aCompanion articles represent reports of published analyses involving the same study population.
Hazard ratios (HRs) were either taken directly from the individual studies or were calculated based on published data. The analyses, using the inverse variance method, were performed with Review Manager software, version 5.2 (Cochrane Collaboration) using random effect models. CAD indicates continuous androgen deprivation; IAD, intermittent androgen deprivation.
eAppendix. Medline search strategy
eFigure 1. Interpretation of the 95% confidence interval
eTable 1. Sensitivity analysis for overall survival
eTable 2. Sensitivity analysis for cancer-specific survival
eTable 3. Sensitivity analysis for progression-free survival
eTable 4. Sensitivity analysis for time to progression
eTable 5. Side effects reported by studies
eTable 6. Pooled estimates of side effects
Magnan S, Zarychanski R, Pilote L, Bernier L, Shemilt M, Vigneault E, Fradet V, Turgeon AF. Intermittent vs Continuous Androgen Deprivation Therapy for Prostate CancerA Systematic Review and Meta-analysis. JAMA Oncol. 2015;1(9):1261–1269. doi:10.1001/jamaoncol.2015.2895
Copyright 2015 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
Androgen deprivation is the standard therapy for patients with advanced or recurrent prostate cancer. However, this treatment causes adverse effects, alters quality of life, and may lead to castration-resistant disease. Intermittent androgen deprivation has been studied as an alternative.
To conduct a systematic review and meta-analysis comparing the efficacy and tolerability of intermittent vs continuous androgen deprivation therapy in patients with prostate cancer.
We searched Cochrane CENTRAL, Medline, Embase, Web of Science, Biosis, National Technical Information Service, OpenSIGLE, and Google Scholar from inception of each database through March 2014. References from published guidelines, reviews, and other relevant articles were also considered.
We selected randomized clinical trials comparing intermittent vs continuous androgen deprivation therapy in patients with prostate cancer.
Data Extraction and Synthesis
Two reviewers performed study selection, data abstraction, and risk of bias assessment. We calculated hazard ratios (HRs) with the inverse variance method and risk ratios with the Mantel-Haenszel method, using random effect models. A noninferiority analysis was conducted for overall survival with a margin of 1.15 for the upper boundary of the HR. We assessed heterogeneity using the I2 index.
Main Outcomes and Measures
Primary outcomes were overall survival and quality of life. Secondary outcomes were cancer-specific survival, progression-free survival, time to castration resistance, skeletal-related events, and adverse effects.
From 10 510 references, we included 22 articles from 15 trials (6856 patients) published between 2000 and 2013. All but 1 study had an unclear or high risk of bias. We observed no significant difference between intermittent and continuous therapy for overall survival (HR, 1.02; 95% CI, 0.93-1.11; 8 trials, 5352 patients), cancer-specific survival (HR, 1.02; 95% CI, 0.87-1.19; 5 trials, 3613 patients), and progression-free survival (HR, 0.94; 95% CI, 0.84-1.05; 4 trials, 1774 patients). There was minimal difference in patients’ self-reported quality of life between the 2 interventions. Most trials observed an improvement in physical and sexual functioning with intermittent therapy.
Conclusions and Relevance
Intermittent androgen deprivation was not inferior to continuous therapy with respect to the overall survival. Some quality-of-life criteria seemed improved with intermittent therapy. Intermittent androgen deprivation can be considered as an alternative option in patients with recurrent or metastatic prostate cancer.
Prostate cancer is the second most common cancer and the fifth leading cause of death from cancer in men worldwide, with an estimated 1 112 000 new diagnoses and 307 000 deaths in 2012.1 For patients with advanced or recurrent prostate cancer, androgen deprivation is the standard initial therapeutic approach,2 but it carries many life-altering adverse effects including fatigue, sexual dysfunction, and cognitive changes.3- 6 Despite initial response rates of 80% to 90%, almost all men eventually develop castration-resistant disease. At this stage, although multiple new agents have become available in the past few years, survival remains limited.7
To overcome castration resistance and prolong the duration of response, androgen deprivation administered intermittently has been investigated as an alternative to continuous therapy. Studies performed on tumor models have observed that alternation between testosterone deprivation and replacement therapy induced multiple apoptotic regressions and a longer time to castration resistance.8,9 In addition, it may allow an increase in testosterone levels between cycles, reducing adverse effects of medical castration and improving quality of life.
Although intermittent androgen deprivation appears to be an attractive option, there is a paucity of evidence for its safety and efficacy,10 and so its use remains controversial, and recommendations in current guidelines are variable.2,11,12 Given an increase in the number of clinical trials, including 3 large randomized clinical trials,13- 15 and the growing use of intermittent androgen deprivation therapy in clinical practice, assessment of this treatment option has grown in importance.
We conducted a systematic review and meta-analysis of randomized clinical trials to compare the efficacy and tolerability of intermittent vs continuous androgen deprivation in the treatment of patients with prostate cancer. We hypothesized that intermittent androgen deprivation therapy is not inferior to the continuous regimen in terms of overall survival.
The goal of this systematic review and meta-analysis was to compare the efficacy and tolerability of intermittent vs continuous androgen deprivation in the treatment of patients with prostate cancer.
Intermittent androgen deprivation is not inferior to continuous therapy with respect to overall survival. No major difference in quality of life was observed between groups, but some domains seemed improved with intermittent therapy.
High risk of bias in trials, unclear optimal duration of on- and off-treatment periods, and unknown magnitude of effect according to disease stage mandate further research on intermittent therapy before considering it the standard of care.
Using an a priori defined protocol, we conducted our systematic review following the Methodological Expectations of Cochrane Intervention Reviews and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses criteria.16,17 Experts from radiation oncology, urology, hematology, and clinical epidemiology, including investigators with extensive experience in conducting systematic reviews, designed the protocol and conducted the review.
We designed a search strategy to identify randomized clinical trials comparing intermittent with continuous androgen deprivation therapy for prostate cancer from the inception of all searched databases through March 2014 (eAppendix in the Supplement). We carried out an electronic search of Cochrane CENTRAL, Medline, Embase, and Web of Science. We also searched for abstracts and conference proceedings in Biosis Previews. We searched for ongoing trials at clinicaltrials.gov and controlled-trials.com. Searched gray-literature sources included the National Technical Information Service (NTIS), OpenSIGLE, and Google Scholar. Validated filters for randomized clinical trials were used for Embase18 and Medline.19 References from published guidelines and from narrative and systematic reviews identified in the systematic search were also considered. Two of us (S.M. and L.B.) independently screened all trials to determine eligibility first by titles and abstracts, then by full texts. Disagreements were resolved through consensus or consultation with a third reviewer (A.F.T.).
We considered all randomized clinical trials that compared intermittent vs continuous androgen deprivation therapy (any regimen) for patients with prostate cancer (any stage). Our primary outcomes were overall survival and quality of life. Secondary outcomes were cancer-specific survival, progression-free survival (biochemical and/or clinical), time to castration resistance, skeletal-related events, and drug-related adverse effects. We considered any adverse effects reported in trials such as fatigue, vasomotor symptoms, emotional and cognitive changes, decreased libido, sexual dysfunction, thromboembolic and cardiovascular events, osteoporosis, diabetes, gynecomastia, and weight gain. Castration-resistant disease was defined as any type of progression (biochemical and/or clinical) while the patient was undergoing androgen deprivation therapy and his testosterone level was at castration levels. Duration of off-treatment intervals and testosterone levels were also considered as additional outcomes for the intermittent-therapy group. There was no restriction for language and type of publication.
To standardize the data extraction process, we developed a data collection form that was pilot tested with 3 trials. The form included items on study eligibility and design, participants’ characteristics, interventions, outcomes, and risk of bias. Two of us (S.M. and L.P.) independently extracted the data from the selected trials. In instances of disagreement, a third reviewer was consulted (A.F.T.). A translator was consulted for trials in languages other than English.
The risk of bias was assessed with the Cochrane Collaboration’s Risk of Bias tool.16 Selection, performance, detection, attrition, and reporting bias, as well as their relevant domains, were assessed separately according to primary outcomes. When neither overall survival nor quality of life was the trial’s primary outcome, we also assessed the same 4 risk-of-bias categories for the primary outcome of the trial. In our assessment, we considered that overall survival was unlikely to be influenced by lack of blinding. A judgment of low risk was made for this outcome for all trials in terms of performance and detection bias. Assessment of the risk of bias was performed independently by 2 of us (S.M. and M.S.). A third reviewer was consulted in case of disagreements (A.F.T.).
Analyses were performed with Review Manager software, version 5.2 (Cochrane Collaboration), using random effect models. We reported time-to-event outcomes as hazard ratios (HRs) and dichotomous outcomes as risk ratios with 95% confidence intervals. Hazard ratios were calculated with the inverse variance method, and risk ratios with the Mantel-Haenszel method. Hazard ratios of the individual trials were taken directly from the articles when reported or calculated using validated methods20,21 when sufficient data were available. For our primary outcome of overall survival, we conducted a noninferiority analysis. We considered that an absolute difference of 5% at median survival was clinically relevant. This difference is equivalent to an upper boundary of the HR for death of less than 1.15. The 95% confidence interval was interpreted as suggested by the CONSORT Group22 (eFigure in the Supplement).
Heterogeneity was assessed for each outcome using the I2 statistic.23,24 We considered I2 values of 25% to 49%, 50% to 74%, and 75% or greater to represent low, moderate, and high heterogeneity, respectively.24 We planned subgroup and sensitivity analyses based on (1) disease stage (early stage [T1/2N0M0]; locally advanced without prior androgen deprivation; recurrent [biochemical and/or clinical] after primary radical therapy; metastatic hormone sensitive; and metastatic castration resistant); (2) prior therapy; (3) androgen deprivation regimen (combined androgen blockade, medical or surgical orchiectomy with or without flare-up prophylaxis, antiandrogens alone); (4) induction course (yes vs no); (5) treatment intervals (fixed vs based on prostate-specific antigen [PSA] measure or clinical evaluations); (6) PSA threshold to stop or restart therapy; (7) patients’ age (≤70 years vs >70 years); (8) Gleason score at diagnosis (≤6 vs 7 vs 8-10); and (9) risk of bias of included trials (low vs unclear or high). The presence of publication bias was evaluated through visual examination of funnel plots.25
The overall strength of evidence for the 2 primary outcome measures was evaluated independently by 2 of us (S.M. and M.S.) according to the GRADE system.26 A third reviewer was consulted in case of disagreement (A.F.T.).
We identified 10 495 references from electronic database searches and 15 references from other sources (Figure 1). Eighty-two citations were retained for full-text evaluation. Of these, 31 were preliminary results of other publications; 3 were unavailable through extensive library service searches; and 26 did not meet the inclusion criteria. Finally, 22 articles13- 15,27- 45 based on 15 unique trials (7 companion articles) were included in this systematic review, which represents 6856 patients (median number of patients per trial, 201; range, 31-1749) (Table 1).
Articles from the included trials were published between 2000 and 2013: 20 in English, 1 in Portuguese,30 and 1 in Japanese.32 Of the 15 unique trials, 12 were presented as full publications, and the other 3 in abstract form.35,40,42 The disease stage of the study population was as follows: locally advanced in 1 trial,45 recurrent following a primary radical therapy in 2 trials,13,42 metastatic hormone sensitive in 6 trials,14,30,35,36,40,44 metastatic castration resistant in 1 trial,37 mixed in 4 trials,15,27,29,39 and not specified in the remaining trial.31 Eleven trials included an induction course, durations ranging from 3 to 7 months.14,27- 29,35,36,39,40,42,44,45 The androgen deprivation therapy regimen for the intermittent-therapy group was as follows: combined androgen blockade in 7 trials,14,27,29,31,35,36,40 medical orchiectomy with or without flare-up prophylaxis in 5 trials,13,37,39,42,45 and antiandrogen alone in 3 trials.15,30,44 The duration of treatment periods was fixed in 6 trials ranging from 3 to 8 months,13,14,31,36,39,42 variable in 8 trials,27- 30,37,40,44,45 and not specified in 1 trial.35 Almost all trials had variable off-treatment periods except 1 in which a fixed 6-month interval was used31 and 1 in which the duration was not mentioned.35 The mean age of trial participants was 70 years. All but 3 trials30,35,42 reported median follow-up duration, ranging from 23.2 to 117.6 months.
Risk of bias was assessed for overall survival, quality of life, and the primary outcome of each included trial, when applicable (Table 2). In 9 trials, the primary outcome was neither overall survival nor quality of life but rather recurrence and/or progression of the disease.27,29,30,35,39,40,42,44,45 We considered 14 of the 15 trials to have an unclear or high risk of bias for all the evaluated outcomes. The summary of risk of bias assessment across trials was unclear for overall survival and high for quality of life. Six trials were not blinded, and blinding was not clear in the other 9 trials. Two trials were stopped early after interim analysis: one because of poor accrual37 and the other because the intermittent arm proved to be noninferior with 99% confidence.13 In one trial, 214 patients (12%) were found to be ineligible after randomization and were excluded from analysis.14 In another trial, 21 randomized patients from 2 centers (11%) were excluded from efficacy analysis after a quality audit.36 Four trials had a particularly high incidence of withdrawal from therapy (21%,39 36%,14 44%,40 and 61%44) for causes other than death or progression of the disease. Two trials mentioned no loss to follow-up30,39; 6 trials13- 15,27,29,36 reported losses to follow-up that ranged from 1%36 to 15%27; and in 7 trials, loss to follow-up was unclear.
Eleven trials reported data on overall survival,13- 15,27,30,31,35- 37,39,40 but only 6 provided HRs.13- 15,27,31,39 Two other trials provided enough information to enable calculation of HRs.36,37 We observed no difference between intermittent and continuous therapy based on pooled results of those 8 trials (5352 patients, HR for death, 1.02; 95% CI, 0.93-1.11; I2 = 23%) (Figure 2A). The upper boundary of the 95% CI was less than the prespecified 1.15 limit, supporting our hypothesis that intermittent therapy is not inferior to continuous therapy. The small number of trials limited the robustness of sensitivity and subgroup analyses (see eTable 1 in the Supplement). A subgroup analysis of metastatic castration-resistant prostate cancer vs hormone-sensitive prostate cancer was also performed a posteriori. This analysis did not affect the study results.
Quality of life was assessed by patient self-administered questionnaires in 13 trials,13- 15,27,31,32,35- 38,40,42,44 but the disparity of instruments used and the unavailability of quantitative data prevented conducting a meta-analysis. Different versions of the EORTC (European Organization for Research and Treatment of Cancer) QLQ-C30 questionnaire were used in 9 trials,13,15,27,31,36,37,40,42,44 including a prostate cancer–specific complementary module in 8 of them.13,15,27,31,36,37,40,44 One trial did not mention the instrument used,35 and the remaining 3 trials used other questionnaires.14,32,38 Most of the trials assessed quality of life according to a fixed schedule regardless of treatment intervals. Only 2 trials31,38 evaluated quality of life with respect to treatment and off-treatment periods. Two trials reported a better overall quality of life with intermittent therapy,35,42 and 3 trials did not observe any difference between the 2 methods of administration.36,37,40 The other 7 trials observed an increase in quality of life with intermittent therapy but only in certain domains. The most frequently detected differences related to physical and sexual functioning. However, 3 of those 7 trials also noted an improvement of some quality-of-life criteria in the continuous-therapy group.27,31,44
There was no significant difference between the 2 treatment groups with respect to cancer-specific survival based on pooled results from 5 trials13,15,27,31,39 (3613 patients, HR for death, 1.02; 95% CI, 0.87-1.19; I2 = 14%) (Figure 2B). The results were consistent in all subgroup and sensitivity analyses (see eTable 2 in the Supplement).
Twelve trials reported data on disease progression,13,15,27,29- 31,35,36,39,40,42,44 of which 5 provided HRs.13,15,27,31,39 Hazard ratios from 3 other trials were calculated.35,36,44 Progression of the disease was presented as time to progression in some trials and as progression-free survival in other trials, thus precluding combination of these data. We observed no difference in terms of disease progression between intermittent and continuous androgen deprivation therapy based on pooled results of 5 trials for time to progression (3523 patients, HR for progression, 0.96; 95% CI, 0.76-1.21; I2 = 75%) (Figure 2C) and 4 trials for progression-free survival (1774 patients, HR for progression, 0.94; 95% CI, 0.84-1.05; I2 = 0%) (Figure 2D). Inferences from sensitivity and subgroup analyses are weak considering the small number of trials (eTables 3 and 4 in the Supplement).
Of the 4 trials that evaluated time to castration resistance,13,29,30,42 2 observed a statistically significant difference in favor of intermittent therapy.13,29 The difference was not significant in the 2 remaining trials. Only 1 trial provided an HR,13 and the 3 other trials did not provide enough information to allow HR calculation, thus precluding a pooled analysis.
Twelve trials reported data on drug-related adverse effects (eTable 5 in the Supplement),13- 15,27,29,30,35,36,39,40,42,44 but none evaluated them systematically. All trials presented adverse effects as the number of patients in each group who experienced the adverse event at least once during the whole follow-up period. There was no significant difference between groups for all reported adverse effects (eTable 6 in the Supplement). However, as a whole, pooled point estimates favored reduced drug-related adverse effects with intermittent androgen deprivation.
Only 1 trial reported only 1 skeletal-related event: fracture.39 In this trial, 6.9% vs 5.4% of patients had fracture(s) in the intermittent and the continuous therapy group, respectively. This difference was not significant.
Information was available for off-treatment intervals and testosterone levels for the intermittent therapy group in 1313,15,27,29,30,34- 37,39,40,42- 45 and 913,27,30,31,34,36,37,39,43 trials, respectively. However, most of them reported very few data on those 2 outcomes. Moreover, there was a wide variability in the type of data reported. Therefore, we could not outline any trend between trials except that most of them observed that the duration of off-treatment periods decreased from cycle to cycle.
Visual inspection of a funnel plot of the intervention estimate vs the standard error for trials that provided an HR for overall survival did not reveal evidence of publication bias. According to the GRADE methodology,26 the overall strength of evidence for overall survival was considered to be moderate.
In this systematic review, we did not observe a difference in overall survival for intermittent compared with continuous androgen deprivation therapy for the treatment of patients with prostate cancer. Since the observed upper boundary of the 95% CI of the HR for death was lower than the prespecified margin, our findings support our hypothesis that intermittent therapy is not inferior to continuous therapy. Overall, there were minimal differences in patients’ self-reported quality of life between the 2 interventions. However, an improvement in some quality-of-life criteria was observed with intermittent therapy, mostly in relation with physical and sexual functioning. Finally, the use of intermittent androgen deprivation was not associated with increased time to castration resistance.
Our results for the overall survival analysis are in accordance with those of 2 large recent randomized clinical trials13,15 but in contradiction with 1 other14; all 3 studies were included in our systematic review and meta-analysis. However, besides the disease stage of the study population, important methodological limitations of the contrasting study14 in regard to calculation of the noninferiority margin and sample size, number of participants excluded after randomization, and high incidence of withdrawal from therapy may explain the discrepancies. Two recent meta-analyses also observed no difference in overall survival between intermittent and continuous therapy.46,47 However, these systematic reviews had significant methodological weaknesses, including a limited search strategy, lack of exhaustiveness, questionable methodologies and analyses, and no consideration of recent large trials. Moreover, inadequate methods of HR calculation were used for some of the included trials. Furthermore, time to progression and progression-free survival data were inappropriately combined in analyses, thus seriously limiting the validity of the results. Regarding time to castration resistance, although previous trials performed on tumor models8,9 reported an important increase with alternation of testosterone deprivation and replacement compared with definitive castration, those findings did not seem to translate into a significant clinical benefit as observed in our systematic review.
The fact that most trials used a fixed schedule, regardless of treatment and off-treatment intervals, may explain why no major difference was observed between the 2 treatment regimens in quality-of-life outcomes. Testosterone levels decrease rapidly following the introduction of antiandrogen deprivation therapy and take several months to return to normal levels once the medication administration is stopped. As long as patients undergoing intermittent therapy have low testosterone levels, we can expect that they will experience the same adverse effects as patients undergoing continuous therapy, and that these adverse effects will have a comparable impact on their quality of life. Off-treatment periods in these trials may not have been long enough to allow a sufficient increase in testosterone levels, and this may explain the comparable adverse events and quality of life observed. Finally, the lack of a systematic method of evaluation of adverse effects in the included trials could mainly explain the statistical heterogeneity observed in most pooled estimates and could also partly explain why we observed no significant difference between the 2 interventions. In addition, all trials presented adverse effects as cumulative incidences, which may have contributed to the underestimation of the effect of the treatment schedule since, as noted, we expect that patients with low testosterone levels have similar adverse effects regardless of the intervention received.
Our systematic review and meta-analysis was conducted and reported following established methodological guidelines.16,17 We used an a priori defined protocol and carried out an extensive literature search using multiple databases, including both scientific and gray literature. We analyzed overall survival using a noninferiority design with a prespecified threshold of statistical significance allowing us to draw firm conclusions. We also examined the effect of the intervention on several secondary end points allowing a comprehensive review of the current knowledge on the effect of intermittent androgen deprivation.
Our systematic review was limited by the available data, which were sometimes insufficient to conduct pooled analyses or to include all trials in these analyses. Nevertheless, the pooled estimates of most analyses regrouped a large number of patients (5352 patients for overall survival) showing a potentially high accuracy with narrow CIs. The majority of trials reported quality of life in a descriptive fashion thus precluding pooled analyses. In addition, the high risk of bias and the low methodological quality of the included trials reduced the strength of evidence for our primary outcome measure of overall survival. A high risk of bias was associated with quality-of-life assessment, which is a highly subjective outcome, considering that participants were not blinded to the treatment assigned in more than one-third of the trials and that blinding was not mentioned in the remaining trials. Three trials failed to publish their final results in peer-reviewed journals, which increases concern related to internal validity or systematic bias. Finally, we could not perform all of the planned subgroup and sensitivity analyses owing to limited data availability.
In our systematic review, we observed that intermittent androgen deprivation for the treatment of prostate cancer is not inferior to continuous therapy with respect to overall survival. No major difference in quality of life was observed between groups, although some criteria seemed improved in the intermittent groups in relation with physical and sexual functioning. Intermittent androgen deprivation can be considered as an alternative therapeutic option in patients with prostate cancer. However, the high risk of bias observed in some trials, the unclear optimal approach to the duration of treatment and off-treatment periods and criteria on which it should be based, and the unknown magnitude of effect according to the disease stage warrant further research before it becomes the mandatory standard of care.
Accepted for Publication: June 27, 2015.
Corresponding Author: Sindy Magnan, MD, MSc, FRCPC, Division of Radiation Oncology, Department of Medicine, CHU de Québec, Université Laval, Pavillon Carlton-Auger, 25, rue Charlevoix, Québec, QC G1R 2J6, Canada (firstname.lastname@example.org).
Published Online: September 17, 2015. doi:10.1001/jamaoncol.2015.2895.
Author Contributions: Drs Magnan and Turgeon had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Magnan, Vigneault, Turgeon.
Acquisition, analysis, or interpretation of data: Magnan, Zarychanski, Pilote, Bernier, Shemilt, Fradet, Turgeon.
Drafting of the manuscript: Magnan, Zarychanski, Turgeon.
Critical revision of the manuscript for important intellectual content: Magnan, Zarychanski, Pilote, Bernier, Shemilt, Vigneault, Fradet, Turgeon.
Statistical analysis: Magnan, Zarychanski, Shemilt, Fradet, Turgeon.
Obtained funding: Turgeon.
Administrative, technical, or material support: Shemilt.
Study supervision: Magnan, Vigneault, Turgeon.
Conflict of Interest Disclosures: Dr Vigneault has received honoraria for consultation for Abvie, Sanofie, Jansen, Astellas, Peladin, and Amgen. No other disclosures are reported.
Funding/Support: Dr Magnan is a recipient of a Resident Physician Health Research Career Training Grant from the Fonds de Recherche du Québec–Santé (FRQS). Drs Fradet and Turgeon are recipients of a Clinician-Scientist Award from the FRQS.
Role of the Funder/Sponsor: The supporting institution had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: We are thankful to Caroline Léger, PhD, for editing assistance, and Brice Lionel Batomen Kuimi, MSc, for statistical analysis. Dr Léger and Mr Kuimi are both affiliated with the Population Health and Optimal Health Practices Unit of the CHU de Québec Research Center, Université Laval, Québec City, Québec, Canada. They were not compensated for their contributions beyond their normal employment compensation.