn. Another limitation of the study is the definition of the frequency of HDD. The respondents were asked to check one of 11 categories to represent how often they drank five (four) or more drinks in a single day as opposed to reporting the exact number of HDD. Finally, findings about the role of age, smoking and depression/dysthymia need to be independently replicated in other national surveys to rule out the possibility of spurious findings due to type 1 error since the P-values in this study were not adjusted for multiple comparisons.

Despite these limitations, our findings suggest that smoking status and endorsement of the DSM-IV criteria of tolerance are associated with change in the frequency of HDD in AD subjects. This natural history observation requires replication and further investigation because it may have implications for long-term follow-up of AD patients as well as in the selection of the kinds of interventions required for AD patients to achieve favorable outcomes. In general, the results support the utility of HDD as a measure of outcome in research involving individuals with AD.

Funding

This study was funded by Eli Lilly and Company.

Conflict of interest statement. K.S., J.A.J., D.R.M. and C.M. are full time employees and stock holders of Eli Lilly and Company. At the time this study was conducted A.D. was a full time employee at Eli Lilly and Company. H.M. received no financial compensation for his scientific contributions to this study.

Acknowledgments

The authors thank Rachelle Dawn Rodriguez for helpful comments on an earlier draft of this manuscript.

REFERENCES

View Abstract