Impact of androgen deprivation therapy on depressive symptoms in men with nonmetastatic prostate cancer


  • Narhari Timilshina MPH,

    1. Department of Medicine, University Health Network, Toronto, Ontario, Canada
    Search for more papers by this author
  • Henriette Breunis CRA,

    1. Department of Medicine, University Health Network, Toronto, Ontario, Canada
    Search for more papers by this author
  • Shabbir Alibhai MD, MSc

    Corresponding author
    1. Department of Medicine, University Health Network, Toronto, Ontario, Canada
    2. Geriatric Program, Toronto Rehabilitation Institute, Toronto, Ontario, Canada
    3. Department of Medicine, University of Toronto, Toronto, Ontario, Canada
    4. Department of Health Policy, Management and Evaluation, University of Toronto, Toronto, Ontario, Canada
    • University Health Network, Room EN14-214, 200 Elizabeth Street, Toronto M5G 2C4, Canada
    Search for more papers by this author
    • Fax: (416) 595-5826



Up to 50% of prostate cancer (PC) patients receive androgen deprivation therapy (ADT), often for several years. Although depression has been reported after a diagnosis of PC, whether ADT leads to or worsens depression is not clear.


Three groups were assembled: ADT users (men initiating continuous ADT), PC controls (PC patients who were not on ADT), and healthy controls. All 3 cohorts were matched on age, education, and physical function, and none had metastases. Depression was measured at study entry and again at 3, 6, and 12 months using the 15-item Geriatric Depression Scale (GDS). Our primary outcomes were worsening depressive symptoms and incident depression (defined as a GDS score ≥5), analyzed using adjusted linear regression and logistic regression, respectively.


Of the 257 participants (mean age, 69.1 years), baseline characteristics including GDS score and prior depression were similar across cohorts. In adjusted analyses of initially nondepressed patients, ADT use was not a significant predictor of change in GDS score at 3 months (P = .42), 6 months (P = .25), or 12 months (P = 0.19). Among ADT users, 8%-9% of participants developed incident depression compared with 0%-4% among PC controls and 4%-6% among healthy controls over 3-12 months (P>.05 at all time points). In a separate analysis of patients with depression at baseline, there was no effect of ADT on depressive symptoms at 3, 6, or 12 months (P = .11, .74, and .12, respectively).


Twelve months of ADT use were not associated with worsening depressive symptoms among nondepressed or depressed patients with nonmetastatic PC. Cancer 2012;. © 2011 American Cancer Society.


Prostate cancer (PC) is the most common nonskin malignancy, with more than 200,000 cases diagnosed in 2010.1 More than 80% of cases occurred in men >65 years of age. A cancer diagnosis can mean a significant threat to the patient and evoke strong psychological responses. In addition to developing PC, many of these men, by virtue of being middle-aged or elderly, are experiencing other significant stresses, such as the death of a spouse, retirement, loss of health due to comorbid illness, and so forth.2

Depression often follows a diagnosis of cancer. Pirl3 suggested that depression is seen in approximately 10%-25% of patients with various kinds of cancer. Similarly, Roth et al4 found that the prevalence of depression was 15.2% among men with PC. Various other studies have suggested that depression can lead to decreased medical adherence as well as increased hospital stay, morbidity, and mortality.5-7 Therefore, it is important to understand psychiatric comorbidities in patients with cancer.

Based in large part on clinical trial evidence for improved outcomes, approximately 50% of PC patients will receive androgen deprivation therapy (ADT) at some point after diagnosis; most will take it for at least 2-3 years.8-10 The most common adverse effects of ADT include fatigue, hot flashes, loss of sexual interest, anemia, decreased muscle mass, and osteoporosis.9, 11, 12 ADT may also affect the brain, given the widespread distribution of both estrogen and testosterone receptors in the brain.13-15 Minor cognitive changes have been described after 3-12 months of treatment.14, 16, 17

Many men who are on ADT also describe emotional lability and related psychological changes, including depressed mood.18 Given that older patients who have cancer and are on ADT are vulnerable to depression, and given the negative impact of depression on quality of life, medical adherence, and morbidity, this relationship is important to clarify. However, few studies have examined the impact of ADT on depression. These studies have generally included men with advanced disease, featured small sample sizes, lacked appropriate controls, and had varying periods of follow-up (range, 0-12 months), limiting the interpretation of findings.19-22 Therefore, the impact of ADT on depression remains unclear, particularly in men with nonmetastatic disease, which is presently the most common indication for ADT.

We examined both depressive symptoms and incident depression outcome as prespecified secondary outcomes in a prospective matched cohort study of men with PC evaluated from the time of initiation of ADT along with 2 control groups. Based in part on previous publications reviewed above,19, 20 we hypothesized that ADT use would be associated with worsening depression in men with nonmetastatic PC.


This study is a secondary analysis from a prospective cohort study. Our primary outcomes were physical function, quality of life, and cognitive function, which have been published recently.12, 16

Three separate cohorts were enrolled from PC and urology clinics at the Princess Margaret Hospital, a major academic cancer center in Toronto, Canada. ADT users consist of men who were initiating continuous ADT for at least 1 year. Two controls groups were also enrolled, consisting of men with PC who were not on ADT (PC controls) and healthy cohorts without PC. All 3 cohorts were matched for age, education, and baseline physical function. All men with PC had nonmetastatic disease.

Patients were excluded if they had another active malignancy, had a life expectancy of less than 1 year, had a major neuropsychiatric abnormality (including active, severe depression, but not previously treated or current nonsevere depression), or were unable to ambulate without assistance. All of the patients provided written informed consent. The study was approved by the Institutional Research Ethics Board.

Patients were asked to complete questionnaires at study entry, providing socio-demographic, comorbidity,23 and medication information. PC-specific information (including prostate-specific antigen [PSA] values) were obtained from the patient's medical record. Height and weight were measured using standardized instruments at baseline to compute body mass index. Follow-up questionnaires were administered at 3 months, 6 months, and 12 months. Loss to follow-up was minimal.12

Depression Measurements

Patients in all 3 cohorts completed the 15-item version of the Geriatric Depression Scale (GDS)24 at 4 time points over 12 months. Yesavage et al.24 designed the GDS to be a rapid and reliable screening tool for depression in older adults. The short version of GDS consists of 15 “yes/no” question, with a score of ≥5 suggestive of depression and score of >10 indicating moderate to severe depression. The GDS has been extensively validated in adults aged 65 years or older25, 26 and in middle-aged patients >40 years old.27, 28 Based on a recent meta-analysis, a cut-off of ≥5 on the 15-item GDS has a sensitivity of 81.3% and a specificity of 78.4% when compared with a validated psychiatric diagnosis.25, 29 Although responsiveness of the GDS does not appear to have been formally assessed,30 based on statistical grounds a 1-point change in GDS score would be considered a small to moderate change.31 At the baseline visit, the GDS was administered to measure the level of depressive symptoms. Participants were re-evaluated at 3, 6, and 12 months.

Our primary outcomes were change in depressive symptoms (GDS score) and the development of incident depression (defined as a GDS score ≥5 in a previously nondepressed patient). At each visit, we also asked the patient about any new antidepressant or other psychiatric medication prescriptions or a physician diagnosis of depression.

Statistical Analyses

Descriptive analyses of baseline characteristics were conducted to compare the 3 cohorts. Continuous data were reported as the mean ± SD or median and interquartile range depending on the distribution of the data. Frequencies were reported for categorical data. To compare baseline characteristics across the 3 cohorts, analysis of variance and chi-square tests were used for continuous and categorical variables, respectively.

Patients with prevalent depression (defined as a GDS score ≥5, previous physician diagnosis of depression, or antidepressant medication use) at baseline were excluded from analyses of change in depressive symptoms over time.

The mean GDS score at baseline and change in depression score at 3, 6, and 12 months were evaluated. Multivariable linear regression was used to examine for differences in change in GDS score across cohorts over time, adjusting for age, education, smoking, alcohol use, income, and baseline GDS score. These covariates were selected based on clinical judgment using variables from the literature that were associated with depression in other populations. We also examined whether the proportion of patients developing incident depression were different across cohorts using Fisher's exact tests and pair-wise contrasts between ADT users and healthy controls; this was a prespecified analysis. Given the few events, we did not attempt multivariable logistic regression modeling. In an exploratory analysis, we examined the impact of PSA progression over 12 months on GDS score.

In a subset analysis of patients with prevalent depression, we examined whether ADT led to changes in GDS scores compared with PC controls and healthy controls. However, because few patients had prevalent depression, this analysis was primarily descriptive, and multivariable modeling was not performed.

All data were entered in duplicate in electronic databases. Statistical analyses were performed using the SAS statistical package version 9.1 (SAS institute Inc., Cary, NC).


Demographic Characteristics of Patients

Of the 260 enrolled participants, 257 completed the baseline GDS assessment. Patients ranged from 60-80 years of age (mean, 69.1 years; SD 7.0 years). The majority of patients were white (87.3%) and had a mean of 15.3 years of formal education. Most baseline characteristics were similar across cohorts at baseline (Table 1). At baseline, ADT users had slightly higher GDS scores than PC controls or healthy controls (mean, 1.75, 1.00, and 1.49, respectively; P = .062) (Table 2). PSA levels were higher in ADT users (P = .001) compared with PC controls, but comorbidity (as measured by the Charlson score23) was not different among the 3 cohorts (P = .41).

Table 1. Baseline Characteristics of Patients
VariableADT Users (n = 85)PC Controls (n = 86)Healthy Controls (n = 86)Total (n = 257)P
  • ADT, androgen deprivation therapy; BMI, body mass index; IQR, interquartile range; NA, not available; PC, prostate cancer; PSA, prostate-specific antigen; RP, radical prostatectomy; RT, radiation therapy; SD, standard deviation.

  • a, b

    Analysis of variance.

  • b

    Chi-square test.

  • c

    None or planned RP/RT (but not yet occurred at baseline).

Mean age, y (SD)69.7 (6.9)69.8 (6.7)67.8 (7.3)69.1 (7.0).13a
Mean education, y (SD)14.6 (4.1)15.6 (3.9)15.9 (3.4)15.4 (3.8).08a
Annual income, CAD$ (%)     
 <20,0009 (10.6)8 (9.3)14 (16.3)31 (12.1).18b
 20,000-40,00027 (31.8)14 (16.3)21 (24.4)62 (24.1) 
 40,000-60,00012 (14.1)24 (27.9)17 (19.8)53 (20.6) 
 60,000-100,00020 (23.5)21 (24.4)21 (24.4)62 (24.1) 
 >100,00017 (20.0)19 (22.1)13 (15.1)49 (19.1) 
Race, no. (%)     
 White72 (84.7)77 (89.5)76 (88.4)225 (87.5).61b
 Other13 (15.3)9 (10.5)10 (11.6)32 (12.5) 
Current smoker, no. (%)60 (70.5)47 (54.6)56 (65.8)163 (63.7).08b
Current regular alcohol use, no. (%)34 (40.5)45 (54.7)39 (45.3)118 (46.6).19b
Currently working, no. (%)28 (32.9)29 (34.1)30 (34.8)87 (33.9).96b
Living alone, no. (%)20 (23.5)9 (10.5)22 (25.6)51 (19.8).026b
Gleason score, median (IQR)7 (7-9)7 (6-7)NA7 (6-8).001a
Clinical stage, no. (%)     
 T1c23 (27.1)40 (46.5)NA63 (36.9).0002b
 T2a19 (22.4)27 (31.4)NA46 (26.9) 
 T2b18 (21.2)15 (17.4)NA33 (19.3) 
 T2c7 (8.2)2 (2.3)NA9 (5.3) 
 T318 (21.2)2 (2.3)NA20 (11.7) 
PSA level, median (IQR)10.3 (5.6 -20.6)0.83 (0.05-4.4)NA2.81 (0.12-8.3).001a
Type of previous treatment, no. (%)     
 RT56 (65.8)42 (48.8)NA98 (57.3).015b
 RP12 (14.1)29 (33.7)NA41 (23.9) 
 RP+RT9 (10.6)5 (5.8)NA14 (8.2) 
 Nonec8 (9.4)10 (11.6)NA18 (10.5) 
Median Charlson score (IQR)0 (0-1)0 (0-1)0 (0-1)0 (0-1).41a
Mean BMI (SD)27.4 (4.6)27.3 (3.6)27.6 (3.9)27.8 (4.0).88a
Depression at baseline, no. (%)8 (9.4)3 (3.5)7 (8.1)18 (7.0).24b
Other psychiatric illness, no. (%)4 (4.7)5 (5.8)4 (4.7)13 (5.1).93b
Table 2. Depression Scores Over Time and by Cohort
VariableADT Users (n = 85)PC Controls (n = 86)Healthy Controls (n = 86)Total (n = 257)P
  • ADT, androgen deprivation therapy; GDS, Geriatric Depression Scale, 15-item (higher scores = more depression); PC, prostate cancer; SD, standard deviation.

  • All values are presented as the mean (SD).

  • a

    From adjusted regression model.

Baseline GDS score1.75 (2.1)1.00 (1.7)1.49 (2.4)1.41 (2.1).062a
Change over 3 months0.38 (2.5)0.0 (0.89)0.19 (1.5)0.18 (1.7).42a
Change over 6 months0.52 (2.3)0.04 (0.77)0.21 (1.9)0.25 (1.8).25a
Change over 12 months0.59 (2.0)0.13 (1.2)0.20 (1.5)0.30 (1.6).19a

Eighteen men (7%) met the criteria for prevalent depression at baseline and were thus excluded from the main analyses. There were similar numbers of men with prevalent depression in each cohort (P = .24) (Table 1).

Impact of ADT Use in Patients Who Were Not Depressed at Baseline

On univariate linear regression, ADT use was not a significant predictor of change in GDS score over 3 months (P = .42), 6 months (P = .25), and 12 months (P = .19) (Table 2) among patients who did not have prevalent depression. The proportion of patients developing incident depression among ADT users was approximately 9% compared with 4% in PC controls and 6% in healthy controls over 12 months (Table 3). Pairwise contrasts between ADT users and healthy controls showed no difference in the incidence of depression at either 3 (P = .52), 6 (P = .32), or 12 months (P = .53). Four men started an antidepressant medication between the baseline and 12-month visits (2 ADT users, 1 PC control, and 1 healthy control). PSA progression, analyzed as a continuous or categorical variable, was not associated with change in GDS score over time (data not shown).

Table 3. Development of Incident Depression Over 12 Months (After Excluding Prevalent Depression at Baseline)
Time PointIncident Depression (GDS Score ≥5)Pa
ADT UsersPC ControlsHealthy CohortsTotal
  • ADT, androgen deprivation therapy; GDS, Geriatric Depression Scale; PC, prostate cancer.

  • All values are presented as no. (%).

  • a

    Based on pairwise contrasts between ADT users and healthy controls.

3 mo    .52
 Yes6 (8.8)0 (0.0)4 (5.4)10 (4.6) 
 No62 (91.2)76 (100.0)70 (94.6)208 (95.4) 
6 mo    .32
 Yes6 (8.7)0 (0.0)3 (4.2)9 (4.1) 
 No63 (91.3)76 (100.0)70 (95.9)209 (95.9) 
12 mo    .53
 Yes6 (9.1)3 (4.2)4 (5.8)13 (6.3) 
 No60 (90.9)69 (95.8)65 (94.2)194 (93.7) 

Impact of ADT Use in Patients With Prevalent Depression

Among patients with prevalent depression, ADT use did not appear to be associated with any impact in the change in GDS score from baseline to 3 months, 6 months, or 12 months compared with PC controls and healthy cohorts (P = .11, .74, and .12, respectively) (Table 4). PC patients, both ADT users and PC controls, demonstrated small declines in GDS scores over time, whereas healthy cohorts had relatively stable GDS scores (Table 4).

Table 4. GDS Scores at Baseline and Over Time in Patients With Prevalent Depression
Time PointADT Users (n = 8)PC Controls (n = 3)Healthy Cohorts (n = 7)Total (n = 18)Pa
  • a

    Based on medical chart diagnosis, chronic use of antidepressant medications, or geriatric depression scale score >5 at baseline.

  • b

    Chi-square test.

Depression at baseline (%)a8 (9.4)3 (3.5)7 (8.1)18 (7.0).24b
GDS score at baseline, mean (SD)6.6 (1.9)8.3 (2.1)8.6 (1.5)7.7 (1.9).13
Mean change GDS score at 3 months−2.9 (2.7)−3.7 (2.5)0.7 (4.0)−1.6 (3.6).11
Mean change GDS score at 6 months−1.7 (2.1)−0.6 (4.7)0.0 (5.2)−0.8 (3.8).74
Mean change GDS score at 12 months−2.0 (2.7)−3.3 (2.5)0.7 (3.7)−1.2 (3.1).12


We examined the changes over 12 months in depressive symptoms using a validated self-reported questionnaire in ADT users, PC controls, and healthy controls. This prospective study demonstrated that ADT does not appear to be associated with an increased risk of depression over 12 months compared with controls in men who were not depressed at baseline. Additionally, among patients with prevalent depression at baseline, ADT use did not appear to be associated with a significant worsening of depressive symptoms.

Previous studies reporting on the risk of depression in patients with PC receiving ADT are primarily restricted to case series32 or uncontrolled studies with small samples. For example, in an early report, Higano et al33 studied 22 men on intermittent ADT. In this study, depressive symptoms were reported by patients or their spouses, and formal quality of life or depression questionnaires were not used. In 2 studies, Pirl et al examined depressive symptoms in up to 52 ADT users.19, 20 In 1 study, 26.3% of subjects had metastatic disease.20 None of the above studies had a control group of either men with PC who were not using ADT or healthy controls.19, 20, 33 A large population-based study by Shahinian et al34 assessed the risk of depression in a large sample of men with and without PC. In their study, men with or without PC developed depression in similar proportions (9.6% and 9.5%, respectively), but depression was more frequent among ADT users than nonusers with PC (12.1%; P<.001). However, the excess risk of depression was substantially attenuated after adjusting for comorbidity, tumor characteristics, and age (relative risk, 1.08; 95% confidence interval, 1.02-1.15),34 and depression was diagnosed based on unvalidated diagnostic codes.

Our rates of incident and prevalent depression are slightly lower than those reported by Pirl et al20 and Roth et al,4 although many of their patients had advanced PC and greater comorbidity and were older than our study participants, which may explain the differences.

Although our findings appear somewhat discrepant with those of other authors, as mentioned earlier, most previous studies lacked control groups. Several authors have noted that conditions such as depression may be due to the cancer itself or to other medical comorbid conditions.35, 36 Hence there is a need to include controls, ideally both with and without PC. Although Shahinian et al34 found a slightly higher rate of depression among ADT users compared with controls, the relationship was attenuated after adjustment for age, stage, and comorbidity. Additionally, the diagnosis of depression was based on administrative data coding rather than more validated methods such as screening questionnaires or structured psychiatric interviews. These differences likely account for the slightly different results observed between studies.

To our knowledge, no previous prospective study has examined the impact of ADT on depressive symptoms in men with prevalent depression. If ADT is indeed associated with worsening depressive symptoms and/or incident depression, one would expect that men with prevalent depression might be at even greater risk of worsening depressive symptoms. This was not observed in our study, suggesting that our finding of a lack of effect of ADT on depressive symptoms in men without prevalent depression is likely real. However, there are several limitations that must be considered. Our study featured a relatively small cohort of patients, which might lead to us missing a small increased risk of depression. Equally important, we examined depressive symptoms rather than clinical depression. Although the GDS has very good sensitivity and specificity, ideally any patients who are screen-positive (ie, GDS scores ≥5) should be evaluated by a psychiatrist to exclude other factors such as fatigue and effects of physical illness. Finally, our cohort included relatively few men with prevalent depression and excluded men with severe active depression from study participation. Thus, our observations on the effects of ADT in men with prevalent depression must be interpreted cautiously and warrant confirmation in other studies.

In conclusion, our study suggests that 12 months of ADT use is not associated with worsening depressive symptoms or greater incident depression in men with nonmetastatic PC. These findings should be reassuring to primary care physicians and PC clinicians who manage the adverse affects of ADT during long-term follow-up of men with PC. Conversely, in men with significant depressive symptoms who are on ADT, other causes should be considered, and these men may benefit from formal psychiatric evaluation.


Funded by grants from the Canadian Cancer Society.


The authors made no disclosures.