Cognitive Behavioral TherapyVersus Sertraline in Patients WithDepression and Poorly ControlledDiabetes: The Diabetes andDepression (DAD) StudyA Randomized Controlled MulticenterTrialDiabetes Care 2015;38:767–775 | DOI: 10.2337/dc14-1599

OBJECTIVE

This study compared the long-term efficacy of a diabetes-specific cognitive be-havioral group therapy (CBT) with sertraline in patients with diabetes and de-pression who initially responded to short-term depression treatment.

RESEARCH DESIGN AND METHODS

A randomized controlled single-blind trial was conducted in 70 secondary carecenters across Germany comparing 12weeks of CBTwith sertraline in 251 patientswith type 1 or 2 diabetes (mean HbA1c 9.3%, 78 mmol/mol) and major depression(Structured Clinical Interview for DSM-IV [SCID]). After 12 weeks, treatment re-sponders (‡50% reduction Hamilton Depression Rating Scale [HAMD-17]) wereincluded in the 1-year study phase where CBT patients were encouraged to usebibliotherapy and sertraline patients received continuous treatment. We ana-lyzed differences for HbA1c (primary outcome) and reduction (HAMD-17) or re-mission (SCID) of depression from baseline to the 1-year follow-up using ANCOVAor logistic regression analysis.

RESULTS

After 12 weeks, 45.8% of patients responded to antidepressant treatment andwere included in the 1-year study phase. Adjusted HbA1cmean score changes frombaseline to the end of the long-term phase (20.27, 95% CI20.62 to 0.08) revealedno significant difference between interventions. Depression improved in bothgroups, with a significant advantage for sertraline (HAMD-17 change: 22.59,95% CI 1.15–4.04, P < 0.05).

CONCLUSIONS

Depression improved under CBT and sertraline in patients with diabetes and de-pression, with a significant advantage for sertraline, but glycemic controlremained unchanged. CBT and sertraline as single treatment are insufficient totreat secondary care diabetes patients with depression and poor glycemic control.

1Department of Psychosomatic Medicine andPsychotherapy, LWL-University Clinic Bochum,Ruhr-University Bochum, Bochum, Germany2Center for Psychotherapy Wiesbaden, Wiesba-den, Germany3Department of Psychosomatic Medicine andPsychotherapy, University of Cologne, Koln,Germany4Diabetes Center Mergentheim, Bad Mergen-theim, Germany5Department of Psychiatry and Psychotherapy,University Medical Centre, Johannes GutenbergUniversity, Mainz, Germany6Department of Clinical Psychology and Psycho-therapy, Institute of Psychology, JohannesGutenberg University Mainz, Mainz, Germany7Interdisciplinary Centre for Clinical Trials Mainz(IZKS Mainz), University Medical Centre,Johannes GutenbergUniversity,Mainz, Germany8Clinic for Psychosomatic and Psychotherapy,University Clinic Gießen/Marburg, Philipps Uni-versity Marburg, Marburg, Germany9Department of Health Psychology, Institute ofPsychology, Johannes Gutenberg UniversityMainz, Mainz, Germany10Vitos Clinical Centre Gießen-Marburg andJustus LiebigUniversityGießen,Marburg,Germany

Corresponding author: Frank Petrak, mail@dr-frank-petrak.de.

Received 30 June 2014 and accepted 11 January2015.

Clinical trial registration: ISRCTN89333241,http://www.isrctn.com.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc14-1599/-/DC1.

A slide set summarizing this article is availableonline.

© 2015 by the American Diabetes Association.Readers may use this article as long as the workis properly cited, the use is educational and notfor profit, and the work is not altered.

Frank Petrak,1,2 Stephan Herpertz,1

Christian Albus,3 Norbert Hermanns,4

Christoph Hiemke,5 Wolfgang Hiller,6

Kai Kronfeld,7 Johannes Kruse,8

Bernd Kulzer,4 Christian Ruckes,7

Daniela Zahn,9 and Matthias J. Muller10

Diabetes Care Volume 38, May 2015 767

CLIN

CARE/ED

UCATIO

N/N

UTR

ITION/PSYC

HOSO

CIAL

Depression is a common and potentiallylife-threatening comorbidity in diabe-tes. The known adverse effects includepoor treatment adherence (1), hyper-glycemia (2), increased health care costs(3), increased complications (4), cogni-tive decline (5), and increased mortality(6). Hence, the treatment of comorbiddepression is essential for the clinicalcare of patients with diabetes (7). Treat-ment goals focus not only on remissionor the improvement of depression butalso on the improvement of poor glyce-mic control (8) to prevent or delay long-term complications (9).Interventions were evaluated in ran-

domized controlled trials and summa-rized in two recent meta-analyses(10,11). Both meta-analyses concludedthat depression could be treated withmoderate success in patients with diabe-tes using psychological, pharmacological,or combined interventions. The overallresult for glycemic control is controver-sial, mostly due to severe methodologicallimitations and the heterogeneity of theexamined populations and interventions(11). In addition, collaborative care stud-ies with soundmethodology (e.g., studiesby Ell et al. [12] and Katon et al. [13]),tested the efficacy of this approach as awhole. However, these studies were notable to identify single effective compo-nents of compound treatments or evalu-ate the superiority of one single treatmentover another because of a design thatallowed a switch between pharmacologi-cal and psychological interventions. Mostrecently, beneficial long-term effects ofCBT to improved treatment adherence(compared with treatment as usual) inpatients with diabetes with depressionwere observed for adherence and glyce-mic control; however, the hypothesizedsuperior long-term treatment effects fordepression were not confirmed (14). Insummary, the evidence for a single treat-ment that could significantly improvedepression outcomes and glycemic con-trol at the same time remains so farinconclusive (10,11).One potential reason for contradic-

tory long term-results regarding depres-sion outcomes is a lack of distinctionbetween the initial treatment responseand maintenance of the treatment ef-fect. Whereas pharmacological and psy-chological treatments of depressionhave similar efficacy for the reductionof depression (10,11), whether the

same strategies have a similar efficacyon the maintenance of reduced depres-sion is unclear. Randomized trials compar-ing pharmacological and psychologicaltreatments for maintenance of treatmentresponse on depression are not yet pub-lished in people with diabetes. This wouldrequire including only people with diabe-tes and depression who initially yield aclinically significant depression reductionafter treatment and to test different treat-ment strategies for maintenance.

Against this background, we con-ducted the Diabetes and Depression(DAD) Study, a randomized controlledmulticenter trial in which the efficacy ofpsychotherapy (diabetes-specific cognitivebehavioral therapy [CBT]) was comparedfor the first time with a pharmacologicaltreatment for depression (sertraline) inpatients with poorly controlled diabetesand major depression who respondedinitially to these treatments with a clini-cally significant reduction of depression.

RESEARCH DESIGN AND METHODS

The study protocol was published in de-tail previously (15).

Study Participants and SettingFour coordinating trial centers (locatedin Bochum/Dortmund, Mainz, Dusseldorf/Koln, andBadMergentheim) organized therecruitment and treatment in seventytrial centers of outpatient secondarycare (specialized diabetes practices andambulatory care health services in clin-ics) located in different parts of Germany(predominantly in the Rhine-Main area,Ruhr area, and Dusseldorf/Koln) fromApril 2006 through May 2009. Eligibilitycriteria included insulin-treated diabetes(type 1 or 2), 21–69 years of age, majordepression according to DSM-IV criteria,and HbA1c .7.5% (58 mmol/mol) withinthe 9 preceding months and again in thescreening measurement. Key exclusioncriteria were suicidal ideations, psy-chotic symptoms, bipolar disorder, sub-stance abuse or dependence in the past6 months, psychotherapy in the preced-ing 3 months, current use of mood sta-bilizers, neuroleptics, antidepressants,or benzodiazepines, and liver enzyme el-evations to exclude severe liver dysfunc-tion (for a detailed list, see [15]).

The following eligibility and exclusioncriteria were revised and amended tothe protocol in August 2006 (when 60patients were randomized) and April

2007 (when 198 patients were random-ized): the age range was changed from21–65 years to 21–69 years to increasethe number of eligible subjects. The in-clusion criterion of HbA1c .8% (64mmol/mol) was changed to .7.5% (58mmol/mol) because we faced seriousdifficulties finding poorly controlled pa-tients in the secondary care recruitmentcenters. To enhance comparability withinternational studies and to prevent aselection bias, our independent scien-tific advisory board (15) recommendedexcluding patients only in case of clini-cally significant suicide risk or history ofattempted suicide in the past 12months. Therefore, patients with valuesabove 1 in the Hamilton Depression Rat-ing Scale (HAMD-17) “suicide” item(item 3, range 0–4) were not includedin the trial. The exclusion criterion re-garding the current use of psychotropicdrugs was modified to allow the inclu-sion of patients treated with low-potency neuroleptic drugs in low doses(,300 mg chlorpromazine dose equiva-lents per day) because these drugs areoften used to treat sleeping disordersand restlessness in patients with diabe-tes. Post hoc analysis revealed that noneof the patients received low-potencyneuroleptic drugs at the beginning orduring the study.

Recruitment Procedures andMeasuresThe medical records of the patientswere reviewed to assess age, type of di-abetes, insulin treatment, and HbA1c

levels. Patients were contacted by tele-phone to confirm basic eligibility criteriaand to invite them to a baseline screen-ing. Depression was measured with aquestionnaire-based screening (Centerfor Epidemiologic Studies DepressionScale [CES-D]) (16,17), followed by theStructured Clinical Interview for DSM-IV(SCID) (18) for patients with a CES-Dvalue .22. The severity of depressionwas measured with the HAMD-17 (19).Health-related quality of life was as-sessed with the Short Form Health Sur-vey (SF-36) (20), and diabetes-specificstress was assessed with the ProblemAreas in Diabetes Questionnaire (PAID)(21). Medical diagnostics included docu-mentation of current medication, concom-itant diseases, onset and treatment ofdiabetes, liver enzymes, and HbA1c (usinghigh-performance liquid chromatography).

768 CBT vs. Sertraline in Diabetes and Depression Diabetes Care Volume 38, May 2015

All blood samples were analyzed in acentral laboratory (Bioscientia Labora-tory, Mainz, Germany). Moreover, pa-tient had to participate in a shortdiabetes education program to be ran-domized (see INTERVENTIONS).

Randomization and BlindingPatients were randomized using blockrandomization. For each coordinatinginstitution, a separate randomizationprocedure using permuted blocks strat-ified by type of diabetes was performed.Randomization lists were computer gen-erated and maintained by the Interdis-ciplinary Centre for Clinical Trials Mainz(IZKS) and conducted by fax.The study was a single-blind trial; that

is, the treatment evaluation was con-ducted by research psychologists un-aware of the treatment allocation andnot involved in the recruitment or treat-ment of the patients.

Study DesignEligible patients were assigned to CBT orsertraline treatments (Fig. 1). Diabetes

treatment was not part of the trial pro-tocol and was continued “as usual.” Af-ter 12 weeks, those of both groups whoresponded to the short-term therapyphase ($50% reduction of the HAMD-17 baseline score or HAMD-17 post-treatment score #7) were includedin a 12-month, long-term phase. Thesepatients constituted the intention-to-treat (ITT) population. Nonrespondersof the short-term phase were excludedfrom the treatment protocol. All pa-tients entering the long-term phase re-ceived diabetes treatment as usual inthe trial center at 3-month intervals dur-ing the following 12 months. Sertralineresponders received continuous treat-ment as relapseprevention. CBT respond-ers did not receive further treatmentbut were encouraged to work with apatients’manual in the sense of a biblio-therapy (22) during the 1-year follow-upphase. The difference in the active treat-ment duration between both inter-ventions corresponds to usual clinicalpractices and thus ensures external

validity. Generally, group CBT is offeredfor a limited time, assuming that “carry-over” effects will stabilize the results(23,24), whereas sertraline is givenfor a longer period as relapse preventionin patients responding to the initialtreatment (25). Patients of both groupsunderwent the same number of visitsto control for the amount of physiciancontact.

At the 12-month follow-up, bothtreatment groups were reexamined re-garding the primary and secondary out-come variables. The primary outcomewas change of glycemic control, definedas the difference in the HbA1c valuefrom baseline to the end of the long-term phase. Secondary outcomes werethe reduction of the HAMD-17 score, re-mission of depression (not fulfilling theDSM-IV-TR criteria for depression ac-cording to the SCID and HAMD scores#7), HbA1c decrease of $1%, andchanges in SF-36 and PAID scores (allanalyses were compared with the base-line values to the end of the trial).

Figure 1—Design of the DAD study. RCT, randomized controlled trial.

care.diabetesjournals.org Petrak and Associates 769

Initially, the primary outcomewas im-provement of glycemic control, definedas a decrease of at least 1% in HbA1cvalue (yes/no) from baseline to theend of the long-term phase. Owing toadvice from the advisory board, theanalysis of the primary outcome param-eter was changed in November 2008 to“change of glycemic control,” defined asthe difference in the HbA1c value frombaseline to the end of the long-termphase. By changing the dichotomousinto a continuous end point, the statis-tical power of the trial could be en-hanced and the planned number oftrial subjects had to be reduced. Allchanges in eligibility criteria or outcomewere amended to the protocol duringthe trial and before breaking of theblinding after approval by the ethiccommittee.

Primary Hypothesis

CBT leads to a better improvement ofglycemic control compared with sertra-line treatment at the 1-year follow-up inpatients who initially responded toshort-term therapy (CBT or sertraline)with regards to improvement in depres-sion. This superiority of the CBT regard-ing glycemic control was expectedbecause the CBT focused not only ondepression but also on diabetes-relatedaspects (e.g., problems with self-management and adherence to treat-ment). In contrast, the sertraline groupreceived a purely psychopharmacologicaltreatment, without further focus on theirdiabetes-related problems.

Secondary Hypothesis

CBT and sertraline are equally effectivein remission of depression a) after 12weeks of treatment and b) at the1-year follow-up.

Interventions

Diabetes Education

Given the poor glycemic control of thepatients and considering that most pa-tients treated with insulin had likely un-dergone diabetes education previously(according to the German guidelines[26]), a short diabetes education pro-gram (2 3 3 h) was offered to all pa-tients as an update by trained diabeteseducators (education manual [15]).

Diabetes-Specific Group CBT

CBT was administered by clinical psy-chologists who had undergone system-atic training regarding the manual (CBT

manual [18]). CBT was delivered ingroups of 4 to 10 patients in an outpa-tient setting within a 12-week period.This treatment consisted of 10 sessions(20 h) using a manualized semistruc-tured CBT for depression, including dif-ferent diabetes-specific aspects, toimprove adherence to diabetes treat-ment and coping with diabetes. Psycho-education included information aboutthe association of mood, activities, andthe development and maintenance ofdepression. Participants learned aboutthe link between diabetes and moodand ways to influence impaired moodwith cognitive techniques.

Furthermore, participants were en-couraged to discuss diabetes-specificgoals, such as HbA1c target values, withtheir diabetologists and to specify be-havioral goals to improve their glycemiccontrol (15). Individual goal achieve-ment was assessed, and possible bar-riers to the goal attainment wereidentified and modified, if possible.Each participant received a patientworkbook that included theoreticalbackground, worksheets, and exercisesfor each session. Patients were encour-aged to continue working with the bookafter the end of the short-term phase tostabilize and generalize the improve-ment (patients’ manual [15]).

Sertraline Treatment

Treatment was started at a dose of 50mg/day and could gradually be raised to200 mg/day, with changes not exceed-ing 50 mg/week. Dose changes accord-ing to response and side effects werebased on the Clinical Global ImpressionRating (27) and the Udvalg for KliniskeUndersøgelser side effects rating scale(28).

Statistical Analysis

Sample Size

The power calculation was based on ex-pected differences in HbA1c levels of thecomparison groups in the ITT samplethat included all randomized patientswho entered the 12-month follow-upphase. Considering randomized con-trolled trials evaluating CBT (29) or se-lective serotonin reuptake inhibitors(30), a treatment difference of 1.0 61.6% HbA1c could be assumed to be rel-evant. Therefore, with 23 46 evaluablesubjects, the trial had an 85% power ofdetecting a treatment difference of1.0% HbA1c by means of a t test on a

two-sided significance level of a =0.05. Assuming a response rate of 40%after the short-term treatment, 230(23 115) subjects had to be randomized(statistical analyses plan) (15).

Outcome Variables

The primary outcome of the trial wasdefined as the difference in HbA1c valuesfrom the baseline to the end of the long-term phase comparing both treatmentgroups using an ANCOVA controlling forbaseline HbA1c value and a baselineHAMD-17 score. Categorical outcomes(e.g., remission of depression) were an-alyzed using logistic regression analysiscontrolling for baseline HbA1c valuesand baseline HAMD-17 scores. Improve-ment in HAMD-17, SF-36, and PAIDscores were evaluated using ANCOVAcontrolling for baseline HbA1c valuesand respective baseline scores. Analyseswere repeated controlling for potentialconfounders. A correlation analysis wasperformed to identify confounders (age,sex, coordinating institution, diabetestype, diabetes complications, educationyears, income, single/recurrent episode[s],and comorbidity) with other mental dis-orders. Baseline variables associated(P , 0.10) with long-term outcome var-iables (HbA1c, HAMD-17 score, SF-36,and PAID score, respectively) were in-cluded as further control variables. Allanalyses were conducted for the ITTpopulation. Subgroup analyses for thetype of diabetes were also performed.Owing to the purely exploratory charac-ter of these analyses, no adjustment formultiplicity was done.

Quality Assurance and SafetyClinical monitoring, data management,pharmacovigilance, regulatory affairs,and statistical analyses were conductedby the IZKS Mainz and are described indetail elsewhere (15).

c Data management: A data manage-ment plan describing data manage-ment procedures, data collection,data flow, and the data validationwas established for the DAD study.

c Monitoring: Clinical on-site monitoringwas performed by personal visitsfrom a clinical monitor according tostandard operating procedures of theIZKS. The monitor visited each site atregular intervals to ensure compliancewith the study protocol, good clinicalpractice, and legal aspects.

770 CBT vs. Sertraline in Diabetes and Depression Diabetes Care Volume 38, May 2015

c Independent scientific advisoryboard: An independent scientific ad-visory board reviewed the outcomeevery 12 months.

c Quality assurance of depression se-verity assessment: To assure thequality of the HAMD rating, weestablished a validated training pro-cedure (for details [15]).

c Adherence to the CBT manual: CBTsessions were videotaped to ensureadherence to the manual and to givecontinuous supervision by one of theauthors (F.P.), who is a CBT trainerand supervisor.

c Adverse events reported by the pa-tients or detected by the investigatorweremonitored and recorded contin-uously according to good clinicalpractice (15).

The trial was approved by theMedicalEthics Committee Hessen and was con-ducted according to the Declaration ofHelsinki (31). All patients gave writteninformed consent to participate in thetrial.

RESULTS

Study ParticipantsThe recruitment is described in Fig. 2.The baseline characteristics of the pa-tients showed no significant differencesbetween intervention groups as testedwith t tests or x2 tests (Table 1). Theresults of the drop-out analyses are re-ported in Supplementary Appendix 1.

OutcomeThe response rate after 12 weeks oftreatment was 45.8% (n = 115 of 251)for the total sample, with 42.1% (n = 53of 126) for CBT and 49.6% for sertralinetreatment (n = 62 of 125). This differ-ence was not statistically significant(P = 0.231). The 115 treatment respondersof both groups constituted the ITT anal-ysis group and were included in the long-term phase of the study.

Primary and Secondary OutcomePrimary and secondary outcomes are re-ported in Table 2. After 15 months, themean HbA1c remained nearly un-changed compared with the baselinemeasures in both intervention groups,with no significant difference betweenthe groups. The estimated treatmentdifference from the ANCOVA modelamounted to 20.27% (95% CI 20.62 to0.08, P = 0.129), indicating no significant

difference between the groups. Hence,the main hypothesis of the study, whichexpected an advantage for CBT treat-ment, could not be confirmed. Themean HbA1c values did not change sub-stantially during the course of the studyin either group. Depressive symptomsassessed by the HAMD-17 strongly de-creased in both groups, with a significantadvantage for sertraline (P = 0.020) and amoderate effect size difference betweengroups (d = 0.48, 95% CI 0.11–0.86, P =0.011). A nonsignificant trend (P = 0.083)was observed for remission of depres-sion in favor of sertraline.

In contrast with the physical compo-nent of the SF-36, which remainednearly unchanged compared with thebaseline score for both groups, themen-tal component of the SF-36 improvedconsiderably in both treatment groups.This result was similar for diabetes-related stress, with a strong decreasein the PAID score in both interventiongroups (SF-36 and PAID differences be-tween groups were statistically notsignificant).

The results of the per-protocol analy-ses, which included all patients whocompleted the treatment, demon-strated nearly identical results for theprimary outcome and the depression-related secondary outcome (Supple-mentary Appendix 2).

Subgroup AnalysesAlthough patients with type 1 diabe-tes showed a slight increase in HbA1c

after 15 months compared with thebaseline, a decrease was observed inpatients with type 2 diabetes (Supple-mentary Fig. 1). The unadjusted meandifference between groups was 0.63%(adjusted P = 0.0036). Patients withtype 2 diabetes showed an improvementin HbA1c when treated with CBT com-pared with the sertraline treatment,which was associated with an increasein mean HbA1c values (unadjustedmean difference between treatments0.66%, adjusted P = 0.0475; Supplemen-tary Fig. 1A). Patientswith type 1 diabetestreated with sertraline demonstrated astronger improvement in depressivesymptoms after 15 months comparedwith patients who received the CBT treat-ment (P = 0.0044). In contrast, the differ-ence between the interventions forpatients with type 2 diabetes was notstatistically significant (Supplementary

Fig. 1B). Accordingly, remission rateswere better for patients with type 1 di-abetes in the sertraline group comparedwith the CBT group (odds ratio 0.28, 95%CI 0.095-0.810; P = 0.017; Supplemen-tary Fig. 1C).

Adverse EventsThroughout the trial, 73 patients had atleast one hospitalization (CBT: n = 44;sertraline: n = 29), and 2 patients died(CBT: n = 1; sertraline: n = 1).

CONCLUSIONS

Contrary to our primary hypothesis, CBTdid not lead to a better improvement ofglycemic control compared with sertra-line treatment after 15 months. In fact,the very poor glycemic control remainednearly unchanged during the entire trialin both intervention groups.

The initial response rate regarding de-pressive symptoms after 12 weeks oftreatment was 45.8% for all randomizedpatients (in agreement with the firstpart of our secondary hypothesis), withsimilar response rates for the CBT andsertraline groups.

For the patients who qualified for thelong-term phase of the study by theirinitial response to treatment, we ob-served a significantly better depressionoutcome for patients treated with ser-traline after 15 months, which wasmainly due to better maintenance of re-duced depression in the sertralinegroup. This overall difference in theHAMD score (0.48 effect size) can beconsidered as moderate (32).

For the remission rates of depression,the observed differences betweengroups failed to achieve statisticalsignificance.

Subgroup analyses showed differen-tial treatment effects depending onthe type of diabetes. A moderateHbA1c improvement was shown whenpatients with type 2 diabetes were trea-ted with CBT and a slight deteriorationwas shown when treated with sertra-line, which led to a moderate differenceof 0.66% HbA1c between treatments.This effect was not seen in patientswith type 1 diabetes.

In contrast, patients with type 1 dia-betes benefited more from sertralinetreatment than the CBT group with re-gard to depressive symptoms. This wasalso the case for remission of depressionin patients with type 1 diabetes, with

care.diabetesjournals.org Petrak and Associates 771

nearly 74% of patients treated with ser-traline achieving remission comparedwith only 44% of the patients treatedwith CBT. This differential effect was

not observed in patients with type 2diabetes.

The short-term treatment responseregarding depression after CBT or

sertraline treatment in the current studyis in linewith randomized controlled trialsin major depression with and withoutcomorbid diabetes (33–37). Although

Figure 2—Enrollment, treatment, and follow-up of the study participants.

772 CBT vs. Sertraline in Diabetes and Depression Diabetes Care Volume 38, May 2015

negative results regarding glycemic con-trol were observed inmost of the studiesin the field (10,11), it was still unex-pected for us. Our expectation regardingthe assumed advantage of CBT wasbased on the specific intervention,targeting not only depression but alsoadhering to diabetes treatment recom-mendations, which was a unique ap-proach in published studies when westarted our trial in 2006. Meanwhile, re-sults of a trial with some similarities toour study became available (14): CBTwith a focus on increased treatment foradherence and depression was com-pared with enhanced treatment as usualfor patients with uncontrolled type 2 di-abetes and depression. After 4 monthsof treatment, CBT for adherence and

depression led to better results for ad-herence, depression, and glycemic con-trol. Those differences were no longerobserved after 8 and 12 months for de-pression but remained stable for adher-ence and glycemic control. Hence, like inour study, these results demonstrate thedifficulties identifying a specific treat-ment with simultaneous effect ondepression and glycemic control. How-ever, the observed long-term influenceon glycemic control makes it promisingto include the topic of adherence indepression treatments for this group ofpatients.

Study Limitations and StrengthsThe inclusion of secondary care of pa-tients with poor diabetes control with

major depression and the unintendedfact that we included only Caucasiansrestricts the generalizability of our re-sults. Adherence to diabetes treatmentwas not directly assessed, which is a fur-ther limitation of our study.

A strength of our work is that we con-ducted the study with a sound method-ology, includingmultiple quality assuranceaspects. These included the measure-ment of depression with SCID andHAMD interviews by trained clinical psy-chologists (instead of questionnaires, asin some important trials on this topic[12,13]), the video-controlled monitor-ing of adherence of psychologists tothe CBT manual, the establishment ofan independent scientific advisoryboard, the data management, and datavalidation with intense validation strate-gies. Finally, we limited bias due to cen-ter or investigator effects by includingeight clinical psychologists and 70 trialsites across Germany.

Several facts might explain the lack ofeffects regarding control in our study.

Selection of PatientsPrevious trials in the research field ofdiabetes and depression to a large ex-tent included primary care patients (e.g.,[12,13]) and the rare positive effects onglycemic control were mainly seen instudies conducted in primary care pa-tients (13,38,39). In contrast, we re-cruited participants exclusively fromsecondary care study sites. In Germany,it can be expected that most patientswith type 1 diabetes are treated in sec-ondary care (40), whereas almost all pa-tients with type 2 diabetes are treated inprimary care and are referred to second-ary care only when treatment fails. Thus,recruiting patients with type 2 diabetesin secondary care is already selecting pa-tients in whom usual treatment is failing.When considering the mean HbA1c base-line value of 9.3% (78 mmol/mol) in oursample, we had to assume that werecruited a group of very “difficult-to-treat patients” even in the secondarycare setting and at least with respect totheir glycemic control.

Severity of DepressionThe depression in most of our patientswas moderate to severe. Consideringthat trials with positive results in glyce-mic control preferentially used ques-tionnaires to assess depression (13,39),we likely included patients with a more

Table 1—Characteristics of the randomized patients at baseline

CBT group Sertraline group Total sampleCharacteristics n = 126 n = 125 N = 251

Age (years) 49.0 6 10.6 47.9 6 12.8 48.5 6 11.7

Female sex 79 (62.7) 77 (61.6) 156 (62.2)

Caucasians 126 (100) 125 (100) 251 (100)

Years of formal education,10 62 (49.2) 56 (44.8) 118 (47.0)10–14 53 (42.1) 64 (51.2) 117 (46.6).14 10 (7.9) 5 (4.0) 15 (6.0)

EmploymentEmployed or in training 64 (50.8) 62 (49.6) 126 (50.2)Retired 21 (16.7) 23 (18.4) 44 (17.5)Unemployed or disabled 22 (17.5) 21 (16.8) 43 (17.1)Homemaker/other 19 (15.1) 19 (15.2) 38 (15.2)

Type 1 diabetes 65 (51.6) 64 (51.2) 129 (51.4)

Type 2 diabetes 61 (48.4) 61 (48.8) 122 (48.6)

Diabetes duration (years) 15.7 6 10.4 15.0 6 10.6 15.3 6 10.5

HbA1c (%) 9.30 6 1.49 9.20 6 1.44 9.25 6 1.46

HbA1c (mmol/mol) 78 6 7.21 77 6 7.76 78 6 7.76

Retinopathy, nephropathy, or neuropathy 72 (57.1) 72 (57.6) 144 (57.4)

Macrovascular complications 23 (18.3) 23 (18.4) 46 (18.3)

Coronary heart disease 13 (10.3) 19 (15.2) 32 (12.7)

Major depression (SCID)Single episode 62 (49.2) 64 (51.2) 126 (50.2)Recurrent episodes 61 (48.4) 61 (48.8) 122 (50.2)Recurrent episodes with seasonal pattern 3 (2.4) 0 (0.0) 3 (1.2)

HAMD-17 scores (range 0–52) 18.3 6 4.8 18.8 6 5.0 18.5 6 4.9

Depression severity by HAMD-17 scoresMild (8–13) 21 (16.7) 16 (12.8) 37 (14.7)Moderate (14–19) 58 (46.0) 63 (50.4) 121 (48.2)Severe (20–25) 38 (30.2) 35 (28.0) 73 (29.1)Very severe (26–52) 9 (7.1) 11 (8.8) 20 (8.0)

PAID sum score (range 0–100) 47.5 6 18.4 49.1 6 16.5 48.3 6 17.5

SF-36 HRQoLMental component (z values) 22.9 6 1.1 22.8 6 1.0 22.9 6 1.1Physical component (z values) 21.1 6 1.2 21.2 6 1.2 21.1 6 1.2

Data are shown as means 6 SD or as n (%). Higher HAMD-17 scores indicate more diabetes-related burden. Higher PAID scores indicate more diabetes-related burden. HRQoL, health-related quality of life.

care.diabetesjournals.org Petrak and Associates 773

severe depression and potentiallypoorer treatment adherence, which inturn might have contributed to ourresults (41).

Group CBT and Single InterventionIt might be the case that treatment hasto be individualized evenmore distinctlyto achieve better glycemic control (i.e.,not only with respect to the type of di-abetes), as suggested by our subgroupresults, but also regarding the group in-tervention approach.

ConclusionBoth treatments were effective to im-prove depression in poorly controlledpatients with diabetes, but sertralinewas slightly more effective to maintainthese effects comparedwith CBT. Despitean intense treatment focus on the im-provement of adherence to diabetestreatment in the CBT group, glycemiccontrol did not improve in either inter-vention. The selection of very difficult-to-treat patients with poorly controlled

diabetes and depression and currentlyunknown aspects of the comorbidity ofdiabetes and depressionmight have con-tributed to the results of our trial and thedifferences that we observed with re-gards to the results of previous studies.

Our results demonstrate that CBT andsertraline, two widely used interven-tions in clinical practice, cannot be con-sidered as sufficiently effective in thetreatment of secondary care patientswith poor diabetes control and depres-sion. “Mood repair” alone does not au-tomatically result in improved diabetesoutcomes. Further research needs toaddress potential additional mecha-nisms that mediate the effect of depres-sion on glycemic control.

Acknowledgments. The authors thank all ofthe patients for their participation in the DADtrial.Funding. This work was supported by theKompetenznetz Diabetes mellitus (Competence

Network for Diabetes Mellitus) funded by theGerman Federal Ministry of Education andResearch (BMBF) (No. 01KG0505). All authorsreceived financial support from the BMBF forthe submitted work. The BMBF had no role inthe design and conduct of the study; collection,management, analysis, and interpretation ofthe data; and preparation, review, or approvalof the manuscript.Duality of Interest. C.H. has served as aconsultant for Servier (Paris, France) and Janssen-Cilag (Beerse, Belgium); has served on thespeakers bureaus of Bristol-Myers Squibb, EliLilly, Janssen-Cilag, Pfizer, and Servier; and is themanaging director of the psiac GmbH, Mainz,which provides an Internet-based drug interac-tion program for psychoactive drugs. No otherpotential conflicts of interest relevant to thisarticle were reported.Author Contributions. All authors designedthe study, were responsible for its conduct, andcontributed to the writing and editing of themanuscript. F.P., K.K., and C.R. were responsiblefor study management and data collection. C.R.undertook data analysis. F.P. is the guarantor ofthis work and, as such, had full access to allthe data in the study and takes responsibility forthe integrity of the data and the accuracy of thedata analysis.

Table 2—Differences between interventions from baseline to the end of the long-term phase in diabetes patients who initiallyresponded to short-term depression treatment (ITT analysis)

Unadjusted estimated

CBT group Sertraline groupOutcomes n = 53 Change n = 62 Change Adjusted between-group differences (95% CI)

HbA1c, % (mmol/mol)Baseline 9.37 6 1.63 (79) 9.15 6 1.37 (76)3 months 9.12 6 1.61 (76) 8.90 6 1.43 (74)15 months 9.22 6 1.67 (77) 20.15 9.41 6 1.36 (79) +0.26 20.27 (20.62 to 0.08)†

HbA1c decrease of $1% after15 months vs. baseline 5 (9.4) 4 (6.5) OR 1.43 (0.28–7.65)‡

HAMD-17Baseline 18.04 6 4.62 18.87 6 5.143 months 5.40 6 3.03 5.35 6 3.7215 months 7.83 6 6.49 210.21 5.46 6 5.75 213.41 2.59 (1.15–4.04)§*

Remission of depression, %15 months 27 (50.9) 41 (66.1) OR 0.47 (0.20–1.11)|

SF-36 (z values)Physical componentBaseline 20.88 6 1.06 21.14 6 1.20 d

3 months 21.04 6 1.01 21.29 6 1.1315 months 21.03 6 1.25 20.15 21.04 6 1.13 +0.10 0.16 (20.60 to 0.28)#

Mental componentBaseline 22.69 6 1.05 22.86 6 1.063 months 20.97 6 1.37 20.81 6 1.4415 months 21.09 6 1.58 +1.6 20.65 6 1.30 +2.21 20.36 (20.94 to 0.22)**

PAIDBaseline 45.93 6 17.89 50.31 6 15.833 months 37.12 6 18.86 37.92 6 16.6815 months 34.96 6 21.01 210.97 31.43 6 20.94 218.88 7.13 (20.87 to 15.12)††

Data are shown as mean 6 SD, as n (%), or as indicated. Change is the baseline mean minus the 15-month mean. OR, odds ratio. *Significantdifferences between groups P, 0.05. †Mean adjusted for baseline HbA1c and baseline HAMD-17 in ANCOVA. ‡Mean adjusted for baseline HbA1c,baseline HAMD-17, and sex in logistic regression analysis. §Mean adjusted for baseline HbA1c, baseline HAMD-17, and years of formal education inANCOVA. |Mean adjusted for baseline HbA1c, baseline HAMD-17, and years of formal education in logistic regression analysis. #Mean adjusted forbaseline HbA1c and baseline SF-36 (physical component) in ANCOVA. **Mean adjusted for baseline HbA1c, baseline SF-36 (mental component), age,and baseline macrovascular complication in ANCOVA. ††Mean adjusted for baseline HbA1c and baseline PAID in ANCOVA.

774 CBT vs. Sertraline in Diabetes and Depression Diabetes Care Volume 38, May 2015

References1. Gonzalez JS, Peyrot M, McCarl LA, et al. De-pression and diabetes treatment nonadher-ence: a meta-analysis. Diabetes Care 2008;31:2398–24032. Lustman PJ, Anderson RJ, Freedland KE, deGrootM, Carney RM, Clouse RE. Depression andpoor glycemic control: ameta-analytic review ofthe literature. Diabetes Care 2000;23:934–9423. Hutter N, Schnurr A, Baumeister H. Health-care costs in patients with diabetes mellitus andcomorbid mental disordersda systematic re-view. Diabetologia 2010;53:2470–24794. de Groot M, Anderson R, Freedland KE,Clouse RE, Lustman PJ. Association of de-pression and diabetes complications: a meta-analysis. Psychosom Med 2001;63:619–6305. Slomski A. Depression promotes cognitivedecline in patients with diabetes. JAMA 2013;310:21396. Egede LE, Nietert PJ, Zheng D. Depressionand all-cause and coronary heart disease mor-tality among adults with and without diabetes.Diabetes Care 2005;28:1339–13457. Holt RI, Katon WJ. Dialogue on Diabetes andDepression: Dealing with the double burden ofco-morbidity. J Affect Disord 2012;142(Suppl.):S1–S38. Petrak F, Herpertz S. Treatment of depres-sion in diabetes: an update. Curr Opin Psychia-try 2009;22:211–2179. American Diabetes Association. Standards ofmedical care in diabetesd2013. Diabetes Care2013;36(Suppl. 1):S11–S6610. van der Feltz-Cornelis CM, Nuyen J, Stoop C,et al. Effect of interventions for major depres-sive disorder and significant depressive symp-toms in patients with diabetes mellitus:a systematic review and meta-analysis. GenHosp Psychiatry 2010;32:380–39511. Baumeister H, Hutter N, Bengel J. Psycho-logical and pharmacological interventions fordepression in patients with diabetes mellitusand depression. Cochrane Database Syst Rev2012;12:CD00838112. Ell K, Katon W, Xie B, et al. Collaborativecare management of major depression amonglow-income, predominantly Hispanic subjectswith diabetes: a randomized controlled trial. Di-abetes Care 2010;33:706–71313. Katon WJ, Lin EH, Von Korff M, et al. Collab-orative care for patients with depression andchronic illnesses. N Engl J Med 2010;363:2611–262014. Safren SA, Gonzalez JS, Wexler DJ, et al. Arandomized controlled trial of cognitive behav-ioral therapy for adherence and depression(CBT-AD) in patients with uncontrolled type 2diabetes. Diabetes Care 2014;37:625–63315. Petrak F, Herpertz S, Albus C, et al. Studyprotocol of the Diabetes and Depression Study(DAD): a multi-center randomized controlled trialto compare the efficacy of a diabetes-specific

cognitive behavioral group therapy versus sertra-line in patients withmajor depression and poorlycontrolled diabetes mellitus. BMC Psychiatry2013;13:20616. Hautzinger M, Bailer M. Die AllgemeineDepressionsskala. Weinheim, Beltz, 199317. Radloff L. The CES-D scale: a self-reportdepression scale for research in the generalpopulation. Appl Psychol Meas 1977;1:385–40118. Wittchen HU, Zaudig M, Fydrich T. SKID.Strukturiertes Klinisches Interview fur DSM-IV.Gottingen, Hogrefe, 199719. U.S. Department of Health and Human Ser-vices. HAMD Depression Scale. In ECDEU Assess-ment Manual. Rockville, MD, Public HealthService - Alcohol, Drug Abuse, and Mental HealthAdministration Rev, 1976, p. 180–19220. Bullinger M, Kirchberger I. SF-36 Fragebo-gen zum Gesundheitszustand. Gottingen, Hog-refe, 199821. Kulzer B, Hermanns N, Ebert M, Kempe J,Kubiak T, Haak T. Problembereiche bei Diabetes(PAID)dein neues Meßinstrument zur Erfassungder emotionalen Anpassung an Diabetes. DiabetesStoffwechsel 2002;11(Suppl. 1):14422. Cuijpers P. Bibliotherapy in unipolar de-pression: a meta-analysis. J Behav Ther Exp Psy-chiatry 1997;28:139–14723. Gloaguen V, Cottraux J, Cucherat M,Blackburn IM. A meta-analysis of the effects ofcognitive therapy in depressed patients. J AffectDisord 1998;49:59–7224. DeRubeis RJ, Gelfand LA, Tang TZ, SimonsAD.Medications versus cognitive behavior ther-apy for severely depressed outpatients: mega-analysis of four randomized comparisons. Am JPsychiatry 1999;156:1007–101325. DGPPN, BAK, KBV, et al. S3-Leitlinie/Nationale VersorgungsLeitlinie UnipolareDepression-Langfassung (Version 1.3), [Internet]2012. Available from http://www.leitlinien.de/mdb/downloads/nvl/depression/unipolare-depression-vers1.3-lang.pdf. Accessed 30 June201426. Petrak F, Herpertz S, Albus C, Hirsch A,Kulzer B, Kruse J. Psychosocial factors and dia-betes mellitus: evidence-based treatmentguidelines. Curr Diabetes Rev 2005;1:255–27027. Guy W. Clinical Global Impressions (CGI)Scale.Modified fromAPAHandbookofPsychiatricMeasures. Rush AJ, Pincus HA, First M, et al.Washington, DC, APA, 200028. Lingjaerde O, Ahlfors UG, Bech P, Dencker SJ,Elgen K. The UKU side effect rating scale. A newcomprehensive rating scale for psychotropicdrugs and a cross-sectional study of side effectsin neuroleptic-treated patients. Acta PsychiatrScand Suppl 1987;334:1–10029. Lustman PJ, Griffith LS, Freedland KE, KisselSS, Clouse RE. Cognitive behavior therapy fordepression in type 2 diabetes mellitus. A

randomized, controlled trial. Ann Intern Med1998;129:613–62130. Lustman PJ, Clouse RE, Nix BD, et al. Sertra-line for prevention of depression recurrence indiabetes mellitus: a randomized, double-blind,placebo-controlled trial. Arch Gen Psychiatry2006;63:521–52931. World Medical Association. World MedicalAssociation Declaration of Helsinki: ethical prin-ciples for medical research involving humansubjects. JAMA 2013;310:2191–2194.32. Cohen J. Statistical Power Analysis forthe Bahavioral Sciences. Hillsdale, LawrenceErlbaum, 198833. Atlantis E, Fahey P, Foster J. Collaborativecare for comorbid depression and diabetes:a systematic review and meta-analysis. BMJOpen 2014;4:e00470634. Baumeister H, Hutter N, Bengel J. Psycho-logical and pharmacological interventions fordepression in patients with diabetes mellitus:an abridged Cochrane review. Diabet Med2014;31:773–78635. Cuijpers P, Karyotaki E, Weitz E,Andersson G, Hollon SD, van Straten A. Theeffects of psychotherapies for major depres-sion in adults on remission, recovery and im-provement: a meta-analysis. J Affect Disord2014;159:118–12636. Huang Y, Wei X, Wu T, Chen R, Guo A.Collaborative care for patients with depres-sion and diabetes mellitus: a systematic re-view and meta-analysis. BMC Psychiatry2013;13:26037. Huang KL, Lu WC, Wang YY, et al. Compar-ison of agomelatine and selective serotonin re-uptake inhibitors/serotonin-norepinephrinereuptake inhibitors in major depressive disor-der: A meta-analysis of head-to-head random-ized clinical trials. Aust N Z J Psychiatry 2014;48:663–67138. Lamers F, Jonkers CC, Bosma H, KnottnerusJA, van Eijk JT. Treating depression in diabetespatients: does a nurse-administered minimalpsychological intervention affect diabetes-specific quality of life and glycaemic control? Arandomized controlled trial. J Adv Nurs 2011;67:788–79939. Bogner HR, Morales KH, de Vries HF,Cappola AR. Integrated management of type2 diabetes mellitus and depression treatmentto improve medication adherence: a random-ized controlled trial. Ann Fam Med 2012;10:15–2240. Bohm BO, Dreyer M, Fritsche A,Fuchtenbusch M, Golz S, Martin M. Therapy oftype 1 diabetes - practitioner guidelines of theGermanDiabetes Association, DDG. Diabetologie2011;6:S120–S130 [in German]41. Ciechanowski PS, Katon WJ, Russo JE. De-pression and Diabetes: impact of depressivesymptoms on adherence, function, and costs.Arch Intern Med 2000;160:3278–3285

care.diabetesjournals.org Petrak and Associates 775