Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
October 2002

Consequences and Correlates of Adolescent Depression

Author Affiliations

From the Department of Health Policy and Management, Mailman School of Public Health, Columbia University, New York, NY (Dr Glied); and Section on Development and Affective Neuroscience, Mood and Anxiety Disorders Program, National Institute of Mental Health–Intramural Research Program, Bethesda, Md (Dr Pine).


Copyright 2002 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2002

Arch Pediatr Adolesc Med. 2002;156(10):1009-1014. doi:10.1001/archpedi.156.10.1009

Objective  To examine the correlates and consequences of high levels of depressive symptoms among adolescents.

Design  Secondary analysis of the 1997 Commonwealth Fund Survey of the Health of Adolescent Girls, a survey of a nationally representative sample of 4648 adolescent boys and girls between the ages of 10 and 18 years, inclusive, conducted in school settings. The self-administered questionnaire contains a screening instrument for depression based on the Children's Depression Inventory.

Outcome  Days of school missed, performance at grade level, alcohol use, drug use, smoking, and bingeing.

Results  After controlling for sociodemographics, life events, sexual abuse, physical abuse, and exposure to violence, relative to other children, children and adolescents with high degrees of depressive symptoms missed about 1 day more of school in the month preceding the survey (P<.05) and had higher odds of smoking (odds ratio, 1.84; P<.001), bingeing (odds ratio, 2.02; P<.001), and suicidal ideation (odds ratio, 16.59; P<.001).

Conclusion  High levels of depressive symptoms are correlated with serious and significant consequences, even after controlling for life circumstances.

THE DIRECT COSTS of medical care utilization and the morbidity and mortality costs of mental health problems for children younger than 15 years has been estimated at $2 billion in 1985, but estimates for youth younger than 19 years that include all related costs, including costs of juvenile justice and educational programs, have ranged as high as $20 billion.1,2

Depression in adolescence might also generate important nonmedical costs in several ways. First, depression may lead girls and boys to miss school or to fall behind in school. Education is a critical determinant of adult earnings, so if school attendance and performance are substantially affected by depression, adolescents may lose earnings in the future. Depression may inhibit school performance of children and adolescents, just as such symptoms reduce work performance among adults.35 Second, depression may affect other aspects of well-being. Such effects could occur through a connection between depression and dangerous behaviors, such as alcohol and drug use, bingeing, and smoking. Children with emotional and behavioral disorders in general are significantly more likely to experience substance use and are at higher risk of involvement with the juvenile justice system.6,7 There is also a suggestion that adolescent depression affects susceptibility to infectious disease.8 Risky behaviors are quite prevalent among youth. Data from the 1999 Youth Risk Behavior Surveillance Survey indicate that more than one third of youth in grades 9 through 12 currently smoke cigarettes, one half currently use alcohol, and more than one fourth currently use marijuana.9 Further, depression may raise the risk of suicide in children and adolescents, as it does in adults.10,11 According to data from the Youth Risk Behavior Surveillance Survey, nearly 20% of youth seriously considered attempting suicide during the preceding year.9

It is difficult to assess the consequences of depressive symptoms because depression in adolescents is often associated with many other factors that also raise the risk of undesirable behaviors and outcomes. Mental health problems in adolescents tend to be concentrated in the most disadvantaged groups—children from minority groups, from single-parent families, and from low-income families. Furthermore, family studies suggest that the prevalence of depression is higher among adolescents from families that include a parent with depression, and these children may be at risk for other poor outcomes as well.12

Adolescent depression may also be associated with environmental adversity. The relationship between depression and extreme stress has been demonstrated in children subjected to natural disasters, children who are homeless, and children subjected to physical or sexual abuse.1315 While these studies consistently note associations between depression and extreme adversity, the findings are limited by the nature of the generally nonrepresentative samples in most studies.

The relationship between adolescent depression and other, less extreme life events has been examined primarily in clinic-based samples. While these studies consistently note an association between life events and adolescent depression, the findings are limited by the referred nature of these samples.1619 This relationship has been examined in 3 epidemiologic samples, with each study noting a consistent relationship between the 2 constructs.2022

In this article, we examine the prevalence and correlates of depressive symptoms among children and adolescents. We then turn to the consequences of depression in adolescence: the degree to which depressive symptoms are correlated with school performance and with dangerous behaviors, particularly alcohol and drug use and eating problems. We examine the extent to which these negative outcomes associated with depression persist after controlling for sociodemographic and other risk factors that are associated with both higher rates of depression and higher risk of problem indicators. Previous research has shown associations between depression and some of these risk behaviors. This analysis broadens the range of behaviors considered to include school performance as well as a range of unhealthy behaviors. It describes these associations in a large community sample. Assessing the relationship between depression and adverse outcomes is complicated by the fact that depression is also correlated with other factors that raise the risk of adverse outcomes, such as life events and abuse. Moreover, some correlates of adolescent depression, such as abuse or adverse life circumstances, may carry independent risks for adverse consequences, such as excessive alcohol use.

Under these circumstances, relatively large studies in representative samples of adolescents, such as the Commonwealth Fund Survey of the Health of Adolescent Girls (CFSAG), provide particular advantages. Specifically, relative to smaller or less representative studies, studies based in large representative samples facilitate more precise, generalizable estimation of independent association that both adolescent depression and correlates of depression, such as adverse life events or sexual abuse, exhibit with various types of adverse outcomes.


We used data from the CFSAG, a nationally representative sample of adolescents.23 The CFSAG was a classroom-based study. Classrooms were selected for the survey in a 2-stage, stratified, and clustered sampling process (stratification variables involve school type, grade coverage, urbanicity, and region). Replacements for schools choosing not to participate were selected by nearest ZIP code within the same cell. This survey design shows variations similar to a random sample about 88% as large. In-class questionnaires were administered to 6748 students (3586 girls and 3162 boys) in grades 5 through 12. Approximately half (3216) were in grades 5 through 8 and half (3532) were in grades 9 through 12. All of the students in the classes selected completed the survey. A total of 297 schools, including public, private, and parochial schools, participated in the nationally representative survey. An oversample of 32 urban schools was included to facilitate a more complete analysis of responses by ethnicity and income. Fieldwork was conducted by Louis Harris and Associates Inc, Rochester, NY, from December 1996 to June 1997 and sponsored by the Commonwealth Fund, New York, NY.23 Data from students were collected in compliance with each participating school's own informed consent procedure. The protocol for this analysis of the data obtained from the Commonwealth Fund survey received expedited review approval from the Columbia-Presbyterian Medical Center Institutional Review Board, New York. For the present study, the sample was restricted to those 4648 respondents who were between the ages of 10 and 18 years and had completed survey questions on depressive symptoms, family structure, income, and ethnicity. Respondents with incomplete data were significantly more likely to be behind a grade and to express suicidal ideation (P<.001), were somewhat less likely to drink (P<.02), and (among the small minority of subject with incomplete data who responded to questions on depressive symptoms) had slightly, but not significantly, more depressive symptoms than did those with complete data. Thus, while these results may understate the negative correlates of depressive symptoms, it is unlikely that other missing subjects can account for the positive associations we found in the remaining sample. Other studies have used these data for a variety of purposes, including to describe prevalence and risk factors for mental illness, to study adolescent access to medical care, and to explore adolescents' preferences about physician characteristics.2427

The survey asked adolescents extensive questions about their health status, risk behaviors, and school performance. Respondent demographic and socioeconomic information was also reported. This included age, sex, race, region, urban or suburban residence, and family financial status. Family finances were assessed on the basis of whether the family had money problems frequently, occasionally, rarely, or never. One question was also asked concerning parental history of depression. The survey asked adolescents whether they engaged in risk behaviors (overeating, smoking, and drinking alcohol), how many days of school they missed because of illness in the preceding month, and their grade level. We used the information on age and grade to assess whether the child was below the age-appropriate grade level. The survey also asked adolescents whether they engaged in risk behaviors when experiencing stress. The format of the questions was multiple choice, with responses grouped into categories. Additional details on the student questionnaire can be found in Simantov et al.28

Depressive symptoms were assessed with the Children's Depression Inventory (CDI). This self-report questionnaire was originally developed by Kovacs in 198529 as a screening measure for the diagnosis of major depression in children. The questionnaire consists of a series of forced-choice items, for which an adolescent marks one of the statements most consistent with his or her current mental state. This allows a rating of depressive symptoms to be made on a 3-point scale, from absent to definitely present. These questions rate current depressive symptoms that had been present during the preceding 2 weeks. The CDI exists as both a 26-item and a 10-item questionnaire, with the 10-item version being used to screen for major depression. The measure exhibits satisfactory internal reliability and test-retest reliability, as well as satisfactory predictive validity for the clinical diagnosis of major depression in children.30

The CFSAG used a 9-item version of the CDI. This version included all of the items in the 10-item CDI except for item 3 ("I do most things OK"). The highest possible score on our 9-item CDI is 18. We used a score of 9 or higher on the CDI to estimate the prevalence of major depression symptoms in this sample. Previous clinical validations of the short-form administration of the CDI have established that a cutoff score of 9 optimizes the sensitivity and the specificity of the screening instrument.30 Since our data used the 9-item version of the CDI, a cutoff score of 9 should produce lower-bound estimates of the prevalence of depressive symptoms. We repeated these analyses with raw CDI scores used in place of the cutoff indicator. These results were uniformly stronger than those reported in this article.


We conducted the analysis with Software for Statistical Analysis of Survey Data.31 Our analysis incorporated CFSAG sample weights. We used χ2 and t tests for bivariate relationships between depression, incomes, factors, and problem indicators. We used multivariate logistic and linear regressions to estimate multivariate relationships. The prevalence of major depression in children and adolescents has been examined in many relatively large epidemiologic surveys that have relied on standardized interviews.3236 As reviewed by Harrington,37 the prevalence in the community is approximately 5%, although there is variability across studies, depending on the stringency of the applied impairment threshold and the measure used to make the diagnosis. In the CFSAG, the overall prevalence of depression in the (weighted) sample was 7.1%.


In the CFSAG, the prevalence of depression in boys was 5%, whereas that in girls was 9%. Figure 1 shows the prevalence of depression by age for boys and girls. As expected, rates were quite low among 10-year-old girls and boys. Rates for girls rapidly rose above those for boys and were much higher by age 14 years. Table 1 provides prevalence rates for boys and girls by race, family structure, family history of depression, and family income.

Image not available

Rates of depression among boys and girls.

Table 1. 
Image not available
Prevalence of Depression by Socioeconomic Characteristics

In these data, rates of depression do not vary substantively or significantly among race and ethnic groups. Previous studies have shown higher rates of mental health problems among children from divorced or single-parent families.3840 In these data there was an increased prevalence of depression in boys with divorced parents. Among those who responded to the question (three fourths of the full sample), nearly one third reported that a family member had had depression. Adolescents from such families were much more likely to meet criteria for depression than were those who did not report a family history of depression. Finally, rates of depression were much higher—4 times as high—among girls and boys in very-low-income families than among girls and boys in high-income families. This finding suggests that environmental factors may contribute to depression risk.


Table 2 provides prevalence rates for depression in girls and boys, by history of physical or sexual abuse, and by the number of life events experienced in the preceding year. For convenience, we report results for quartiles of life events. (In the multiple regression analyses that follow, we used direct counts of events.)

Table 2. 
Image not available
Prevalence of Depression by Risk Factors

Nearly 10% of girls and 3% of boys in the sample reported a history of sexual abuse. A history of physical abuse was reported by 13% of girls and 9% of boys in the sample. For both girls and boys, a history of sexual or physical abuse was strongly related to depressive symptoms. Indeed, almost one fourth of girls with a history of either type of abuse met criteria for depression. The survey asked whether violence at home had ever been so serious that an adolescent contemplated leaving home. About one fourth of girls and boys reported this high level of violence. Again, a history of violence was strongly and significantly related to depression in both girls and boys. The relationship between life events and depression was positive in both girls and boys. Consistent with previous studies,1622 girls and boys who had experienced severe life stresses in the year before the interview were much more likely to meet criteria for depression than were those whose lives had been less stressful.


We next examined the correlations between depressive symptoms and problem indicators. We examined 2 measures of school performance: days of school missed because of illness and whether a child was in the expected grade level (age minus 6). We examined the 4 measures of problem behaviors included in the CFSAG survey: alcohol use, drug use, smoking, and bingeing. Finally, we examined whether a girl or boy reported suicidal ideation. Table 3 reports rates of these problem indicators and behaviors among depressed and nondepressed adolescent girls and boys.

Table 3. 
Image not available
Problem Outcomes and Depression

Depression was correlated with a significant increase in the number of school days missed. Depressed adolescent girls were also almost twice as likely to be behind a grade in school as those who were not depressed. Both girls and boys who were depressed reported much higher rates of use of alcohol, drugs, smoking, and bingeing. Indeed, more than 65% of depressed girls and boys engaged in at least 1 of these risk behaviors. Finally, suicidal ideation was substantially more frequent among depressed adolescents than among those who were not depressed.

As we noted already, depression is correlated with environmental risk factors that occur disproportionately in families that are also socioeconomically disadvantaged.41 Risk behaviors that are associated with depression may, instead, be a consequence of socioeconomic disadvantage or of environmental risk factors.

We next examined these outcomes in multivariate analyses that also control for sociodemographic characteristics and risk factors. The odds ratios and regression coefficients for depression, after adjusting for sociodemographic characteristics and risk factors, are reported in Table 4. We found that, for most problem indicators, high levels of depressive symptoms were positive predictors of problems, even after controlling for a wide range of socioeconomic variables and environmental risk factors. As expected, many of these risk factors were also independent predictors of problem indicators.

Table 4. 
Image not available
Problem Indicators and Depression (Multivariate Analysis)*

In the multivariate analysis, depressive symptoms were most strongly associated with smoking, bingeing, and suicidal ideation (P<.05). Adolescents with high levels of depressive symptoms missed, on average, nearly 1 day more of school in the preceding month than similar counterparts.


While depression may raise the risk of any or all of these negative outcomes, the outcomes themselves may place adolescents at risk of depression. For example, girls and boys who are doing poorly at school may become depressed in consequence. Alternatively, these negative outcomes, and depression, may be a consequence of other underlying problems. For example, those who use alcohol or drugs or engage in self-destructive behavior may also be depressed. Finally, suicidal thoughts are a marker of depression (so that the relationship between depression and suicidal thoughts cannot be separated).

While we cannot determine causality in these data, we can use information in the CFSAG about whether adolescents drink alcohol, use drugs, smoke, or eat when they are stressed. Depressed girls were more likely than nondepressed girls to report that they ate when stressed. Both depressed girls and depressed boys were more likely than their nondepressed counterparts to report that they stopped eating when stressed, that they drank alcohol when stressed, that they smoked when stressed, and that they used drugs when stressed. These results suggest that depression does, indeed, raise the risk of engaging in high-risk behaviors.

The present study draws on observational cross-sectional data. Thus, the correlations observed herein are not necessarily causal. Adolescents may become depressed because they are performing poorly in school or are using drugs. We cannot exclude the possible effects of reverse causality on our results. The CFSAG relies on adolescent self-report of mental health symptoms, risk behavior, and family characteristics. The validity of these self-reports has not been established.


The CFSAG data suggest that there are important consequences of depression in adolescents. The data strongly suggest that adolescents who have been subject to traumatic life events and to abuse (sexual or physical) are at significantly higher risk of depression. These results hold for both girls and boys. Depressed adolescents are at much higher risk of poor performance at school, of using drugs and alcohol, and of bingeing. Together, these findings suggest that depression is an especially serious problem among children who live in risky environments and that depression is, in turn, associated with other serious risks.

The results of this study show that school attendance, smoking, bingeing, and suicidal ideation are significantly correlated with depression. Information about these indicators and behaviors as well as the presence of traumatic life events could be powerful tools for physicians in the difficult task of identifying adolescent depression and initiating treatment. Overall, studies show that about 1 in 20 adolescents currently suffers from depression, suggesting that routine screening for depression has considerable merit. In this study, among adolescents who missed more than 10 days of school in the preceding month, smoked, engaged in bingeing, or had suicidal thoughts, rates of elevated depressive symptoms were more than twice as high. Thorough screening for depression in this group is critical.

Back to top
Article Information

Accepted for publication May 29, 2002.

This project was supported by the Commonwealth Fund, New York, NY.

We thank Kathrine Jack for research assistance.

Reprints are not available from the authors.

What This Study Adds

Previous research has documented the high prevalence of depression among adolescents. Several researchers have also argued that depression in adolescence is correlated with serious sequelae. Assessing the relationship between depression and adverse outcomes is complicated by the fact that depression is also correlated with other factors that raise the risk of adverse outcomes, such as life events and abuse.

This study used data collected on a large random sample of adolescents to assess the relationship between depressive symptoms and adverse outcomes. The data also contained measures of family background, abuse, and life events, so we were able to control for these factors. We found evidence that, after controlling for these factors, depressive symptoms are correlated with missing school, smoking, bingeing, and suicidal ideation.

Rice  DPKelman  SMiller  LSDunmeyer  S The Economic Costs of Alcohol and Drug Abuse and Mental Illness: 1985: Report Submitted to the Office of Financing and Coverage Policy of the Alcohol, Drug Abuse, and Mental Health Administration, US Department of Health and Human Services.  San Francisco Institute for Health and Aging, University of California1990;
Jaskulski  T Child Mental Health.  Washington, DC Intergovernmental Health Policy Project, George Washington University1993;
Greenberg  PEStiglin  LEFinkelstein  SNBerndt  ER The economic burden of depression in 1990. J Clin Psychiatry. 1993;54405- 418
Ryan  NDPuig-Antich  JAmbrosini  P  et al.  The clinical picture of major depression in children and adolescents. Arch Gen Psychiatry. 1987;44854- 861Article
Puig-Antich  JKaufman  JRyan  ND  et al.  The psychosocial functioning and family environment of depressed adolescents. J Am Acad Child Adolesc Psychiatry. 1993;32244- 253Article
Koyanagi  CGaines  C All Systems Failure.  Washington, DC National Mental Health Association and the Federation of Families for Children's Mental Health1993;
Lewis  DOShanok  SS Medical histories of delinquent and nondelinquent children: an epidemiological study. Am J Psychiatry. 1977;1341020- 1025
Cohen  PPine  DSMust  AKasen  SBrook  J Prospective associations between somatic illness and mental illness from childhood to adulthood. Am J Epidemiol. 1998;147232- 239Article
Kann  LKinchen  SAWilliams  BI  et al.  Youth risk behavior surveillance—United States, 1999. MMWR Morb Mortal Wkly Rep. 2000;49 ((SS-5)) 1- 32
Shaffer  DGould  MSFisher  P  et al.  Psychiatric diagnosis in child and adolescent suicide. Arch Gen Psychiatry. 1996;53339- 348Article
Gould  MSKing  RGreenwald  S  et al.  Psychopathology associated with suicidal ideation and attempts among children and adolescents. J Am Acad Child Adolesc Psychiatry. 1998;37915- 923Article
Rende  RWeissman  MRutter  MWickramaratne  PHarrington  RPickles  A Psychiatric disorders in the relatives of depressed probands, II: familial loading for comorbid non-depressive disorders based upon proband age of onset. J Affect Disord. 1997;4223- 28Article
Fergusson  DMHorwood  LJLynskey  MT Childhood sexual abuse and psychiatric disorder in young adulthood, II: psychiatric outcomes of childhood sexual abuse. J Am Acad Child Adolesc Psychiatry. 1996;351365- 1374Article
Goenjian  AKPynoos  RSSteinberg  AM  et al.  Psychiatric comorbidity in children after the 1988 earthquake in Armenia. J Am Acad Child Adolesc Psychiatry. 1995;341174- 1184Article
Zima  BTWells  KBBenjamin  BDuan  N Mental health problems among homeless mothers: relationship to service use and child mental health problems. Arch Gen Psychiatry. 1996;53332- 338Article
Goodyer  IM Recent undesirable life events: their influence on subsequent psychopathology. Eur Child Adolesc Psychiatry. 1996;5(suppl 1)33- 37Article
Goodyer  IMHerbert  JTamplin  ASecher  SMPearson  J Short-term outcome of major depression, II: life events, family dysfunction, and friendship difficulties as predictors of persistent disorder. J Am Acad Child Adolesc Psychiatry. 1997;36474- 480Article
Tamplin  AGoodyer  IMHerbert  J Family functioning and parent general health in families of adolescents with major depressive disorder. J Affect Disord. 1998;481- 13Article
Williamson  DEBirmaher  BAnderson  BPal-Shabbout  MRyan  ND Stressful life events in depressed adolescents: the role of dependent events during the depressive episode. J Am Acad Child Adolesc Psychiatry. 1995;34591- 598Article
DuBois  DLFelner  RDMeares  HKrier  M Prospective investigation of the effects of socioeconomic disadvantage, life stress, and social support on early adolescent adjustment. J Abnorm Psychol. 1994;103511- 522Article
Ge  XLorenz  FOConger  RDElder Jr  GHSimons  RL Trajectories of stressful life events and depressive symptoms during adolescence. Dev Psychopathol. 1994;30467- 483Article
Monck  EGraham  PRichman  NDobbs  R Adolescent girls, II: background factors in anxiety and depressive states. Br J Psychiatry. 1994;165770- 780Article
Schoen  CDavis  KCollins  KLGDes Roches  CAbrams  M The Commonwealth Fund Survey of the Health of Adolescent Girls.  New York, NY Commonwealth Fund1997;
Schraedley  PKGotlib  IHHayward  C Gender differences in correlates of depressive symptoms in adolescents. J Adolesc Health. 1999;2598- 108Article
Hayward  CGotlib  IHSchraedley  PKLitt  IF Ethnic differences in the association between pubertal status and symptoms of depression in adolescent girls. J Adolesc Health. 1999;25143- 149Article
Klein  JDWilson  KMMcNulty  MKapphahn  CCollins  KS Access to medical care for adolescents: results from the 1997 Commonwealth Fund Survey of the Health of Adolescent Girls. J Adolesc Health. 1999;25120- 130Article
Kapphahn  CJWilson  KMKlein  JD Adolescent girls' and boys' preferences for provider gender and confidentiality in their health care. J Adolesc Health. 1999;25131- 142Article
Simantov  ESchoen  CKlein  J Health-compromising behaviors: why do adolescents smoke or drink? identifying underlying risk and protective factors. Arch Pediatr Adolesc Med. 2000;1541025- 1033Article
Kovacs  M The Children's Depression Inventory (CDI). Psychopharmacol Bull. 1985;21995- 998
Carlson  GACantwell  DP Unmasking masked depression in children and adolescents. Am J Psychiatry. 1980;138445- 449
Stata Corp, Stata Statistical Software: Release 7.0.  College Station, Tex Stata Corp2001;
Anderson  JCWilliams  SMMcGee  ROSilva  PA DSM-III disorders in preadolescent children: prevalence in a large sample from the general population. Arch Gen Psychiatry. 1987;4469- 76Article
Bird  HRCanino  GRubia-Stipec  M  et al.  Estimates of the prevalence of childhood maladjustment in a community survey in Puerto Rico: the use of combined measures. Arch Gen Psychiatry. 1988;451120- 1126Article
Cohen  PCohen  JBrook  J An epidemiological study of disorders in late childhood and adolescence, II: persistence of disorders. J Child Psychol Psychiatry. 1993;34869- 877Article
McGee  RFeehan  MWilliams  SAnderson  J DSM-III disorders from age 11 to age 15 years. J Am Acad Child Adolesc Psychiatry. 1992;3150- 59Article
Shaffer  DFisher  PDulcan  MK  et al.  The NIMH Diagnostic Interview Schedule for Children Version 2.3 (DISC-2.3): description, acceptability, prevalence rates, and performance in the MECA study. J Am Acad Child Adolesc Psychiatry. 1996;35865- 877Article
Harrington  R Affective disorders. Rutter  MedTaylor  EedHersov  LedChild and Adolescent Psychiatry: Modern Approaches London, England Blackwell Scientific Publications1995;330- 350
Bolgar  RZweig-Frank  HParis  J Childhood antecedents of interpersonal problems in young adult children of divorce. J Am Acad Child Adolesc Psychiatry. 1995;34143- 150Article
Wallerstein  JS The long-term effects of divorce on children: a review. J Am Acad Child Adolesc Psychiatry. 1991;30349- 360Article
Fergusson  DMHorwood  LJLynskey  MT Parental separation, adolescent psychopathology, and problem behaviors. J Am Acad Child Adolesc Psychiatry. 1994;331122- 1131Article
Johnson  JGCohen  PDohrenwend  BPLink  BGBrook  JS A longitudinal investigation of social causation and social selection processes involved in the association between socioeconomic status and psychiatric disorders. J Abnorm Psychol. 1999;108490- 499Article