An Integrated Psychosomatic Treatment Program for People with Diabetes (psy-PAD)

Abstract

Background

Many people with diabetes have permanently elevated blood sugar concentrations and a high level of diabetes-related psychological stress, also called “diabetes distress.” In clinical practice, diabetes distress is often an impediment to successful self-management. psy-PAD is a psychodynamically oriented short-term therapy program whose goal is to reduce diabetes distress and improve glycemic control.

Methods

A randomized controlled trial was conducted with 143 patients with either type 1 or type 2 diabetes who were being treated in eleven specialized diabetological practices. psy-PAD in the intervention group (eight sessions) was compared with optimized standard care as the control condition. The inclusion criteria were HbA1c ≥ 7.5% combined with diabetes distress (PAID >35, or doctor’s determination). The primary endpoint was the HbA1c at six months (t₁). Diabetes-related distress (PAID), depressive symptoms (HADS-D, PHQ-9), anxiety symptoms (HADS-A), health-related quality of life (SF-36), panic (short form of the PHQ-D), body mass index (BMI), and triglyceride levels were secondary endpoints. Follow-ups were conducted at six (t₁) and 12 months (t₂) (trial registration: DRKS00003247).

Results

The intergroup comparison at t₁ revealed a significant, clinically relevant reduction of HbA1c by -0.53 percentage points (95% confidence interval [-0.89; -0.16], p = 0.005). The secondary analyses revealed relevant differences in the point estimators for diabetes distress at t₁ and t₂, depressive symptoms at t₂ and BMI at t₁.

Conclusion

For people with diabetes and diabetes distress who do not achieve satisfactory glycemic control despite intensive treatment in specialized diabetological practices, integrated psychosomatic-psychotherapeutic treatment can lower blood sugar levels over the intermediate term and also reduce diabetes distress and depressive symptoms over a one-year period.

With rising prevalences, more than 8 million people were diagnosed with type 2 diabetes mellitus in Germany in 2020. To this figure one can add approximately 32,000 children and adolescents, as well as around 340,000 adults, diagnosed with type 1 diabetes (1). Acute metabolic crises in the form of severe hyperglycemia, life-threatening hypoglycemia, and the development of micro-/macrovascular complications contribute to the increased risk of mortality compared with the normal population (2– 5). Therefore, one of the primary treatment goals of diabetes therapy is to achieve balanced blood sugar metabolism control in order to prevent the risk of acute crises and the development of diabetic complications. Despite intensive health policy efforts, for example in the form of disease management programs (DMPs) or specific diabetes education/training programs, unsatisfactory metabolic control is still found in a relevant proportion of patients. For example, according to the North Rhine DMP report, 58.4% of all people with type 1 diabetes and 40.4% with type 2 diabetes fail to achieve the blood sugar target levels agreed upon (usually < 7.5%) (6).

Patients’ self-management behavior plays a central role in blood sugar control. Reasons for unfavorable self-management behavior can be patient-related, practitioner-related, or environment-related. Against this backdrop, psychosomatic aspects increasingly come to the fore. Many people with diabetes experience a high level of diabetes-related psychological stress, also referred to as “diabetes distress,” (type 1: ˜ 44%; type 2: ˜ 25 % [7]), as well as comorbid mental illnesses, such as depressive disorders (type 1 ˜ 6% [8]; type 2 ˜ 15% [9]) and anxiety disorders (type 1: ˜ 8% [8]; type 2: ˜ 7 % [10]). These result not only in impediments to diabetes self-management and reduced quality of life but also in poorer glycemic control (11, 12), an increased risk of complications (13) and mortality (2– 5), as well as significantly increased costs (14). By focusing psychosocial treatment approaches on reducing diabetes distress, it may be possible to help affected individuals reduce impediments to treatment and, thus, also achieve improvements in their glycemic control. On an international level, integrated treatment approaches enabling access to structured interdisciplinary care concepts that are integrated in primary care have been shown to be effective, with heterogeneous results seen in terms of HbA1c improvement (15, 16).

This study evaluates a cross-sectoral, psychodynamically oriented, integrated psychosocial and psychosomatic treatment program for patients with diabetes (psy-PAD), which is implemented in a collaboration between psychosomatic outpatient clinics and diabetologists in specialized practices. The assumption is that psy-PAD can achieve a significant improvement in glycemic control (HbA1c) compared to optimized standard care. Reductions in diabetes distress, depressive symptoms, and anxiety, as well as an increase in health-related quality of life, were also anticipated.

Methods

The study was conducted as a two-armed randomized controlled trial (RCT). In contrast to the data held by the German Registry of Clinical Trials (DRKS), the HbA1c at t₂ that was erroneously filed there does not represent a primary endpoint; this also applies to all prospectively submitted documents, such as the ethics application and the application to the German Medical Association (Bundesärztekammer, BÄK) for funding. Due to the complexity of practical implementation in routine care in specialized diabetological practices, and in contrast to the original study protocol, the inclusion criterion “significant distress in the management of diabetes” was not evaluated by means of PAID alone, but also by the clinical impression of the treating diabetologists. Therefore, inclusion criteria included diagnosed type 1/type 2 diabetes, age between 18–70 years, previous completion of diabetes training, HbA1c≥ 7.5%, and significant diabetes distress (determined by PAID > 35 or identified by the treating diabetologist). Exclusion criteria included severe comorbid physical illness (for example, oncological disease), dementia, severe mental illness (for example, severe depressive episode, psychosis, addiction disorder), as well as insufficient knowledge of the German language. Recruitment was carried out in 11 specialized diabetological practices by means of informed consent.

In addition to standard care, patients in the intervention group (IG) received the short intervention psy-PAD. This is based on a low-threshold, psychodynamically oriented short-term therapy program, the goal of which is to achieve a reduction in psychosocial impediments to treatment and diabetes distress, thereby also improving metabolic control. In terms of content, it integrates elements from self-management therapy and solution-oriented psychotherapy in four phases (see also Köhler and Kruse [17] as well as eMethods Section 1). The psy-PAD program comprised eight individual sessions, which were initially held weekly (4 ×) and later monthly (4 ×) by psychotherapists from the psychosomatic outpatient department, Gießen, Germany, in the specialized practice. Patients in the control group (CG) received standard care in the specialized practices and also received a consultation in the psychosomatic outpatient department at baseline, as well as brief verbal contact with the psychosomatic department at 6 and 12 months. If, as part of this, any evidence of mental illness came to light, the patients were informed about treatment options. The treating diabetologist was also accordingly informed (= optimized standard care).

Data collection was conducted at three measurement time points, t₀ (= pre-intervention), t₁ (= 6 months post intervention), and t₂ (= 12 months post intervention). The primary outcome was HbA1c at t₁ (Bio-Rad VARIANT II; centralized determination at the certified laboratory of the German Diabetes Center, Düsseldorf). Secondary outcomes included body mass index (BMI) and triglyceride levels, as well as diabetes-related distress (PAID) (18– 20), anxiety and depressive symptoms (HADS) (21), health-related quality of life (SF-36) (22), and panic and depressive symptoms (PHQ-D and PHQ-9, respectively) (23), as recorded using questionnaires. Screening to exclude dementia and severe depressive episodes was performed using DemTect (24) and SCID-I (25). Further details on the instruments used are provided in the eMethods Section 2. Information on sample size, power, randomization, and blinding can be found in eMethods Section 3.

Statistical analyses

The analysis was performed according to the intention-to-treat (ITT) principle. All participants that had taken part in the first follow-up at t₁ were selected as the study population. Baseline characteristics were described separately for the intervention and control groups by frequency distributions, means (M) ± standard deviations (SD), and, for triglyceride levels, by geometric means (GM)*/: standard deviation factors (SDF). Normal distributions were assumed for the primary outcome HbA1c and the (continuous) secondary outcomes, while a log-normal distribution was assumed for the triglyceride levels distributed with a right skew. In accordance with the sample size determination, individual HbA1c differences (Delta t₁–t₀) between the intervention and control groups were compared using the independent groups t-test. In addition, sensitivity analyses using last observation carried forward (LOCF) analysis were performed based on the ITT population for the primary parameter of difference in HbA1c between t₀ and t₁ with group comparison by means of t-test. Here, in accordance with LOCF, the HbA1c difference value of 0 was imputed between t₀ and t₁. Secondary analyses were performed using linear mixed models with adjustment for baseline values at t₀ and for dependencies arising from repeated measures in both groups. For adjustment following repeated measures, the general covariance structure was selected. Analyses were performed using SAS Version 9.4 (STAT 15.2).

Results

Of the N = 213 patients screened in the specialized diabetological practices, n = 178 met the inclusion or exclusion criteria and were randomized to one of the two study conditions (see the Figure for details on patient flow and drop-out reasons). The sensitivity analysis for the primary outcome HbA1c at t₁ was based on the ITT population (N = 177), which also included the n = 34 (IG: n = 18; CG: n = 16) patients hitherto not included, for which no data regarding HbA1c at t₀ or t₁ were available. There was also no baseline value at t₀ for n = 4 patients in the IG and n = 5 patients in the CG.

Figure.

Flowchart on inclusion and course of study

Overall, data on t₀ and t₁ were available for n = 143 individuals, meaning that these were included in the evaluations (IG: n = 69; CG: n = 74, drop-out rate 19.7%). There were no patients for whom data were available at t₂ but not at t₁. Patient recruitment took place between February 2010 and June 2014. Tables 1 and 2 provide an overview of sociodemographics as well as information on mental health at t₀. A representation of sociodemographics stratified according to diabetes type can be viewed in eTable 1.

Table 1. Sociodemographic data at t0.

	Total (N = 143)	IG (n = 69)	CG (n = 74)
Sex n (%) – Females – Males	92 (64.3) 51 (35.7)	43 (62.3) 26 (37.7)	49 (66.2) 25 (33.8)
Age M (SD) – In years	48.0 (11.7)	49.4 (11.6)	46.7 (11.8)
Marital status n (%) – Single – Married/living in a partnership – Divorced/separated/widowed	25 (17.5) 98 (68.5) 20 (14.0)	12 (17.4) 49 (71.0) 7 (10.1)	13 (17.6) 49 (66.2) 10 (13.5)
Highest school-leaving qualification n (%) – Lower secondary school or elementary school-leaving certificate – Secondary school certificate/polytechnic high school – (Subject-specific) university entrance qualification/general university entrance qualification – Other	43 (30.1) 45 (31.5) 51 (35.7) 4 (2.8)	23 (33.3) 23 (33.3) 20 (29.0) 3 (4.4)	20 (27.0) 22 (29.8) 31 (41.9) 1 (1.4)
Occupational activity in the preceding 6 months n (%) – Employed full-time – Employed part-time – Unemployed – Unable to work – Pension/retirement/early retirement – Other activity	54 (37.8) 33 (23.1) 6 (4.2) 6 (4.2) 28 (19.6) 16 (11.2)	28 (40.6) 17 (24.6) 2 (2.9) 3 (4.4) 12 (17.4) 7 (10.1)	26 (35.1) 16 (21.6) 4 (5.4) 3 (4.1) 16 (21.6) 9 (12.2)
Incapacity for work in the preceding 6 months n (%) – Yes	69 (48.3)	36 (52.2)	33 (44.6)
Duration of incapacity to work*1 M (SD) – In days – Quartiles (Q1/median/Q3)	35.4 (58.5) 5.0/14.0/35.0	47.1 (70.6) 8.5/21.0/53.0	22.7 (38.7) 5.0/8.0/23.0
Diabetes type n (%) – Type 1 diabetes – Type 2 diabetes	68 (47.6) 75 (52.5)	31 (44.9) 38 (55.1)	37 (50.0) 37 (50.0)
Time since diabetes diagnosis M (SD) – In years – Quartiles (Q1/median/Q3)	16.1 (10.6) 8.0/13.0/24.0	15.2 (10.5) 7.0/12.0/22.0	16.9 (10.8) 8.0/14.5/24.0
Glucose metabolism regulation M (SD) – HbA1c	8.7 (1.1)	8.8 (1.2)	8.6 (1.1)
Concomitant diseases and complications*2 n (%) – Nephropathy – Retinopathy – Neuropathy – Diabetic foot syndrome/amputation on the feet – Cardiovascular diseases – Stroke – Other diseases	27 (19.4) 37 (26.6) 58 (41.7) 16 (11.5) 24 (17.3) 2 (1.4) 116 (83.5)	13 (19.7) 16 (24.2) 26 (39.4) 8 (12.1) 10 (15.2) 1 (1.5) 56 (84.8)	14 (19.2) 21 (28.8) 32 (43.8) 8 (11.0) 14 (19.3) 1 (1.4) 60 (82.2)
Inpatient treatment in the preceding 6 months*3 n (%) – Yes	30 (21.1)	13 (18.8)	17 (23.3)

*1 In relation to participants unable to work in the preceding 6 months; *2 n = 3 missing in the IG and n = 1 missing in the CG; *3n = 1 missing in the CG

IG, intervention group; CG, control group; M, mean; SD, standard deviation

eTable 1. Sociodemographic data at t0 broken down according to diabetes type.

	Type 1 diabetes			Type 2 diabetes
	Total (n = 68)	IG (n = 31)	CG (n = 37)	Total (n = 75)	IG (n = 38)	CG (n = 37)
Sex n (%) – Females – Males	49 (72.1) 19 (27.9)	24 (77.4) 7 (22.6)	25 (67.6) 12 (32.4)	43 (57.3) 32 (42.7)	19 (50.0) 19 (50.0)	24 (64.9) 13 (35.1)
Age M (SD) – In years	41.6 (10.6)	44.0 (11.4)	39.6 (9.7)	53.8 (9.5)	53.8 (9.9)	53.7 (9.3)
Marital status n (%) – Single – Married/living in a partnership – Divorced/separated/widowed	13 (19.1) 47 (69.1) 8 (11.8)	6 (19.4) 21 (67.7) 4 (12.9)	7 (18.9) 26 (70.3) 4 (10.8)	12 (16.0) 51 (68.0) 12 (16.0)	6 (15.8) 28 (73.7) 4 (10.5)	6 (16.2) 23 (62.2) 8 (21.6)
Highest school-leaving qualification n (%) – Lower secondary school or elementary school-leaving certificate – Secondary school certificate/polytechnic high school – (Subject-specific) university entrance qualification/general university entrance qualification – Other	13 (19.1) 25 (36.8) 29 (42.7) 1 (1.5)	6 (19.4) 13 (41.9) 11 (35.5) 1 (3.2)	7 (18.9) 12 (32.4) 18 (48.6) 0 (0.0)	30 (40.0) 20 (26.6) 22 (29.3) 3 (4.0)	17 (44.7) 10 (26.3) 9 (23.7) 2 (5.3)	13 (35.1) 10 (27.0) 13 (35.1) 1 (2.7)
Occupational activity in the preceding 6 months n (%) – Employed full-time – Employed part-time – Unemployed – Unable to work – Pension/retirement/early retirement – Other activity	25 (36.8) 22 (32.4) 2 (2.9) 3 (4.4) 4 (5.9) 12 (17.6)	12 (38.7) 12 (38.7) 0 (0.0) 1 (3.2) 2 (6.5) 4 (12.9)	13 (35.1) 10 (27.0) 2 (5.4) 2 (5.4) 2 (5.4) 8 (21.6)	29 (38.7) 11 (14.7) 4 (5.3) 3 (4.0) 24 (32.0) 4 (5.3)	16 (42.1) 5 (13.2) 2 (5.3) 2 (5.3) 10 (26.3) 3 (7.9)	13 (35.1) 6 (16.2) 2 (5.4) 1 (2.7) 14 (37.8) 1 (2.7)
Incapacity for work in the preceding 6 months n (%) – Yes	39 (57.4)	22 (71.0)	17 (45.9)	30 (40.0)	14 (36.8)	16 (43.2)
Duration of incapacity to work*1 M (SD) – In days – Quartiles (Q1/median/Q3)	35.3 (64.9) 5.0/12.0/35.0	48.8 (82.1) 7.0/18.0/56.0	17.9 (24.2) 4.0/6.0/21.0	35.6 (50.0) 7.0/15.5/40.0	44.6 (50.2) 14.0/23.0/50.0	27.8 (50.1) 5.0/13.5/26.5
Time since diabetes diagnosis M (SD) – In years – Quartiles (Q1/median/Q3)	20.3 (10.7) 12.5/21.0/28.0	18.6 (10.5) 12.0/16.0/28.0	21.8 (10.9) 13.0/24.0/29.0	12.2 (9.0) 6.0/10.0/17.0	12.5 (9.8) 6.0/10.0/19.0	1.0 (8.2) 6.0/10.0/16.0
Glucose metabolism regulation M (SD) – HbA1c	8.6 (1.0)	8.8 (1.2)	8.4 (0.8)	8.8 (1.2)	8.9 (1.2)	8.8 (1.3)
Concomitant diseases and complications*2 n (%) – Nephropathy – Retinopathy – Neuropathy – Diabetic foot syndrome/amputation on the feet – Cardiovascular diseases – Stroke – Other diseases	12 (18.8) 22 (34.4) 20 (31.3) 3 (4.7) 8 (12.5) 0 (0.0) 50 (78.1)	4 (14.3) 7 (25.0) 8 (28.6) 1 (3.6) 3 (10.8) 0 (0.0) 23 (82.1)	8 (22.2) 15 (41.7) 12 (33.3) 2 (5.6) 5 (14.0) 0 (0.0) 27 (75.0)	15 (20.0) 15 (20.0) 38 (50.7) 13 (17.3) 16 (21.3) 2 (2.7) 66 (88.0)	9 (23.7) 9 (23.7) 18 (47.4) 7 (18.4) 7 (18.4) 1 (2.6) 33 (86.8)	6 (16.2) 6 (16.2) 20 (54.1) 6 (16.2) 9 (24.3) 1 (2.7) 33 (89.2)
Inpatient treatment in the preceding 6 months*3 n (%) – Yes	13 (19.1)	6 (19.4)	7 (18.9)	17 (23.0)	7 (18.4)	10 (27.8)

*¹ In relation to participants unable to work in the preceding 6 months

*² n = 3 Missing in the IG and n = 1 missing in the CG

*³ n = 1 Missing in the CG

IG, intervention group; CG, control group; M, mean; SD, standard deviation

The main analysis of the primary outcome included mean change 6 months post intervention (Delta t₁–t₀). The secondary analyses include mean changes and estimated effects at both 6 months (Delta t₁–t₀) and 12 months (Delta t₂–t₀). Tables 3a and 3b provide a detailed overview of group differences (between-subject effects). Changes over time (within-subject effects) can be found in eTable 2.

Table 3a. Primary analysis: differences between IG and CG (between-subject effects)*.

	M (SD)	M (SD)	Group difference [95% CI]	p-Value
HbA1c (IG: n = 69; CG: n = 74)
t0 t1	8.84 (1.15) 8.48 (1.34)	8.61 (1.09) 8.78 (1.54)	Delta t1–t0: –0.53 [–0.89; –0.16]	0.005

*t-Test for independent groups

IG, intervention group; CG, control group; CI, confidence interval; M, mean; SD, standard deviation

Table 3b. Secondary analyses: differences between the IG and CG (between-subject ?effects)*.

	M (SD)	M (SD)	Estimated effects IG–CG (between-subject effects) Estimate [95% CI]
HbA1c (IG: n = 69/69/66; CG: n = 74/74/65)
t0 t1 t2	8.84 (1.15) 8.48 (1.34) 8.52 (1.63)	8.61 (1.09) 8.78 (1.54) 8.61 (1.65)	Delta t1–t0: –0.48 [−0.85; −0.12] Delta t2 –t0 −0.29 [−0.77; 0.20]
PAID (IG: n = 69/69/66; CG: n = 74/72/64)
t0 t1 t2	41.59 (18.23) 30.54 (20.03) 29.70 (18.71)	39.04 (19.02) 34.31 (19.17) 33.42 (19.77)	Delta t1–t0: −5.25 [−9.74; −0.77] Delta t2 –t0 : −5.35 [−10.26; −0.45]
PHQ-9 (IG: n = 69/69/66; CG: n = 74/72/65)
t0 t1 t2	9.86 (5.11) 7.80 (4.92) 7.11 (4.54)	9.84 (4.76) 8.68 (5.07) 8.60 (5.61)	Delta t1–t0: −0.86 [−2.19; 0.47] Delta t2 –t0 : −1.24 [−2.67; 0.18]
HADS: depression (IG: n = 69/69/66; CG: n = 74/72/65)
t0 t1 t2	7.38 (4.55) 6.28 (4.26) 5.63 (3.92)	7.08 (4.25) 7.04 (4.85) 6.99 (4.75)	Delta t1–t0: −0.92 [−2.02; 0.19] Delta t2 –t0 : −1.25 [−2.48; −0.02]
HADS: anxiety (IG: n = 69/69/66; CG: n = 74/72/65)
t0 t1 t2	7.32 (3.94) 6.17 (3.83) 6.29 (3.70)	7.27 (3.71) 6.79 (3.87) 7.05 (4.40)	Delta t1–t0: −0.56 [−1.48; 0.35] Delta t2 –t0 : −0.79 [−1.97; 0.40]
SF-36: total (IG: n = 69/69/65; CG: n = 74/72/65)
t0 t1 t2	55.39 (18.70) 61.09 (18.24) 63.33 (19.04)	55.57 (18.57) 59.79 (21.58) 57.97 (23.76)	Delta t1–t0: 1.12 [−4.11; 6.34] Delta t2 –t0 : 3.47 [−2.65; 9.58]
SF-36: mental health (IG: n = 69/69/65; CG: n = 74/72/65)
t0 t1 t2	38.18 (12.59) 43.37 (11.87) 45.24 (12.07)	38.37 (10.89) 40.77 (12.95) 41.47 (12.80)	Delta t1–t0: 2.67 [−0.83; 6.18] Delta t2 –t0 : 3.67 [−0.01; 7.36]
SF-36: physical health (IG: n = 69/69/65; CG: n = 74/72/65)
t0 t1 t2	44.30 (9.28) 44.26 (9.52) 43.97 (9.45)	44.45 (11.18) 45.14 (10.77) 43.42 (11.78)	Delta t1–t0: −0.96 [−3.34; 1.41] Delta t2 –t0 : −0.62 [−3.44; 2.21]
BMI (IG: n = 66/57/59; CG: n = 72/54/52)
t0 t1 t2	34.16 (7.94) 33.07 (8.13) 33.56 (8.21)	32.95 (8.27) 32.35 (8.33) 32.90 (8.17)	Delta t1–t0: −0.71 [−1.36; −0.06] Delta t2 –t0 : −0.12 [−0.83; 0.58]
	GM (SDF)	GM (SDF)	Estimate [95% CI]
log-triglycerides (IG: n = 69/69/67; CG: n = 74/74/65)
t0 t1 t2	138.42 (1.99) 134.92 (1.92) 137.15 (2.19)	131.22 (2.00) 135.50 (2.07) 136.85 (1.95)	Delta t1–t0: 0.95 [0.83; 1.09] Delta t2 –t0: 1.02 [0.87; 1.20]

* Linear mixed models with adjustment for baseline values at t0 and for dependencies due to repeated measures in both groups;

BMI, body mass index; CI, confidence interval; M, mean; SD, standard deviation

Instruments: PAID (diabetes-related distress), HADS-D and PHQ-9 (depression), HADS-A (anxiety), and SF-36 (health-related quality of life)

eTable 2. Secondary analyses*: differences within the IG and CG (within-subject effects).

	M (SD)	M (SD)	Mean changes tx–t0 (within-subject effects) Estimate [95% CI]
HbA1c (IG: n = 69/69/66; CG: n = 74/74/65)
t0 t1 t2	8.84 (1.15) 8.48 (1.34) 8.52 (1.63)	8.61 (1.09) 8.78 (1.54) 8.61 (1.65)	Delta t1–t0: IG: −0.33 [−0.60; −0.07] Delta t1 –t0 : CG: 0.15 [−0.11; 0.40] Delta t2 –t0 : IG: −0.31 [−0.65; 0.03] Delta t2 –t0 : CG: −0.02 [−0.37; 0.32]
PAID (IG: n = 69/69/66; CG: n = 74/72/64)
t0 t1 t2	41.59 (18.23) 30.54 (20.03) 29.70 (18.71)	39.04 (19.02) 34.31 (19.17) 33.42 (19.77)	Delta t1–t0: IG: −10.63 [−13.84; −7.42] Delta t1 –t0 : CG: −5.38 [−8.51; −2.24] Delta t2 –t0 : IG: −11.53 [−14.99; −8.07] Delta t2 –t0 : CG: −6.18 [−9.65; −2.70]
PHQ-9 (IG: n = 69/69/66; CG: n = 74/72/65)
t0 t1 t2	9.86 (5.11) 7.80 (4.92) 7.11 (4.54)	9.84 (4.76) 8.68 (5.07) 8.60 (5.61)	Delta t1–t0: IG: −2.04 [−2.99; −1.09] Delta t1 –t0 : CG: −1.18 [−2.11; −0.24] Delta t2 –t0 : IG: −2.54 [−3.55; −1.53] Delta t2 –t0 : CG: −1.30 [−2.31; −0.29]
HADS: depression (IG: n = 69/69/66; CG: n = 74/72/65)
t0 t1 t2	7.38 (4.55) 6.28 (4.26) 5.63 (3.92)	7.08 (4.25) 7.04 (4.85) 6.99 (4.75)	Delta t1–t0: IG: −1.04 [−1.83; −0.25] Delta t1 –t0 : CG: −0.13 [−0.90; 0.64] Delta t2 –t0 : IG: −1.49 [−2.36; −0.62] Delta t2 –t0: CG: −0.24 [−1.11; 0.64]
HADS: anxiety (IG: n = 69/69/66; CG: n = 74/72/65)
t0 t1 t2	7.32 (3.94) 6.17 (3.83) 6.29 (3.70)	7.27 (3.71) 6.79 (3.87) 7.05 (4.40)	Delta t1–t0: IG: −1.14 [−1.79; −0.49] Delta t1 –t0 : CG: −0.58 [−1.22; 0.06] Delta t2 –t0: IG: −1.04 [−1.88; −0.21] Delta t2 –t0 : CG: −0.26 [−1.09; 0.58]
SF-36: total (IG: n = 69/69/65; CG: n = 74/72/65)
t0 t1 t2	55.39 (18.70) 61.09 (18.24) 63.33 (19.04)	55.57 (18.57) 59.79 (21.58) 57.97 (23.76)	Delta t1–t0: IG: 5.50 [1.76; 9.23] Delta t1 –t0: CG: 4.38 [0.72; 8.04] Delta t2 –t0: IG: 6.81 [2.48; 11.13] Delta t2 –t0: CG: 3.34 [−0.98; 7.66]
SF-36: mental health (IG: n = 69/69/65; CG: n = 74/72/65)
t0 t1 t2	38.18 (12.59) 43.37 (11.87) 45.24 (12.07)	38.37 (10.89) 40.77 (12.95) 41.47 (12.80)	Delta t1–t0: IG: 5.27 [2.76; 7.77] Delta t1 –t0: CG: 2.59 [0.14; 5.05] Delta t2 –t0 : IG: 6.82 [4.21; 9.44] Delta t2 –t0: CG: 3.15 [0.55; 5.75]
SF-36: physical health (IG: n = 69/69/65; CG: n = 74/72/65)
t0 t1 t2	44.30 (9.28) 44.26 (9.52) 43.97 (9.45)	44.45 (11.18) 45.14 (10.77) 43.42 (11.78)	Delta t1–t0: IG: 0.05 [−1.64; 1.75] Delta t1 –t0 : CG: 1.02 [−0.64; 2.68] Delta t2 –t0 : IG: −0.60 [−2.61; 1.40] Delta t2 –t0 : CG: 0.01 [−1.98; 2.01]
BMI (IG: n = 66/57/59; CG: n = 72/54/52)
t0 t1 t2	34.16 (7.94) 33.07 (8.13) 33.56 (8.21)	32.95 (8.27) 32.35 (8.33) 32.90 (8.17)	Delta t1–t0: IG: −0.75 [−1.22; −0.29] Delta t1 –t0 : CG: 0.04 [−0.50; 0.41] Delta t2 –t0 : IG: −0.51 [−1.00; −0.02] Delta t2 –t0 : CG:−0.39 [−0.90; 0.12]
	GM (SDF)	GM (SDF)	Estimate [95% CI]
log-triglycerides (IG: n = 69/69/67; CG: n = 74/74/65)
t0 t1 t2	138.42 (1.99) 134.92 (1.92) 137.15 (2.19)	131.22 (2.00) 135.50 (2.07) 136.85 (1.95)	Delta t1–t0: IG: 0.98 [0.89; 1.08] Delta t1 –t0: CG: 1.03 [0.94; 1.13] Delta t2 –t0 : IG: 0.99 [0.88; 1.11] Delta t2 –t0 : CG: 0.97 [0.86; 1.09]

* Linear mixed models with adjustment for baseline values at t0 and for dependencies due to repeated measures in both groups; for more details see eMethods Section 2.

BMI, body mass index; GM, geometric mean; IG, intervention group; CG, control group; CI, confidence interval; M, mean; SD, standard deviation; SDF, standard deviation factors; instruments: PAID (diabetes-related emotional distress), HADS-D and PHQ-9 (depression), HADS-A (anxiety); SF-36 (health-related quality of life)

Secondary analyses: HbA1c

For HbA1c at t₁, a relevant difference in point estimators was seen only in the IG (-0.33 %, 95% CI: [-0.60; -0.07]), not in the CG. When comparing t₂ to t₀ , the within-subject effect was still clinically relevant, but differences in point estimators were no longer visible for either group

Secondary analyses: diabetes-related distress (PAID)

For the PAID, relevant differences in point estimators were found for both groups and at both measurement time points (IG t₁ : -10.63 [-13.84; -7.42]; CG t₁ : -5.38 [-8.51; -2.24]; IG t₂: -11.53 [-14.99; -8.07]; CG t₂ : -6.18 [-9.65; -2.70]).

Secondary analyses: psychological distress (PHQ-9, HADS)

For depression (HADS-D), relevant differences in point estimators were seen in the IG at both measurement time points (t₁ : -1.04 [-1.83; -0.25]; t₂ : -1.49 [-2.36; -0.62]), but not in the CG. In the PHQ-9, on the other hand, relevant differences in point estimators were seen in both groups and at both measurement time points (IG t₁ : -2.04 [-2.99; -1.09]; CG t₁ : -1.18 [-2.11; -0.24]; IG t₂ : -2.54 [-3.55; -1.53]; CG t₂ : -1.30 [-2.31; -0.29]).

For anxiety, the HADS-A revealed relevant differences in point estimators in the IG at both measurement time points (t₁ : -1.14 [-1.79; -0.49]; t 2 -1.04 [-1.88; -0.21]). As with the HADS-D, no relevant difference in point estimators could be seen for the HADS-A in the CG.

Secondary analyses: health-related quality of life (SF-36)

For mental health as determined using SF-36, relevant differences in point estimators were seen in the total score in both groups at t₁ (IG: 5.50 [1.76; 9.23]; CG: 4.38 [0.72; 8.04]). At t₂ , relevant differences in point estimators were seen only in the IG (IG: 6.81 [2.48; 11.13]).

For mental health, relevant differences in point estimators were seen in both groups and at both measurement time points (IG t₁ : 5.27 [2.76; 7.77]; CG t₁ : 2.59 [0.14; 5.05]; IG t₂ : 6.82 [4.21; 9.44]; CG t₂ : 3.15 [0.55; 5.75]).

With regard to physical health, no relevant differences in point estimators were seen for either group or at any measurement time point.

Secondary analyses: BMI

With regard to BMI, relevant differences in point estimators were seen at both measurement time points only in the IG (t ₁ : -0.75 [-1.22; -0.29]; t₂ : -0.51 [-1.00; -0.02]), not in the CG.

Secondary analyses: log-triglycerides

For log-triglycerides, no relevant differences in point estimators were seen in either group or at any measurement time point.

Primary analysis: HbA1c at t1

In the primary analysis, HbA1c differences between t₀ and t₁ differed statistically significantly and clinically relevantly from each other between IG (M = - 0.36 percentage points [SD = 1.06]) and CG (M = 0.16 percentage points [SD = 1.16]) with a group difference of -0.53 percentage points (95% confidence interval: [-0.89; -0.16], p = 0.005).

Likewise in the sensitivity analysis of the ITT population, the reduction in HbA1c in the IG was significantly greater compared to the CG. The mean difference in Delta t₁-t₀ HbA1c values between IG and CG was -0.42 percentage points ([0.72; -0.12], p= 0.006).

Secondary analyses

With a mean change of -0.48 percentage points in HbA1c, the comparison between the IG and the CG at t₁ showed a relevant difference in the point estimators in favor of the IG; however, this difference was no longer evident in the further time course from t₀ to t₂. Since linear mixed models with adjustments were used in the secondary analyses, slight differences emerge in the mean change in HbA1c at t₁ (-0.48 percentage points) compared with the primary analysis (-0.53 percentage points).

For PAID, a relevant difference in point estimators was found in favor of the IG at both measuring time points. At t₁, the mean PAID change between IG and CG was -5.25 points, and at t₂ -5.35 points.

While there was no relevant difference in point estimators regarding depression (HADS-D) in the comparison between the IG and the KG at t₁, a relevant difference was found in favor of the IG (1.25 points) at 12-month follow-up at t₂. The converse was true for the mean difference in BMI: here, with –0.71 BMI points at t₁, a difference in point estimators was found in favor of the IG, but this was no longer evident at t₂.

For PHQ-9, HADS-A, SF-36, and triglyceride levels, a comparison of IG and CG revealed no relevant difference in point estimators at either of the measurement time points.

Discussion

The results on the primary endpoint HbA1c show that individuals with diabetes who had elevated blood sugar levels (HbA1c ≥ 7.5%) and diabetes distress in a specialized diabetology practice were able to achieve clinically relevant improvements in glycemic control through the low-threshold, cross-sector psychodynamically oriented short-term therapy program psy-PAD compared with optimized standard care. With a mean change of -0.53 percentage points in the 6-month mental health follow-up, there was a statistically significant (p = 0.005) and clinically relevant intergroup difference. Secondary analyses also looked at 12-month mental health follow-up. However, with a mean change in HBA1c of -0.29 percentage points, there were no differences in the point estimators. In this context, the effect at 1 year is also very close to the relevance threshold of 0.3 percentage points used by the European Medicines Agency (EMA) to assess a clinically relevant group difference (26). The fact that, despite this, no relevant difference in the point estimators could be achieved may be explained by the fact that sample size determination was primarily geared towards the comparison of HbA1c in the IG and the KG from t₀ to t₁. Although the power for this comparison was sufficient despite the slightly lower number of cases (N = 143 instead of N = 150), it is possible that, at times, this was insufficient to show the t₀-t₂ effect for HbA1c as well as secondary outcomes due to the lower number of cases at t₂.

A significant overall reduction in diabetes distress was seen over the course of the study, thereby making it possible to address one of the therapeutic goals of diabetes treatment. In the group comparison, the reduction was greater in the IG both at 6 months and at 12 months post intervention than under optimized standard care. In addition, the 12-month mental health follow-up revealed a difference with regard to depressive symptoms. This effect was seen even though the intervention was deliberately designed in such as way as to not primarily address depressive symptoms, but instead to focus on diabetes distress and its causes. Thus, the results also provide an indication that depressive symptoms in patients with diabetes are closely linked to diabetes distress. For anxiety, on the other hand, there were no group differences, even though a significant reduction was seen in the IG at both measurement time points. Despite the fact that in the absence of a passive control group (i.e., standard care) it is not possible to make any statements about the effect of an additional psychosomatic outpatient consultation and how the further standard care in the specialized practice would have been affected without such an option, it is reasonable to assume that even a low-threshold option such as a psychosomatic outpatient consultation in the context of dedicated psychotherapeutic consultation hours can be helpful for patients. This is also evident in the fact that there was significant improvement in mental health in the two groups at both the 6-month and 12-month mental health follow-ups.

Neither of the study conditions appeared to have an effect on physical health or triglyceride levels. For BMI, on the other hand, the group comparison showed a medium-term reduction in favor of the IG, but this was no longer evident at 12 months. Thus, overall, it cannot be assumed that offering short-term therapy leads to long-term somatic improvements.

Limitations

As a result of study drop-outs, it is not possible to exclude bias. Drop-out analyses showed that the study population (N = 143) had higher scores for weight and physical health-related quality of life and lower scores for depression compared with dropouts between t₀ and t₁ (N = 34). Moreover, patients that dropped out before t₁ were more likely to be married or widowed and less likely to suffer from a panic disorder or a depressive disorder. There were no differences with regard to the primary outcome HbA1c. Secondary sensitivity analysis of the ITT population regarding the primary outcome HbA1c at t₁ verified the result of the primary analysis. A multiple imputation analysis was not performed since the LOCF analysis revealed no evidence of drop-out bias in the results. The psy-PAD study addressed a highly distressed patient population in specialized practices, meaning that it is not possible to readily extrapolate the results to patients in, for example, primary care or with less pronounced symptoms. Moreover, psy-PAD was a cooperation model between diabetology practices and psychosomatic outpatient departments in the Gießen/Marburg/Wetzlar area. An assessment needs to be made as to whether the regional networks and collaborations could similarly be transferred to other regions in Germany, especially rural areas. Furthermore, it should be noted that a comparison was carried out with an active control group that also received a psychosomatic intervention in the form of one consultation. This means that the results are at times more conservative than would have been the case if they had been compared with conventional routine care, which includes virtually no psychosomatic treatment components. In addition, there is no knowledge regarding which other, for example psychotherapeutic, services were used during the study period.

Conclusion

The psy-PAD study showed that an integrated, cross-sectoral, psychodynamically oriented therapy program in people with diabetes and problematic glycemic control as well as increased diabetes distress can achieve a clinically meaningful and statistically significant reduction in HbA1c in the medium term. In addition, the group comparison demonstrated a reduction in diabetes-related distress and depressive symptoms. Future research should investigate the possibilities (medical, psychosocial, and economic) of further networking specialized diabetology practices and psychosomatic outpatient departments to improve care and evaluate the potential of low-threshold psychotherapeutic/psychosomatic programs.

Supplementary Material

eMethods section 1

Description of the psy-PAD intervention

Content orientation and aims of the psy-PAD intervention (see also [e1])

The psy-PAD intervention is based on a low-threshold psychodynamically oriented short-term therapy program that integrates elements from self-management therapy and solution-oriented psychotherapy. In order to be able to integrate the intervention into diabetes treatment, regular consultations took place between the diabetologists treating the patients in the specialized practices and the staff of the psychosomatic outpatient clinics. The intregration of the latter also facilitated, among other things and when necessary, access to further psychosomatic care.

The aim of psy-PAD is to achieve a reduction in psychosocial treatment impediments and diabetes-related emotional distress and, as a result, improve metabolic control in patients with diabetes.

Structure

psy-PAD comprised eight individual sessions, which were initially held weekly (4 ×) and later monthly (4 ×). Overall, the intervention was divided into four phases:

Phase 1 focuses on establishing a viable therapeutic relationship, identifying individual psychosocial impediments to treatment, communicating diabetes-related and psychological interrelationships to achieve a comprehensive understanding of the disease, and taking the patient’s biographical history. The latter should help in understanding the patients’ structural level and internalized conflicts and in learning to classify accompanying difficulties in dealing with diabetes. Based on the diagnosis, current disease status, and critical disease phases, both illness behavior and treatment behavior, previous coping strategies and their associated affects, as well as psychological symptoms are explored. At the same time, the patient is asked about their resources (for example, social network, occupational stability), as are diabetes-related interactions with family members and current reference persons.

Phase 2 primarily comprises determining specific and diabetes-relevant treatment goals that can be achieved during the sessions. These include, for example, integrating more self-care or exercise sessions into daily life or having conversations with relatives about their disease and how to deal with it. Furthermore, treatment options are discussed regarding other diabetes-related goals that cannot be achieved during therapy, as well as possible comorbid mental disorders. Goals are also set in terms of therapy, such as building up motivation to start or make arrangements for outpatient or inpatient psychotherapy or psychosomatic rehabilitation measures. As part of a participative decision-making process, achievable treatment goals are defined against the background of the individual problem situation and the available personal resources, with particular attention to psychosocial treatment impediments and the current phase of motivation to change (e2). If it becomes apparent during the session that a patient is markedly ambivalent toward the required behavioral changes, the goals should be modified and, in a first step, an understanding of this ambivalence developed. During the therapy session, the therapist adopts an attitude that is understanding of the difficulties in implementing behavioral changes that are a normal part of the process of further development and which follow an internal logic.

Phase 3 is characterized by its orientation toward resource activation and self-management (e3). Since the psychosocial impediments to treatment, the psychological symptoms, and the therapeutic goals derived from these are highly heterogeneous in these patients, different intervention elements are operationalized based on the elements of psychodynamic psychotherapy (e3), which are used according to the individual needs of each patient.

Phase 4 concentrates on working together with the patient to identify possible ways to maintain the improvements that have been achieved. In addition, further therapy options are recommended and, if necessary, appropriate measures initiated.

Detailed description of the treatment modules

Building an understanding of the difficulty of behavior

modification

This intervention element is aimed at patients who, as a result of repeated unsuccessful attempts to implement the behavioral changes necessary for diabetes management (for example, weight reduction), experience shame and guilt and become increasingly withdrawn. The affected patients are supported by the therapist to develop an understanding of the difficulties involved in change and to adopt an observing and non-judgmental attitude toward their own actions. Through empathetic inquiring on the part of the therapist, the patient is freed up to temper their harsh treatment of themselves and to understand their difficulties not in terms of “fault” or “guilt.” Furthermore, a way is elaborated together with the patient as to how they can implement a desire for change in appropriate steps. If, by working through the implementation difficulties, it becomes clear that there is a link to an interactional problem, the patient can be provided with initial explanations of this as part of the short-term therapy, and, if necessary, the initiation of psychotherapeutic treatment is recommended and arrangements made.

Strengthening self-esteem and integrating the disease into

the patient’s self-image

Many patients with diabetes find the change in self-perception due to their chronic physical disease very challenging, which can lead to problems in accepting and coping with the disease. Since the self-esteem of these patients is often linked to an ideal of strength and health, patients with this problem desire recognition and affection and fear being devalued as sick people. Therefore, they often demonstratively exhibit strength and health and disregard medical advice in order not to be different to their healthy counterparts. Patients often react to possible subsequent reproaches from their physician and more checks by further neglecting their disease—thus leading to poorer blood sugar regulation—in order to avoid having to experience the reduced self-esteem and shame. Therefore, the aim of this therapy element is to help the patient understand their own behavior in the context of self-limiting beliefs, put these into perspective, and, if necessary, attribute them to biographical experiences. Integrating the disease into the patient’s self-image is also accompanied by an active consideration of the tension between the patient’s own wishes and needs on the one hand and the requirements imposed by diabetes therapy on the other.

Work on affect perception, affect differentiation, and affect

expression

This module is aimed at patients who have difficulty openly expressing the affects, symptoms, and limitations associated with their disease. This problem is often linked to early biographical experiences, which can be associated with interpersonal difficulties, for example in a partnership or in the family. As a result, messages in interpersonal relationships that cannot be expressed in other ways, for example, anger, disappointment, or irritation, as well as desires for solicitude and autonomy, are “regulated” by difficult-to-control diabetes. The aim of this treatment module is to help patients better perceive their own diabetes-related affects, to express them free from fear, and, if necessary, to understand them through the prism of situations of tension in interpersonal relationships. It is made clear to affected patients that their needs and difficulties are seen and heard and that they can unburden themselves through the discussions.

Clarifying how to deal with restrictions on autonomy

Coming to terms with diabetes often (re-)activates an individuation–dependency conflict, since treatment-related restrictions on the individual degree of freedom often conflict with the individual’s desire to make autonomous choices in life. This inner conflict may express itself, for example, in highly irregular blood sugar measurements and “injecting according to gut feel.” As a result, the treating diabetologist recommends more frequent blood sugar monitoring and regular insulin administration, which often worsens the willingness of the affected person to change. Therefore, this element of treatment aims to respond to patients’ desires for independence in an understanding and accepting manner, while at the same time working with them on the extent to which they would like to see changes in the way they deal with their disease and how, where necessary, this can be worked on together.

Problem-solving training

If it is agreed upon that the goal of therapy is to improve specific, individual diabetes-related stress and conflict situations, patients are helped to develop new perspectives and creatively formulate possible new solutions for problematic situations by means of multi-stage problem-solving training. Through this, the affected patients should implicitly acquire a general problem-solving attitude that enables them to additionally apply the problem-solving strategies they have developed to other, possibly diabetes-independent problem areas.

Building-up regenerative resources

If a patient wishes to improve their regenerative activities (also generally in the sense of resource activation), individual ways of balancing out stress are elaborated using a variety of elements from enjoyment training. In this process, previous ways of relieving stress that the affected patients may have neglected can be picked up again and incorporated in everyday life. At the same time, patients are helped to better pick up on signals telling them that they need to recuperate. Patients are taught exercises to promote enjoyment in everyday life and are encouraged to include enjoyable experiences in their daily routine. Having said that, these regenerative activities should not be misconstrued as additional requirements, but should be seen as opportunities for enjoyment and fun.

Strengthening social resources

If the goal of therapy is to improve social contacts and social support, patients are helped as part of psychodynamic and empowerment-oriented interventions to reflect on their current social network and, where necessary, strengthen their support contacts or develop new support contacts. In this context, the diabetes team or diabetes support groups can also be seen as social resources in a broader sense.

eMethods section 2

Description and psychometric properties of the measurement instruments used

Problem Areas In Diabetes Scale: PAID

The PAID (e4) is a 20-point questionnaire to measure diabetes distress. The possible answers to each item range from 0 = “not a problem” to 4 = “a serious problem.” The sum value is found by adding up the scores for each item and multiplying this by 1.25, yielding a total score of between 0 and 100. This makes it possible to measure the severity of diabetes distress, with higher scores being associated with greater distress. A score of > 40 indicates severe diabetes-related emotional distress, which can also point to significant depressive symptoms (e4, e5). With a Cronbach’s α ≥ 0.93 for the scale as a whole, the PAID has good psychometric properties (e6, e7). Data on the minimal clinically important change (MCIC) are not available for the PAID (e8). The short form of the PAID (PAID-1/PAID-5) was used as the screening instrument in patient recruitment (e9).

Patient Health Questionnaire: PHQ-D (short form)

The PHQ-D is a well-validated psychodiagnostic instrument that can be used both in clinical practice and for research questions (e10). Using the short version, it is possible to assess depressive disorders (major depression and other depressive disorders), panic disorder, and psychosocial functioning (e11). The short form of the PHQ-D comprises 15 items (nine on depressive symptoms, five on panic disorder, and one item on psychosocial functioning) and enables the measurement of depression severity. As such, the PHQ-D depression scale corresponds to the established PHQ-9 (Patient Health Questionnaire) (e12). To calculate the total scale score for depressive disorders, the severity scores for the individual items are added up. The score can range from 0 and 27. Scores between 0 and 4 indicate the absence of depression, scores between 5 and 10 indicate mild depression, while 10 points or more indicate major depression.

Hospital Anxiety and Depression Scale: HADS

The HADS (e13, e14) is an instrument to assess anxiety and depression in patients with physical ill health or physical symptoms and is used both in clinical practice and in research. The two subscales of the HADS—HADS-D (depression) and HADS-A (anxiety)—each consist of seven items that ask in a targeted manner about psychological anxiety and symptoms of depression in order to avoid any confounding as a result of somatic comorbidity. The items are answered on a scale between 0 = “only occasionally/never” and 3 = “much of the time.” The severity of symptoms can be quantified using the total score per scale, which ranges from 0–21. Higher scores indicate more severe depression or anxiety, and the cut-off is > 11 points per scale (e15, e16). The total sum score can be used as a measure of general mental health impairment. The HADS scales have good internal reliability with values for Cronbach’s α = 0.76 (HADS-D) and α = 0.80 (HADS-A) (e17).

Health Survey: SF-36

The SF-36 is a measurement instrument to survey health-related quality of life and covers the basic dimensions of physical and mental health with a total of 36 items (e18, e19). The questionnaire was developed as a disease-specific measurement tool and makes a distinction between: physical ability to function, physical ability to fulfill usual roles, physical pain, general health perceptions, vitality, ability to function socially, emotional ability to fulfill usual roles, and psychological well-being. The total score can be between 0 and 100, with higher scores indicating greater health-related quality of life. With a Cronbach’s α between 0.76 and 0.90 (e20), internal consistency is deemed to be good.

Structured clinical interview for DSM-IV: SCID-I

The SCID (e21, e22) is a structured standardized clinical interview to diagnose mental disorders according to DSM-IV (e23) and is considered the gold standard for the reliable determination of mental disorders. SCID-I is used to identify mental disorders (DSM-IV Axis-I Disorders). The responses given by subjects to the predefined questions are coded by the interviewing clinician as either met or not met; this prompts the clinician to ask further questions in order to make a differentiated diagnosis. SCID-I is made up of 10 modules covering different categories of mental disorders, whereby these modules can also be used independently of one other. For the present study, patients were interviewed solely regarding affective disorders (major depression, bipolar affective disorder, dysthymic disorder [Section A], and adjustment disorder [Section I]). The inter-rater reliability of SCID-I for affective disorders, anxiety disorders, schizophrenia, and alcohol abuse is over r = 0.70 (e24). SCID-I takes approximately 30–90 min to administer.

Dementia detection: DemTect

Der DemTect (e25) is an economical five-item screening instrument for use in clinical practice to detect early-stage dementia. The test is administered in the form of an interview, as part of which subjects are set five different tasks relating to the following areas of ability: attention, word recall, word fluency, and cognitive flexibility. Raw scores are transformed using two age-standardized conversion tables (under- and over-60-year-olds). The resulting total score ranges between 0 and 18. Scores between 13 and 18 points indicate appropriate cognitive performance, while scores between 9 and 12 point to mild cognitive impairment. Scores < 8 prompt suspicion of dementia. The results should be independent not only of age-related decline in cognitive function but also of educational level. The DemTect showed good sensitivity, inter-rater reliability (r = 0.99), and validity in a number of studies (e26, e27).

eMethods section 3

Sample size and power, as well as recruitment, randomization, and blinding

Sample size and power

To calculate the sample size, we assumed that a 0.6% greater reduction in HbA1c as a result of psy-PAD compared to the control group should be revealed by the t-test at 6-month follow-up (t₁) with a power of 80% at a two-tailed 5% significance level. A standard deviation of ± 1.3% was assumed for the difference value t₀–t₁ in the intervention and control groups. As such, the required sample size was N = 150 patients (n = 75 per group). Power was calculated using PROC POWER in SAS Version 9.4. Given an expected drop-out rate of 25% (by t₁), a total of N = 200 patients were to be recruited.

Overall, the assumptions of the planned analyses were well fulfilled, meaning that the planned power at t₁ was met. Although the total number of patients recruited overall was lower (n = 178) than the planned N = 200, the drop-out rate of 19.7% was lower than expected (25%). With a case number of n = 143 evaluable subjects in the study population (per protocol), this was only slightly smaller than the number calculated using power analysis (N= 150).

Recruitment, randomization, and blinding

Recruitment was carried out in two stages, firstly as:

Measures by physicians and diabetes advisors at the respective specialized practices to motivate patients, including consent for data transfer to the trial coordinating center (fulfillment of inclusion and exclusion criteria as well as contact data), followed by

Renewed summoning of patients by psychotherapists in the psychosomatic outpatient department for detailed patient information. Following informed written consent, stratified randomization took place, with patients being individually assigned per specialized practice to one of the two study conditions (psy-PAD versus optimized standard care) by the clinical trial center. To this end, and depending on the a priori targeted case numbers per specialized practice, a 50:50 distribution to the intervention and control conditions was made on paper and placed, concealed, in an urn (one urn per specialized practice). If the recruitment rate was higher than assumed beforehand, the procedure was repeated. At the trial coordinating center, a random draw from the urn of the respective specialized practice was then performed. Assignment results—intervention or control condition—were then reported back to the patients and the respective specialized practices. Assignment to the intervention or control condition was set in motion by the trial coordinating center. As such, following assignment to the intervention or control condition, group assignment based on type of intervention was disclosed, meaning that patients, practitioners, specialized practices, and the trial coordinating center became unblinded.

Table 2. Mental health at t0.

	Total (N = 143)	IG (n = 69)	CG (n = 74)
PAID M (SD)	40.27 (18.62)	41.59 (18.23)	39.04 (19.02)
SCID-I: mood disorder n (%) – Currently – In previous history	67 (46.85) 64 (44.76)	31 (44.93) 33 (47.83)	36 (48.65) 31 (41.89)
SCID-I: major depression n (%) – Currently – In previous history	34 (23.78) 63 (44.06)	18 (26.09) 32 (46.38)	16 (21.62) 31 (41.89)
HADS: depression M (SD)	7.22 (4.38)	7.38 (4.55)	7.08 (4.25)
HADS: anxiety M (SD)	7.29 (3.81)	7.32 (3.94)	7.27 (3.71)
SF-36 M (SD) – Mental health – Physical health	38.28 (11.70) 44.38 (10.27)	38.18 (12.59) 44.30 (9.28)	38.37 (10.89) 44.45 (11.18)
PHQ-9: depressive symptoms n (%) – Minimal (1–4 points) – Mild (5–9 points) – Moderate (10–14 points) – Severe (15–27 points)	21 (14.69) 47 (32.87) 48 (33.57) 27 (18.88)	11 (15.94) 22 (31.88) 23 (33.33) 13 (18.84)	10 (13.51) 25 (33.78) 25 (33.78) 14 (18.92)
PHQ-D: panic disorder n (%) – Yes – No	9 (6.29) 134 (93.71)	8 (11.59) 61 (88.41)	1 (1.35) 73 (98.65)
Contact with a psychotherapist in the preceding 6 months n (%) – Yes	22 (15.4)	10 (14.5)	12 (16.2)

IG, intervention group; CG, control group; M, mean; SD, standard deviations