DIABETES/METABOLISM RESEARCH AND REVIEWS RESEARCH ARTICLE Diabetes Metab Res Rev 2015; 31: 515–529. Published online 6 April 2015 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/dmrr.2634
Lower-limb amputation following foot ulcers in patients with diabetes: classification systems, external validation and comparative analysis
Matilde Monteiro-Soares1* Daniela Martins-Mendes2,3,4 António Vaz-Carneiro1,5,6 Mário Dinis-Ribeiro1
Abstract Background This study aimed to validate and compare the existing systems developed to stratify subjects with diabetic foot ulcers by risk of consequent lower extremity amputation.
1
CIDES/CINTESIS, Health Information and Decision Sciences Department, Faculty of Medicine, University of Porto, Porto, Portugal
2
Internal Medicine Department, Centro Hospitalar de Vila Nova de Gaia/Espinho EPE, Vila Nova de Gaia, Portugal
Methods We conducted a prospective cohort study on a consecutive series of patients (mean age of 68 years; 64% male) with active ulcer who were attending our Hospital Diabetic Foot Clinic (n = 293) from January 2010 to March 2013. At baseline, we collected information on the participants’ characteristics and the relevant variables. Afterwards, we assessed the predictive value of each variable and each system’s prognostic accuracy for amputation occurrence.
3
Diabetic Foot Clinic, Endocrinology, Diabetes and Metabolism Department, Centro Hospitalar de Vila Nova de Gaia/Espinho EPE, Vila Nova de Gaia, Portugal
4
Department of Biochemistry, Faculty of Medicine, University of Porto, Porto, Portugal
5
Centre for Evidence-Based Medicine, Faculty of Medicine, University of Lisbon, Lisbon, Portugal
6
Portuguese Collaborating Center of the Iberoamerican Cochrane NetworkLisbon, Portugal * Correspondence to: Matilde Monteiro-Soares, Departamento de Ciências da Informação e da Decisão em Saúde; Faculdade de Medicina da Universidade do Porto (CIM - FMUP), Universidade do Porto, Rua Dr. Plácido da Costa, s/n; 4200-450 Porto; Portugal. E-mail: mat.monteirosoares@gmail. com Received: 30 May 2014 Accepted: 13 December 2014
Copyright © 2014 John Wiley & Sons, Ltd.
Results During a median follow-up of 91 days (interquartile range of 98), ulcers healed in 62% of the subjects. Major amputation occurred in 7% and minor occurred in 17%. Previous ulcer or amputation, ulcer area, and gangrene were associated with amputation occurrence. Nephropathy, pulses number, ulcer aetiology, depth, and number were associated with risk of amputation. Systems typically presented sensitivity values ≥80% and negative likelihood ratios ≤0.5 for the highest risk group; area under the receiver operating characteristic curve ranged from 0.56 to 0.83 and positive likelihood ratios from 1.0 to 5.9. If one chose only major amputation as an outcome, positive predictive values were lower, and negative predictive values tended to be higher. Conclusions System stages, grades, scores, and/or prognostics were generally associated with amputation, presenting overall substantial accuracy values. Nevertheless, great improvement is possible. A multicentre study validating and refining the existing systems is needed to improve clinical decision-making in this area. Copyright © 2014 John Wiley & Sons, Ltd. Keywords clinical prediction rules; diabetic foot; diagnostic accuracy; foot ulcer; classification systems
Abbreviations ABI, ankle–brachial index; ADA, American Diabetes Association; AUC, area under the receiver operating characteristic curve; CHS, Curative Health Services wound grade scale; CI, confidence interval; DEPA, Depth of the ulcer, Extent of bacterial colonization, Phase of ulcer and Association aetiology classification system; DFU, Foot Ulcer in subjects with Diabetes; DPN, diabetic peripheral neuropathy; DUSS, diabetic ulcer severity score; IDSA, Infectious Disease Society of America; IWGDF, International
M. Monteiro-Soares et al.
516
Working Group on Diabetic Foot; LEA, lower-extremity amputation; LR, likelihood ratio; mmHg, millimetres of mercury; NPV, negative predictive value; OR, odds ratio; PAD, peripheral arterial disease; PEDIS, Perfusion, Extent, Depth/tissue loss, Infection, Sensation classification system; PPV, positive predictive value; S (AD)SAD, Size (Area, Depth), Sepsis, Arteriopathy, Denervation system; SEWSS, Saint Elian Wound Score System; SIGN, Scottish Intercollegiate Guidelines Network classification; SINBAD, Site, Ischemia, Neuropathy, Bacterial infection, and Depth; STARD, Standards for the Reporting of Diagnostic Accuracy studies; SWM, Semmes–Weinstein Monofilament; TUC, Texas University Classification; UT, University of Texas
Introduction Foot ulcers in subjects with diabetes DFU frequently result in LEA [1] and increase death risk [2]. They lead to considerable costs in terms of disability, loss of productivity and quality of life [3,4]. Therefore, we should identify the most effective ways to reduce the morbi-mortality related to foot complications in subjects with diabetes. A Eurodiale study reported that referral was delayed in more than one quarter of patients with an infected or necrotic DFU, mainly because of underestimating DFU severity and poor ischemia detection [5]. On the other hand, some authors report that critical patients’ access to specialized care is delayed, because clinics are overbooked with less urgent patients [6]. A systematic and standardized prognostic assessment of subjects with an active DFU is vital for various aspects of clinical care, namely, adequate resource allocation, treatment planning and evaluating its effectiveness, inter-professionals’ communication and quality of practice auditing [7]. There are 15 different systems that can be used to stratify patients with diabetes and active foot ulcer by their risk of LEA. None has been selected for widespread use, as their evidence level is low. A systematic review concluded that, currently, validation studies are scarce, prognostic accuracy measures are poorly or not at all described, and both overall and major LEA rates are inconsistent [8]. Therefore, we have conducted this study to (1) externally and simultaneously validate, (2) compare all the available systems’ accuracy for predicting LEA occurrence in subjects with diabetes and active foot ulcer and (3) discuss the systems’ ease of use and the pertinence of the composing variables. Also, some experts consider that the systems’ ability may be different in what concerns LEA at any level or only Copyright © 2014 John Wiley & Sons, Ltd.
major LEA prediction. Therefore, we have conducted a subgroup analysis using the last as an outcome. Afterwards, we evaluated if the system’s accuracy or the associated predictive variables changed.
Methods Type of study and selection of participants We conducted a prospective cohort study, consecutively including all subjects with diabetes and active foot ulcer attending our Diabetic Foot Outpatient Clinic, at a northern Portuguese Public Hospital, from January 2010 to March 2013. Subjects with post-LEA wounds (n = 17), decubitus ulcers (n = 14) or those that were discharged or hospitalized in the first appointment (n = 137) were excluded.
Data collection At baseline (the first appointment), characterization variables and all those included in the available stratification systems were collected through a structured interview and foot examination, performed by one podiatrist (Matilde Monteiro-Soares). At the end of data collection, all systems’ classifications and/or score were applied. All systems available for LEA prediction in subjects with diabetes and active foot ulcer were applied. They were retrieved through a systematic review previously conducted by our group [8]. The systems found were the following: (1) CHS, (2) DEPA scoring system, (3) DUSS, (4) IDSAIWGDF classification, (5) Margolis et al. classification, (6) Meggitt–Wagner classification, (7) SEWSS, (8) SIGN, (9) SINBAD score, (10) TUC and (11) Van Acker–Peter classification [4,9–19]. We excluded the Lipsky et al. system [9], because it was derived only for hospitalized patients. PEDIS classification [7] does not allow score or overall risk stratification as it was created to permit audits between centres and therefore was also excluded. S(AD) SAD was excluded because it was modified by the authors into the SINBAD score [10]. Margolis et al. [11] proposed several models for nonhealing DFU prediction. We selected the one designated as count model (that uses dichotomous variables) for the following reasons: it was easier to apply and did not present a statistically different AUC value in the derivation/internal validation study. All the subjects’ characterization variables (Table 1) were retrieved by clinical questionnaire and confirmed with the patients’ medical file. Physical impairment was Diabetes Metab Res Rev 2015; 31: 515–529. DOI: 10.1002/dmrr
Non-LEA (n = 225) LEA (n = 68)
p value
Copyright © 2014 John Wiley & Sons, Ltd.
24 (9) 157 (57) 181 (66) 146 (53) 60 (22) 221 (81) 40 (15) 56 (20) 132 (48) 38 (14) 104 (38) 0.8 (0.3) 200 (76) 163 (63) 261 (95)
1.87 (0.79–4.40) 1.48 (0.84–2.62) 1.72 (0.93–3.17) 1.27 (0.73–2.20) 1.99 (1.09–3.61) 1.73 (0.80–3.74) 0.75 (0.33–1.71) 0.92 (0.46–1.83) 0.30 (0.19–0.46) 0.31 (0.09–1.08) 1.20 (0.60–2.39) 1.34 (0.73–2.46) 1.94 (0.47–7.96)
1.06 (0.95–1.18)
3.52 (2.34–5.29)
1.01 (1.00–1.02)
3.16 (1.89–5.29)
1.79 (1.04–3.09) 2.11 (1.11–4.00)
105 (38) 87 (32) 82 (30) 1.0 (23.9)
37 (14) 106 (39) 131 (48) 4.7 (282.5)
171 (62) 3 (1) 20 (7) 80 (29) 118 (43) 49 (18)
135 (49) 27.1 (4.6)
1.22 (0.71–2.10) 1.00 (0.94–1.07)
0.96 (0.79–1.18)
67.2 (11.8) 177 (65) 269 (98) 17.8 (10.7)
Non-major LEA (n = 274)
1.00 (0.98–1.02) 1.07 (0.59–1.82) 1.00 (0.14–11.04) 1.00 (0.98–1.03)
OR (95% CI)
Univariate analysis
Subject characterization Age [mean (SD)] 67.6 (11.7) 67.5 (11.6) 67.8 (12.4) 0.8 a Male gender [n (%)] 188 (64) 144 (64) 44 (65) 1.0 b Type 2 diabetes [n (%)] 288 (98) 221 (98) 67 (99) 1.0 b Diabetes duration 18.1 (10.9) 18.0 (11.0) 18.7 (10.8) 0.6 a (in years) [mean (SD)] Insulin use [n (%)] 144 (49) 108 (48) 36 (53) 0.5b Body mass index 27.1 (4.6) 27.1 (4.7) 27.2 (4.5) 1.0 a [mean (SD)] Lives alone [n (%)] 26 (9) 17 (8) 9 (13) 0.1 a Physical impairment [n (%)] 173 (59) 128 (57) 45 (66) 0.2b Visual impairment [n (%)] 194 (66) 143 (64) 51 (75) 0.1b Retinopathy [n (%)] 159 (54) 119 (53) 40 (59) 0.4 b Nephropathy [n (%)] 69 (24) 46 (20) 21 (31) 0.02 b Hypertension [n (%)] 237 (81) 178 (79) 59 (87) 0.2 b Coronary heart 42 (14) 34 (15) 8 (12) 0.6b disease [n (%)] Stroke [n (%)] 59 (20) 46 (20) 13 (19) 0.9b DFU foot characterization Total foot pulses [n (%)] 0 149 (51) 93 (41) 56 (82)
Lower-limb amputation following foot ulcers in patients with diabetes: classification systems, external validation and comparative analysis
Matilde Monteiro-Soares1* Daniela Martins-Mendes2,3,4 António Vaz-Carneiro1,5,6 Mário Dinis-Ribeiro1
Abstract Background This study aimed to validate and compare the existing systems developed to stratify subjects with diabetic foot ulcers by risk of consequent lower extremity amputation.
1
CIDES/CINTESIS, Health Information and Decision Sciences Department, Faculty of Medicine, University of Porto, Porto, Portugal
2
Internal Medicine Department, Centro Hospitalar de Vila Nova de Gaia/Espinho EPE, Vila Nova de Gaia, Portugal
Methods We conducted a prospective cohort study on a consecutive series of patients (mean age of 68 years; 64% male) with active ulcer who were attending our Hospital Diabetic Foot Clinic (n = 293) from January 2010 to March 2013. At baseline, we collected information on the participants’ characteristics and the relevant variables. Afterwards, we assessed the predictive value of each variable and each system’s prognostic accuracy for amputation occurrence.
3
Diabetic Foot Clinic, Endocrinology, Diabetes and Metabolism Department, Centro Hospitalar de Vila Nova de Gaia/Espinho EPE, Vila Nova de Gaia, Portugal
4
Department of Biochemistry, Faculty of Medicine, University of Porto, Porto, Portugal
5
Centre for Evidence-Based Medicine, Faculty of Medicine, University of Lisbon, Lisbon, Portugal
6
Portuguese Collaborating Center of the Iberoamerican Cochrane NetworkLisbon, Portugal * Correspondence to: Matilde Monteiro-Soares, Departamento de Ciências da Informação e da Decisão em Saúde; Faculdade de Medicina da Universidade do Porto (CIM - FMUP), Universidade do Porto, Rua Dr. Plácido da Costa, s/n; 4200-450 Porto; Portugal. E-mail: mat.monteirosoares@gmail. com Received: 30 May 2014 Accepted: 13 December 2014
Copyright © 2014 John Wiley & Sons, Ltd.
Results During a median follow-up of 91 days (interquartile range of 98), ulcers healed in 62% of the subjects. Major amputation occurred in 7% and minor occurred in 17%. Previous ulcer or amputation, ulcer area, and gangrene were associated with amputation occurrence. Nephropathy, pulses number, ulcer aetiology, depth, and number were associated with risk of amputation. Systems typically presented sensitivity values ≥80% and negative likelihood ratios ≤0.5 for the highest risk group; area under the receiver operating characteristic curve ranged from 0.56 to 0.83 and positive likelihood ratios from 1.0 to 5.9. If one chose only major amputation as an outcome, positive predictive values were lower, and negative predictive values tended to be higher. Conclusions System stages, grades, scores, and/or prognostics were generally associated with amputation, presenting overall substantial accuracy values. Nevertheless, great improvement is possible. A multicentre study validating and refining the existing systems is needed to improve clinical decision-making in this area. Copyright © 2014 John Wiley & Sons, Ltd. Keywords clinical prediction rules; diabetic foot; diagnostic accuracy; foot ulcer; classification systems
Abbreviations ABI, ankle–brachial index; ADA, American Diabetes Association; AUC, area under the receiver operating characteristic curve; CHS, Curative Health Services wound grade scale; CI, confidence interval; DEPA, Depth of the ulcer, Extent of bacterial colonization, Phase of ulcer and Association aetiology classification system; DFU, Foot Ulcer in subjects with Diabetes; DPN, diabetic peripheral neuropathy; DUSS, diabetic ulcer severity score; IDSA, Infectious Disease Society of America; IWGDF, International
M. Monteiro-Soares et al.
516
Working Group on Diabetic Foot; LEA, lower-extremity amputation; LR, likelihood ratio; mmHg, millimetres of mercury; NPV, negative predictive value; OR, odds ratio; PAD, peripheral arterial disease; PEDIS, Perfusion, Extent, Depth/tissue loss, Infection, Sensation classification system; PPV, positive predictive value; S (AD)SAD, Size (Area, Depth), Sepsis, Arteriopathy, Denervation system; SEWSS, Saint Elian Wound Score System; SIGN, Scottish Intercollegiate Guidelines Network classification; SINBAD, Site, Ischemia, Neuropathy, Bacterial infection, and Depth; STARD, Standards for the Reporting of Diagnostic Accuracy studies; SWM, Semmes–Weinstein Monofilament; TUC, Texas University Classification; UT, University of Texas
Introduction Foot ulcers in subjects with diabetes DFU frequently result in LEA [1] and increase death risk [2]. They lead to considerable costs in terms of disability, loss of productivity and quality of life [3,4]. Therefore, we should identify the most effective ways to reduce the morbi-mortality related to foot complications in subjects with diabetes. A Eurodiale study reported that referral was delayed in more than one quarter of patients with an infected or necrotic DFU, mainly because of underestimating DFU severity and poor ischemia detection [5]. On the other hand, some authors report that critical patients’ access to specialized care is delayed, because clinics are overbooked with less urgent patients [6]. A systematic and standardized prognostic assessment of subjects with an active DFU is vital for various aspects of clinical care, namely, adequate resource allocation, treatment planning and evaluating its effectiveness, inter-professionals’ communication and quality of practice auditing [7]. There are 15 different systems that can be used to stratify patients with diabetes and active foot ulcer by their risk of LEA. None has been selected for widespread use, as their evidence level is low. A systematic review concluded that, currently, validation studies are scarce, prognostic accuracy measures are poorly or not at all described, and both overall and major LEA rates are inconsistent [8]. Therefore, we have conducted this study to (1) externally and simultaneously validate, (2) compare all the available systems’ accuracy for predicting LEA occurrence in subjects with diabetes and active foot ulcer and (3) discuss the systems’ ease of use and the pertinence of the composing variables. Also, some experts consider that the systems’ ability may be different in what concerns LEA at any level or only Copyright © 2014 John Wiley & Sons, Ltd.
major LEA prediction. Therefore, we have conducted a subgroup analysis using the last as an outcome. Afterwards, we evaluated if the system’s accuracy or the associated predictive variables changed.
Methods Type of study and selection of participants We conducted a prospective cohort study, consecutively including all subjects with diabetes and active foot ulcer attending our Diabetic Foot Outpatient Clinic, at a northern Portuguese Public Hospital, from January 2010 to March 2013. Subjects with post-LEA wounds (n = 17), decubitus ulcers (n = 14) or those that were discharged or hospitalized in the first appointment (n = 137) were excluded.
Data collection At baseline (the first appointment), characterization variables and all those included in the available stratification systems were collected through a structured interview and foot examination, performed by one podiatrist (Matilde Monteiro-Soares). At the end of data collection, all systems’ classifications and/or score were applied. All systems available for LEA prediction in subjects with diabetes and active foot ulcer were applied. They were retrieved through a systematic review previously conducted by our group [8]. The systems found were the following: (1) CHS, (2) DEPA scoring system, (3) DUSS, (4) IDSAIWGDF classification, (5) Margolis et al. classification, (6) Meggitt–Wagner classification, (7) SEWSS, (8) SIGN, (9) SINBAD score, (10) TUC and (11) Van Acker–Peter classification [4,9–19]. We excluded the Lipsky et al. system [9], because it was derived only for hospitalized patients. PEDIS classification [7] does not allow score or overall risk stratification as it was created to permit audits between centres and therefore was also excluded. S(AD) SAD was excluded because it was modified by the authors into the SINBAD score [10]. Margolis et al. [11] proposed several models for nonhealing DFU prediction. We selected the one designated as count model (that uses dichotomous variables) for the following reasons: it was easier to apply and did not present a statistically different AUC value in the derivation/internal validation study. All the subjects’ characterization variables (Table 1) were retrieved by clinical questionnaire and confirmed with the patients’ medical file. Physical impairment was Diabetes Metab Res Rev 2015; 31: 515–529. DOI: 10.1002/dmrr
Non-LEA (n = 225) LEA (n = 68)
p value
Copyright © 2014 John Wiley & Sons, Ltd.
24 (9) 157 (57) 181 (66) 146 (53) 60 (22) 221 (81) 40 (15) 56 (20) 132 (48) 38 (14) 104 (38) 0.8 (0.3) 200 (76) 163 (63) 261 (95)
1.87 (0.79–4.40) 1.48 (0.84–2.62) 1.72 (0.93–3.17) 1.27 (0.73–2.20) 1.99 (1.09–3.61) 1.73 (0.80–3.74) 0.75 (0.33–1.71) 0.92 (0.46–1.83) 0.30 (0.19–0.46) 0.31 (0.09–1.08) 1.20 (0.60–2.39) 1.34 (0.73–2.46) 1.94 (0.47–7.96)
1.06 (0.95–1.18)
3.52 (2.34–5.29)
1.01 (1.00–1.02)
3.16 (1.89–5.29)
1.79 (1.04–3.09) 2.11 (1.11–4.00)
105 (38) 87 (32) 82 (30) 1.0 (23.9)
37 (14) 106 (39) 131 (48) 4.7 (282.5)
171 (62) 3 (1) 20 (7) 80 (29) 118 (43) 49 (18)
135 (49) 27.1 (4.6)
1.22 (0.71–2.10) 1.00 (0.94–1.07)
0.96 (0.79–1.18)
67.2 (11.8) 177 (65) 269 (98) 17.8 (10.7)
Non-major LEA (n = 274)
1.00 (0.98–1.02) 1.07 (0.59–1.82) 1.00 (0.14–11.04) 1.00 (0.98–1.03)
OR (95% CI)
Univariate analysis
Subject characterization Age [mean (SD)] 67.6 (11.7) 67.5 (11.6) 67.8 (12.4) 0.8 a Male gender [n (%)] 188 (64) 144 (64) 44 (65) 1.0 b Type 2 diabetes [n (%)] 288 (98) 221 (98) 67 (99) 1.0 b Diabetes duration 18.1 (10.9) 18.0 (11.0) 18.7 (10.8) 0.6 a (in years) [mean (SD)] Insulin use [n (%)] 144 (49) 108 (48) 36 (53) 0.5b Body mass index 27.1 (4.6) 27.1 (4.7) 27.2 (4.5) 1.0 a [mean (SD)] Lives alone [n (%)] 26 (9) 17 (8) 9 (13) 0.1 a Physical impairment [n (%)] 173 (59) 128 (57) 45 (66) 0.2b Visual impairment [n (%)] 194 (66) 143 (64) 51 (75) 0.1b Retinopathy [n (%)] 159 (54) 119 (53) 40 (59) 0.4 b Nephropathy [n (%)] 69 (24) 46 (20) 21 (31) 0.02 b Hypertension [n (%)] 237 (81) 178 (79) 59 (87) 0.2 b Coronary heart 42 (14) 34 (15) 8 (12) 0.6b disease [n (%)] Stroke [n (%)] 59 (20) 46 (20) 13 (19) 0.9b DFU foot characterization Total foot pulses [n (%)] 0 149 (51) 93 (41) 56 (82)
