Treating Venous Leg Ulcers in Primary Care

The cluster-randomized Ulcus Cruris Care trial

Background: Venous leg ulcers (VLUs) markedly restrict patients’ everyday activities and impair their quality of life. Compression is the most important treatment measure but is often not consistently implemented in practice. The aim of the Ulcus Cruris Care (UCC) project was to develop a primary care approach for the systematic support of evidence-based, patient-centered treatment.

Methods: UCC consists of online training for care teams, software-supported case management, and patient training. Block randomization (1:1) was performed at the level of individual primary care practices. The practices recruited patients with UCV. Each patient was followed up for 12 months. The primary endpoint was wound healing, as confirmed by two independent blinded investigators. The secondary endpoints were guideline-compliant compression, health-related quality of life (EQ-5D-5L), depression (PHQ-9), and case-related treatment costs.

Results: 59 patients were included. For the primary endpoint of time to wound healing, the intervention yielded a hazard ratio of 2.71 [0.66; 11.06] (p = 0.16). The odds ratio for complete wound healing was 3.60 [0.97; 13.36] (p = 0.049). Guideline-compliant compression was more common in the intervention group (72% versus 33%; p = 0.010). Depression (5.3 ± 6.1 versus 10.0 ± 6.9; p = 0.019) and outpatient treatment costs (€1457 ± 917 versus €3153 ± 1843; p = 0.016) were lower in the intervention group.

Conclusion: This multicenter RCT suggests a potential benefit of the UCC intervention but does not confirm its efficacy. The UCC approach can specifically target deficits in outpatient care for chronic wounds at the primary care level.

Cite this as: Fleischhauer T, Sander N, Feißt M, Awounvo S, Weller L, Poß-Doering R, Laux G, Altiner A, Müller-Bühl U, Szecsenyi J, Senft JD: Treating venous leg ulcers in primary care: The cluster-randomized Ulcus Cruris Care trial. Dtsch Arztebl Int 2026; 123: 9–14. DOI: 10.3238/arztebl.m2025.0207

Venous leg ulcers (VLU) constitute a major medical problem, placing a severe burden of disease on those affected, and the cost of treating them is a relevant factor in health economics (1, 2, 3, 4). The prevalence of VLU in Germany is estimated at 0.1–1.0% and increases steeply with age (5, 6, 7, 8). We believe that the treatment given by primary care hysicians is heterogeneous in the extreme, ranging from simple prescription of wound dressings to active all-round care. Guideline-compliant management often fails owing to limited time and personnel and the lack of standardized structures.

In particular, compression therapy, a causally oriented form of treatment, is often not applied properly. Performed as specified in the guidelines, compression therapy combats venous hypertension and is—as one of the few treatment options with level 1A evidence—highly effective with regard to wound healing (9, 10, 11). Nevertheless, healthcare research has shown that it is used in fewer than 50% of cases and is often applied incorrectly (2, 12, 13). Other effective elements of treatment, e.g., mobilization and physical exercises, are addressed only inadequately (14). Emphasis is frequently placed on local treatment and the use of modern, cost-intensive means of wound management, although the benefit of these methods in VLU has not yet been demonstrated objectively (15). It would be more productive to focus on guideline-compliant causal treatment, accompanied by local wound treatment appropriate to the given stage. Patients often have a passive role in the process: changing of the dressing is seen as the central treatment measure, with self-management, an important resource in wound treatment, largely remaining unexploited.

The project “Ulcus Cruris Care” (UCC) was established with the intention of developing an evidence-based, patient-oriented strategy to improve the treatment of VLU in the primary care setting. The aims were to support general practice teams in their role as primary care providers; to train patients and their relatives, thus enabling them to to take an active part; and to integrate modern IT resources (e-learning, software-based case management) into the treatment process. The study hypothesis was that implementation of the intervention strategy would be conducive to evidence-based treatment—particularly compression therapy—enabling faster wound healing in patients with VLU (primary endpoint). Moreover, positive effects on health-related quality of life and decreased costs were anticipated (secondary endpoints).

Methods

The UCC intervention having been developed in a three-stage establishment and evaluation phase (16, 17, 18, 19, 20), we set out to test it in a multicenter cluster-randomized trial. A detailed description of the methods was published in 2022 (18), and the study was conducted from the last quarter of 2021 to the first quarter of 2024. The presentation of the results follows the Consolidated Standards of Reporting Trials (CONSORT) recommendations (21). A completed CONSORT checklist is provided in eSupplement 2.

Intervention

The UCC intervention strategy comprises three main elements:

Online training for primary care teams

The participating primary care physicians and medical assistants were trained together in guideline-compliant diagnosis and treatment of VLU. The training comprised a live online webinar (90 minutes, Cisco Webex) and supplementary e-learning courses (https://welearn.academy/). The goal was to enhance their knowledge of evidence-based treatment and patient information and thus promote patients’ active participation in the treatment process.

Software-based case management

The case management software CareCockpit (22), already being used by the primary care teams, was experimentally extended to provide systematic support for routine wound treatment. To ensure treatment standardization, the teams received structured algorithms (standard operating procedures, SOPs).

Standardized patient information

A standardized information process should improve the knowledge level of patients and and their relatives. At the first study visit following trial inclusion, the treating primary care team members verbally explained the procedures and gave the patients access to readily understandable information materials and e-learning programs. Open questions and the patients’ information status were regularly assessed and documented as part of case management.

Recruitment and data acquisition

The recruitment phase ran from the first quarter of 2022 to the last quarter of 2022. A total of 965 primary care practices located in the catchment area of Heidelberg University Hospital and engaged in teaching and research were contacted in writing. For a practice to fulfill the inclusion criteria, its medical assistants had to be involved in the care of chronic wounds and have access to the electronic data processing infrastructure necessary for use of the CareCockpit software.

Block randomization was carried out at practice level in a 1:1 ratio with variable block length by the Institute of Medical Biometry. Screened patients were included after giving their informed consent following explanation by a participating primary care physician. To be eligible, patients had to be at least 18 years old and diagnosed with VLU within the previous 12 months. The exclusion criteria were ulceration of other origin, relevant peripheral arterial occlusive disease (PAOD; ankle-brachial index [ABI] < 0.5 or ankle arterial pressure < 60 mm Hg), decompensated heart failure (NYHA III/IV), and inability to give consent.

The primary outcome measure was the time to complete wound healing. Wound healing status was assessed continuously by the study physicians during the course of treatment, and when a wound was reported as fully healed it was assessed from photographs by two independent, blinded investigators, who had to agree that healing was indeed complete.

The following secondary outcome measures were assessed at study inclusion (T₀), after 3 months (T₁), and after 12 months (T₂):

• Guideline-compliant prescription of compression therapy (T₂)
• Health-related quality of life, measured using EQ-5D-5L (T₀, T₁, T₂) (23)
• Depression, using PHQ-9 (T₀, T₁, T₂) (24)
• Patient satisfaction and information, using the PACIC-5A questionnaire (T₀, T₁, T₂) (25).

Patients’ adherence to treatment was assessed by the evaluating physicians using a five-point Likert scale. Compression therapy was deemed to conform with the guidelines when:

• It was applied without interruption during wound healing and a phlebological compression bandage, or another compression agent/device of at least class III, was used.
• There was no contraindication and no patient preference for another form of treatment after assessment of comorbidities.

This definition was chosen because in the presence of florid VLU, particularly in the decongestion phase with high pressure (40 mm Hg), better results can be achieved, providing there are no contraindications and the treatment is tolerated by the patients (11).

For health economic analysis, the following parameters were measured at T₂:

• The frequency of the service coded 02312 in the EBM catalog, the German fee schedule for outpatient treatment
• The number of physician contacts for wound treatment
• Prescription periods for treatment by social carers
• Type and number of prescriptions for auxiliary treatments, auxiliary devices, and wound care products, together with the duration of their use.

Furthermore, the following variables were documented by means of a patient questionnaire at T₂:

• Frequency of wound swabs
• Number and duration of stays in hospital
• Number of contacts with medical specialists
• Number of days off work due to VLU
• Patients’ travel (means of transport, time, costs).

The participating primary care practices transmitted the data to the study center in pseudonymized form using an electronic data sheet. Data collection and processing took place only after informed consent by the participants and in accordance with the General Data Protection Regulation (GDPR).

Statistical analysis

Case number planning is based on the primary endpoint, time to complete wound healing. According to the literature, a healing rate of 60% (control group) can be anticipated after 12 weeks of standard care (26). The expected healing rate at the same time in the intervention group is 80%—oriented on clinical trials of patient-centered interventions (27). With power of 80%, significance level of 0.05, an assumed drop-out rate of 20%, and the assumption of exponentially distributed treatment curves, the planned number of patients per treatment group is n = 55 using a log-rank test in classical two-group comparison. To take account of data clustering (patients at physicians’ practices), the case number per group was adjusted by a design effect of 1.1, calculated from a cluster size of n = 3 patients per practice and a conservative intracluster coefficient (ICC) of 0.05. This resulted in the need to recruit a total of n = 126 patients (n = 63 per group) in n = 42 practices (n = 21 per group). The maximum observation time per patient was 12 months. The primary endpoint was analyzed by means of a Cox proportional hazards regression model (shared frailty model) with a two-sided significance level of 5%. The variables “group,” “total ulcer size,” “time of existence of VLU,” and “level of care required” were considered as fixed effects, the variable „practice“ as shared frailty term. The secondary outcome measures were described by means of appropriate tests (t-tests, chi-squared tests), point estimators (mean differences with standard deviations, difference of frequencies), and the corresponding 95% confidence intervals. In the primary endpoint analysis, p-values < 0.05 were defined as statistically significant. The statistical analysis plan can be found in eSupplement 1.

Results

The planned total of 126 patients could not be attained owing to recruitment problems in relation to the COVID-19 pandemic. Instead, 59 patients were included, recruited by 10 primary care practices in the control group (CG) and 14 practices in the intervention group (IG) (Figure 1). Four patients from the IG and two from the CG did not complete the trial; the data collected up to the time of discontinuation were included for analysis. The mean observation time was 311 ± 89 days in the IG and 322 ± 93 days in the CG. The cluster size varied from n = 1 to n = 5, with the exception of primary care practice P10 (n = 7), which, however, had three drop-outs. The detailed distribution of patients among the clusters can be seen in eTable 1. The baseline (T₀) characteristics of the patients included in the study are summarized in Table 1.

Figure 1

Flow chart of the inclusion of primary care practices and patients in the trial n, Sample size

Table 1

The patients’ characteristics at baseline

eTable 1

Cluster distribution of the patients included among the primary care practices of the control and intervention groups

Wound healing

Photographic documentation of the wounds for evaluation of healing was available for 40 patients. In all 40 cases (100%), the assessors were in agreement. Cox regression analysis of the primary endpoint, time to wound healing, showed a hazard ratio (HR) of 2.71 [0.66; 11.06] (ICC = 0.005; p = 0.16) for the intervention. Figure 2 shows the results of the Kaplan–Meier analysis of wound healing. Due to missing data, the adjustment factor “body mass index” (BMI) was replaced in the final analysis plan by the variable “level of care”. A post-hoc sensitivity analysis including BMI and with imputation of missing values can be seen in eTable 2.

Figure 2

Kaplan–Meier analysis of the probability of wound healing over the course of time in the two study arms, as determined by photographic documentation and evaluation by two blinded investigators (primary endpoint assessment).

eTable 2

Cox regression analysis (shared frailty model) of the impact of the variables on the time to wound healing*

The odds ratio for the occurrence of wound healing within 12 months was 3.60 [0.97; 13.36] in favor of the intervention group (p = 0.049). A post-hoc sensitivity analysis in which non-analyzed cases (Figure 1) were classed as not healed showed a 12-month wound healing rate of 55% (23/42) in the IG and 29% (5/17) in the CG (p = 0.077). In three IG patients and one CG patient, completed wound healing without photographic documentation was reported at T₁ and T₂ respectively. In these cases the healing was not logged in the continual registration of the primary endpoint and was therefore not included in the analysis thereof. Figure 3 shows the results of the post-hoc Kaplan–Meier analysis including all reported cases of healing. Recurrences were found in four of 38 participants in the IG and in one of 15 participants in the CG (p = 0.925).

Figure 3

Kaplan–Meier analysis of the probability of wound healing over the course of time in the two study arms, including all reported healing data (post-hoc).

Compression therapy

Guideline-compliant prescription of compression therapy in accordance with the study protocol took place in 72% of cases (26/38) in the IG, compared with 33% (5/15) in the CG (p = 0.010). In the IG 22 patients (61%) received a compression bandage, in the CG, seven patients (47%). Six participants from the IG (17%) and four from the CG (27%) were given a medical compression stocking. A multicomponent compression system was used in four IG patients (11%) and three CG patients (20%). One patient in the CG (7%) received an ulcer stocking system. No medical adaptive compression systems were prescribed.

Patient-reported endpoints

The EQ-5D-5L index at T₂ was 0.78 ± 0.23 in the IG versus 0.62 ± 0.41 in the CG (p = 0.099). Depression at T₂ was lower in the IG than in the CG (IG 5.3 ± 6.1, CG 9.9 ± 7.3; p = 0.032). Table 2 summarizes the patient-reported secondary endpoints.

Table 2

Comparison of the patient-reported secondary endpoints at T1 (3 months) and T2 (12 months)

Health economic analysis

The average case-related costs for outpatient treatment were € 1457 ± 917 in the IG versus € 3113 ± 1843 in the CG (p = 0.016). eTable 3 gives an overview of the treatment costs during the study period.

eTable 3

Treatment costs in the intervention and control groups

Discussion

Cox regression analysis of the primary endpoint, time to wound healing, revealed a HR of 2.71 [0.66; 11.06] for the intervention (p = 0.16). Therefore, the multicenter randomized controlled trial did not confirm efficacy. However, the results, particularly for the secondary outcomes, point to a possible positive effect of the UCC intervention.

The possible effect of the intervention is presumably attributable primarily to the higher rate of guideline-compliant compression therapy. The multimodal approach of UCC, which supported both training of the treating primary care teams and a wide-reaching patient-centered approach to wound treatment, proved very effective regarding the implementation of continuous compression therapy, reaching a rate of 72%. This is in contrast to the findings of the project-related routine analysis of outpatient care data, according to which compression therapy was prescribed continuously throughout the wound healing process in only 14% of cases. The higher rate in the control group than in standard care could be interpreted as the result of a Hawthorne effect, i.e., participation in the trial had a positive impact on the treatment behavior of those taking part.

The aim of the UCC intervention strategy was to support general practice teams in their role as primary care providers for patients with VLU. Core elements of the intervention were improvement of the competence of physicians and other team members providing care, standardization of treatment and patient education, and a patient-centered approach. The training intervention comprised guideline-compliant treatment, patient education, involvement of patients and relatives in decision-making processes, and encouragement of active participation. A meta-analysis conducted as part of the project confirmed the positive influence of such measures on wound healing (17). The specific promotion of patient education could thus prove an effective means of overcoming the barriers that exist to the implementation of compression therapy. The intervention also addresses interfaces with relevant medical specialties. We chose to dispense with rigid specifications, because care structures and availability of medical specialists varies from region to region and primary care providers are intimately acquainted with local conditions. The practices’ investment in terms of time and effort was limited to the online training session and the e-learning course (each 90 minutes long). Organizational aspects such as case management and patient information were not documented, but were designed to improve efficiency.

Evaluation of the patient-reported endpoints also suggested positive effects of the UCC intervention on depression and other aspects of quality of life. It seems likely that particularly improvements in terms of mobility, participation in daily life, and pain reduction—as a result of compression therapy and faster wound healing—contribute to these results. Studies have already shown the high importance of such considerations for patients with VLU (28, 29).

With regard to the costs of treatment, the analysis showed a case-related cost saving of € 1656 within 12 months in the intervention group. This decrease was due principally to reductions in the outgoings for dressings and for nursing care at home. Despite the uncertainty due to the small size of the sample, it can be assumed that the swifter wound healing as a result of consistent application of compression therapy in the intervention group led to a relevant reduction in management costs. In view of demographic changes and the anticipated increase in the number of chronic wounds, cost-efficient approaches to treatment will become even more important in the future (30).

This multicenter study was negatively affected by the COVID-19 pandemic, meaning that only around half of the originally planned number of participants could be recruited. The result was reduced power to detect a difference in the primary endpoint. The differences in the secondary endpoints support the assumption of a benefit of the intervention, but must be interpreted cautiously due to the low power and the exploratory nature of the trial. Despite randomization, at T₀ the wounds in the control group were larger and the Venous Clinical Severity Score (VCSS) was higher, possibly indicating more marked disease. The primary endpoint was adjusted for wound size; the resulting HR of 2.71 speaks in favor of the intervention. However, the higher comorbidity index in the intervention group must be considered. Finally, it has to be taken into account that the knowledge of the participating primary care teams may have extended beyond standard care; due to the randomization, however, no relevant degree of bias is to be expected.

Conclusions

The multicenter randomized controlled trial yielded no evidence to confirm efficacy of the UCC intervention, but provided consistent indications of potential benefit. The intervention strategy described here appears suited to address the deficits that exist in the outpatient management of chronic wounds in the primary care setting. Extended implementation of the UCC intervention, e.g. by means of selective contracts, would therefore be a logical and promising approach from the viewpoint of patients, care providers, and health insurance funds.

Ethics committee approval

The study protocol was approved by the Heidelberg ethics committee

(S-608/2021) and the ethics committee of the State Medical Association (B-F-2021–101).

Data sharing

The anonymized case-based patient data collected and analyzed in the course of this trial can be provided for strictly scientific purposes up to 10 years after publication on receipt of a methodologically sound and reasonable proposal for evaluation. The study protocol and statistical analysis plan can also be provided.

Funding

The project “Ulcus Cruris Care” was supported by the Innovation Committee of the Federal Joint Committee (G-BA), Berlin (01VSF19043).

Acknowledgments

The authors are grateful to all participants for their valuable contribution to the trial.

Conflict of interest statement
The authors declare that no conflict of interest exists.

Manuscript received on 15 May 2025, revised version accepted on 20 October 2025

Translated from the original German by David Roseveare

Corresponding author
PD Dr. med. Jonas D. Senft

References

1. Singer AJ, Tassiopoulos A, Kirsner RS: Evaluation and management of lower-extremity ulcers. N Engl J Med 2017; 377: 1559–67

2. Heyer K, Protz K, Glaeske G, Augustin M: Epidemiology and use of compression treatment in venous leg ulcers: nationwide claims data analysis in Germany. Int Wound J 2017; 14: 338–43

3. Müller-Bühl U, Leutgeb R, Bungartz J, Szecsenyi J, Laux G: Expenditure of chronic venous leg ulcer management in German primary care: Results from a population-based study. Int Wound J 2013; 10: 52–6

4. Augustin M, Brocatti LK, Rustenbach SJ, Schäfer I, Herberger K: Cost-of-illness of leg ulcers in the community. Int Wound J 2014; 11: 283–92

5. Kreft D, Keiler J, Grambow E, Kischkel S, Wree A, Doblhammer G: Prevalence and mortality of venous leg diseases of the deep veins: An observational cohort study based on German health claims data. Angiology 2020; 71: 452–64

6. Heyer K, Herberger K, Protz K, Glaeske G, Augustin M: Epidemiology of chronic wounds in Germany: Analysis of statutory health insurance data. Wound Repair Regen 2016; 24: 434–42

7. Laible J, Mayer H, Evers GCM: [Prevalence of ulcus cruris in home care nursing. An epidemiological study in North Rhine-Westphalia]. Pflege 2002; 15: 16–23

8. Rabe E, Pannier-Fischer F, Bromen K, et al.: Bonner Venenstudie der Deutschen Gesellschaft für Phlebologie: Epidemiologische Untersuchung zur Frage der Häufigkeit und Ausprägung von chronischen Venenkrankheiten in der städtischen und ländlichen Wohnbevölkerung. Phlebologie 2003; 32: 1–14

9. Shi C, Dumville JC, Cullum N, Connaughton E, Norman G: Compression bandages or stockings versus no compression for treating venous leg ulcers. Cochrane Database Syst Rev 2021; 7: CD013397

10. Deutsche Gesellschaft für Phlebologie und Lymphologie e.V. (eds.): S2k-Leitlinie Diagnostik und Therapie des Ulcus cruris venosum. 2025. https://register.awmf.org/de/leitlinien/detail/037-009 (last accessed on 16 October 2025).

11. Rabe E: S2k-Leitlinie: Medizinische Kompressionstherapie der Extremitäten mit Medizinischem Kompressionsstrumpf (MKS), Phlebologischem Kompressionsverband (PKV) und Medizinischen adaptiven Kompressionssystemen (MAK). www.awmf.org/service/awmf-aktuell/medizinische-kompressionstherapie-der-extremitaeten-mit-medizinischem-kompressionsstrumpf-mks-phlebologischem-kompressionsverband-pkv-und-medizinischen-adaptiven-kompressionssystemen-mak (last accessed on 16 October 2025).

12. Protz K, Heyer K, Dissemond J, et al.: Compression therapy—current practice of care: level of knowledge in patients with venous leg ulcers. J Dtsch Dermatol Ges 2016; 14: 1273–82

13. Heyer K, Protz K, Augustin M: Compression therapy—cross-sectional observational survey about knowledge and practical treatment of specialised and non-specialised nurses and therapists. Int Wound J 2017; 14: 1148–53

14. Jull A, Slark J, Parsons J: Prescribed exercise with compression vs compression alone in treating patients with venous leg ulcers: A systematic review and meta-analysis. JAMA Dermatol 2018; 154: 1304–11

15. Norman G, Westby MJ, Rithalia AD, Stubbs N, Soares MO, Dumville JC: Dressings and topical agents for treating venous leg ulcers. Cochrane Database Syst Rev 2018; 6: CD012583

16. Poß-Doering R, Anders C, Fleischhauer T, Szecsenyi J, Senft JD: Exploring healthcare provider and patient perspectives on current outpatient care of venous leg ulcers and potential interventions to improve their treatment: A mixed methods study in the ulcus cruris care project. BMC Prim Care 2022; 23: 229

17. Bossert J, Vey JA, Piskorski L, et al.: Effect of educational interventions on wound healing in patients with venous leg ulceration: A systematic review and meta-analysis. Int Wound J 2023; 20: 1784–95

18. Senft JD, Fleischhauer T, Frasch J, et al.: Primary care disease management for venous leg ulceration-study protocol for the Ulcus Cruris Care [UCC] randomized controlled trial (DRKS00026126). Trials 2022; 23: 60

19. Senft JD, Fleischhauer T, Poß-Doering R, et al.: Primary care disease management for venous leg ulceration in German healthcare: Results of the Ulcus Cruris Care Pilot Study. Healthcare (Basel) 2023; 11: 2521

20. Fleischhauer T, Poß-Doering R, Sander N, et al.: Pilot implementation of a primary care disease management concept for venous leg ulceration: Results of a mixed-methods process evaluation. Healthccare (Basel) 2024; 12: 2552

21. Hopewell S, Chan A-W, Collins GS, et al.: CONSORT 2025 statement: Updated guideline for reporting randomised trials. Lancet 2025; 405: 1633–40

22. CareCockpit – Hausarztpraxis-basiertes Case Management. www.carecockpit.org (last accessed on 4 November 2025).

23. EQ-5D. EuroQol. https://euroqol.org/ (last accessed on 4 November 2025).

24. Gräfe K, Zipfel S, Herzog W, Löwe B: Screening psychischer Störungen mit dem “Gesundheitsfragebogen für Patienten (PHQ-D)“. Diagnostica 2004; 50: 171–81

25. Rosemann T, Laux G, Droesemeyer S, Gensichen J, Szecsenyi J: Evaluation of a culturally adapted German version of the patient assessment of chronic illness care (PACIC 5A) questionnaire in a sample of osteoarthritis patients. J Eval Clin Pract 2007; 13: 806–13

26. Weller CD, Buchbinder R, Johnston RV: Interventions for helping people adhere to compression treatments for venous leg ulceration. Cochrane Database Syst Rev 2016; 3: CD008378

27. O’Brien J, Finlayson K, Kerr G, Edwards H: Evaluating the effectiveness of a self-management exercise intervention on wound healing, functional ability and health-related quality of life outcomes in adults with venous leg ulcers: a randomised controlled trial. Int Wound J 2017; 14: 130–7

28. Augustin M, Mayer A, Goepel LM, Baade K, Heyer K, Herberger K: Cumulative Life Course Impairment (CLCI): A new concept to characterize persistent patient burden in chronic wounds. Wound Med 2013; 1: 2–6

29. Herber OR, Schnepp W, Rieger MA: A systematic review on the impact of leg ulceration on patients’ quality of life. Health Qual Life Outcomes 2007; 5: 44

30. Probst S, Saini C, Gschwind G, et al.: Prevalence and incidence of venous leg ulcers—a systematic review and meta-analysis. Int Wound J 2023; 20: 3906–21