Predictors of Lower-Extremity Amputation in Patients With an Infected Diabetic Foot Ulcer

Infection is a frequent complication of diabetic foot ulcers, with up to 58% of ulcers being infected at initial presentation at a diabetic foot clinic, increasing to 82% in patients hospitalized for a diabetic foot ulcer (1). These diabetic foot infections (DFIs) are associated with poor clinical outcomes for the patient and high costs for both the patient and the health care system (2). Patients with a DFI have a 50-fold increased risk of hospitalization and 150-fold increased risk of lower-extremity amputation compared with patients with diabetes and no foot infection (3). Among patients with a DFI, ∼5% will undergo a major amputation and 20–30% a minor amputation, with the presence of peripheral arterial disease (PAD) greatly increasing amputation risk (4–6). Furthermore, lower-limb amputation is associated not only with significant morbidity and mortality but also with major psychosocial and financial consequences (7–9).

As infection of a diabetic foot wound heralds a poor outcome, early diagnosis and treatment are important. Unfortunately, systemic signs of inflammation such as fever and leukocytosis are often absent even with a serious foot infection (10,11). As local signs and symptoms of infection are also often diminished, because of concomitant peripheral neuropathy and ischemia (12), diagnosing and defining resolution of infection can be difficult. Amputation, instead of resolution of symptoms or signs of infection, could therefore be a reliable outcome measure.

The system developed by the International Working Group on the Diabetic Foot (IWGDF) and the Infectious Diseases Society of America (IDSA) provides criteria for the diagnosis of infection of ulcers and classifies it into three categories: mild, moderate, or severe. The system was validated in three relatively small cohorts of patients with a DFI (13–15), and increasing severity of infection was associated with toe amputations in one large cohort (16). While the system seems valid in predicting amputation, it does not take into account the results of diagnostic tests such as probing to bone and does not include more general patient characteristics that may have a major impact on the outcome of a DFI.

The European Study Group on Diabetes and the Lower Extremity (Eurodiale) prospectively studied a large cohort of patients with a diabetic foot ulcer (17), enabling us to determine the prognostic value of the IWGDF system for clinically relevant lower-extremity amputations. A previously published study of this cohort identified determinants of minor amputation but did not specifically focus on patients with an infected diabetic foot ulcer (18). The aims of the current study were to 1) identify which signs of infection independently predict amputation within 1 year in patients from the Eurodiale cohort presenting with a DFI, 2) determine the predictive value of the IWGDF system for amputation in this cohort, and 3) develop an alternative risk score for amputation in individuals with a DFI from characteristics of the foot (ulcer), the leg, and the patient that can readily be assessed by the clinician at presentation to guide further management.

Between 1 September 2003 and 1 October 2004, all patients with diabetes presenting with a new foot ulcer to any of 14 diabetic foot centers in 10 European countries were included. Both in- and outpatients were included. The design and rationale of the multicenter, observational, prospective Eurodiale study have previously been described in detail (17). We excluded patients treated in the participating centers for an ulcer of the ipsilateral foot during the previous 12 months and those with a life expectancy of <1 year.

Patients were followed monthly until healing of the foot ulcer(s), major amputation, or death—up to a maximum of 1 year. Healing was defined as complete epithelialization of the whole foot at two consecutive visits. All patients were treated according to protocols based on the International Consensus on the Diabetic Foot (19).

Ulcer characteristics were described according to the PEDIS system of the IWGDF, which classifies foot ulcers according to five categories: perfusion, extent, depth, infection, and sensation (19,20).

Ankle-brachial index (ABI) <0.9 was considered to represent the presence of PAD. A deep ulcer was defined as a lesion of the skin extending through the subcutis with visible muscle, tendon, or bone, whereas a superficial ulcer was limited to the skin. In patients with a deep ulcer, the probe-to-bone test was performed, yielding two categories of deep ulcers: those with a positive and negative probe-to-bone test. Infection was diagnosed according to the IWGDF system by the presence of two or more of the following signs: frank purulence, local warmth, periwound redness, lymphangitis, periwound edema, pain or tenderness on palpation, or fever (21). For the current study, only the data of patients with an infected foot ulcer diagnosed by the IWGDF/IDSA guidelines were analyzed. We defined two separate outcomes: 1) any amputation, including all minor and major amputations, and 2) amputations proximal to and including the hallux, excluding amputations of digits 2–5. We performed analyses for both of these outcomes, as amputations of digits 2–5 could be considered functionally less important for mobility and quality of life and as part of treatment rather than a measure of poor outcome.

Unadjusted associations between the 1-year amputation incidence and patient, leg, ulcer, and infection characteristics were tested with χ² tests (categorical characteristics) and t tests (continuous characteristics). The characteristics studied included sex, age, immobility, serum creatinine, HbA_1c, presence of PAD and polyneuropathy, ulcer size, depth, duration and location, periwound redness, (periwound) edema, pain, foul smell, exudate/pus, increased local skin temperature, lymphadenitis/lymphangitis, fever, and C-reactive protein (CRP) levels. The relation of the different characteristics with amputation was analyzed in multivariable Cox proportional hazards regression models. In these analyses, a hazard ratio (HR) >1 indicates a higher amputation rate for the corresponding category of a certain characteristic compared with a baseline category. For instance, an HR of 2 means that at any time point during follow-up, the patients in the corresponding category are twice as likely to undergo an amputation as the patients in the baseline category.

The patient demographics and foot, ulcer, and infection characteristics were used to construct risk scores for any amputation and for amputations excluding lesser toes in patients with an infected ulcer. Firstly, a score corresponding to a full model was defined as the linear term of a logistic regression model on the 1-year amputation incidence that included all demographics, foot and ulcer characteristics, and infection characteristics. Thereafter, this full model was reduced by iterative (backward) elimination of the least significant predictor until all predictors were significant (P < 0.05); the final risk scores are a version of the reduced models where the parameters are rounded to increase usability. The full models and the risk scores were evaluated, together with the IWGDF/IDSA classification, with receiver operating characteristic (ROC) curves.

Five hundred and ninety-one (57%) of 1,033 patients who completed the Eurodiale study had an infected ulcer, and data were complete for 575 (97.3%) of them. One hundred and ninety-nine patients had a grade 2 (mild) infection, 338 a grade 3 (moderate), and 38 a grade 4 (severe). Amputations were performed on 159 (28%) patients (126 minor and 33 major) within the year of follow-up; 103 patients (18%) underwent amputations proximal to and including the hallux.

Several general patient and ulcer characteristics were univariately associated with any amputation as depicted in Table 1. The incidence of amputation increased with increasing redness, periwound or pretibial edema, the presence of pus, lymphadenitis/lymphangitis, fever (all P < 0.01), and elevated CRP levels (P = 0.01). The same variables were associated with amputations excluding the lesser toes, with the exception of pus and lymphadenitis/lymphangitis. Incidence of amputation was also directly related to increasing severity of infection according to the IWGDF system (P < 0.01), with 50% of patients with systemic inflammatory signs (IWGDF grade 4 infection) undergoing any amputation and 40% undergoing amputations excluding the lesser toes. Immobility, serum creatinine, HbA_1c, presence of polyneuropathy, and ulcer duration and location were not associated with amputation.

Table 2 shows the results of Cox regression analysis corrected for all of the variables as listed in 2research design and methods as well as for center (unadjusted analyses in Supplementary Table 1). The independent predictors of any amputation were as follows: periwound edema, HR 2.01 (95% CI 1.33–3.03); foul smell, HR 1.74 (1.17–2.57); purulent and nonpurulent exudate, HR 1.67 (1.17–2.37) and 1.49 (1.02–2.18), respectively; deep ulcer, HR 3.49 (1.84–6.60); positive probe-to-bone test, HR 6.78 (3.79–12.15); pretibial edema, HR 1.53 (1.02–2.31); fever, HR 2.00 (1.15–3.48); elevated CRP levels but less than three times the upper limit of normal, HR 2.74 (1.40–5.34); and elevated CRP levels more than three times the upper limit, HR 3.84 (2.07–7.12).

The independent predictors of amputations proximal to and including the hallux were as follows: the presence of either periwound or pretibial edema, HR 1.84 (95% CI 1.14–2.98) and 1.89 (1.17–3.05), respectively; a deep ulcer, HR 4.12 (2.01–8.47); positive probe-to-bone test, HR 3.98 (2.05–7.71); and CRP levels more than three times the upper limit, HR 2.71 (1.31–5.61). In summary, the same variables predicted any amputation and amputations excluding the lesser toes, with the exception of foul smell, nonpurulent and purulent exudate, and fever, which predicted any amputation only.

In comparison with mild infection, the presence of a moderate infection increased the hazard for any amputation by a factor of 2.15 (95% CI 1.25–3.71) and 3.01 (1.51–6.01) for amputations excluding the lesser toes. For severe infection, the hazard for any amputation increased by a factor of 4.12 (1.99–8.51) and for amputations excluding the lesser toes by a factor of 5.40 (2.20–13.26). Larger ulcer size and presence of PAD were also independent predictors of both any amputation and amputations excluding the lesser toes, with HRs between 1.81 and 3 (and 95% CIs between 1.05 and 6.6).

The risk score for any amputation that included all the studied variables as reported in 2research design and methods had an area under the ROC curve of 0.84; the risk score for amputations excluding lesser toes had an area under the ROC curve of 0.85. To increase the applicability of the scores, we reduced them as described above to the ones presented in Table 3, which include the variables sex, PAD, pain or tenderness on palpation, ulcer size and ulcer depth, and periwound edema yielding area under the ROC curves of 0.80 and 0.78 for any amputation and amputations excluding the lesser toes, respectively. By way of comparison, the area under the ROC curve for the IWGDF system was 0.67 in our population. Six percent of patients in the lowest tertile underwent any amputation; this increased to 50% of patients in the highest tertile. The respective percentages for amputations excluding the lesser toes were 1 for the lowest tertile and 39 for the highest tertile (Fig. 1).

Previously published studies that have aimed to identify independent risk factors for lower-extremity amputation in patients with a DFI have noted an association with older age (5,22), the presence of fever (5), elevated acute-phase reactants (5,22,23), higher HbA_1c levels (24), and renal insufficiency (5,22). In the current study, we identified the following independent risk factors for lower-extremity amputation in patients with an infected diabetic foot ulcer (in decreasing order of HR): positive probe-to-bone test, deep ulcer, elevated CRP levels, and the presence of periwound or pretibial edema. The presence of increased (non)purulent exudate, foul smell, and fever independently predicted any amputation but not amputations excluding the lesser toes. Thus, we confirmed some of the previously noted risk factors but could not confirm others. This may be explained in part by several differences between our study and those previously published. Lipsky et al. (5) retrospectively studied amputation risk in 3,018 patients hospitalized for a skin or soft tissue infection, but only 16% of these patients had an infected diabetic foot ulcer. Among three other studies that focused on amputation risk in patients with an infected diabetic foot ulcer, one analyzed only patients with large (>3 × 3 cm) and deep ulcers (23), while we included all ulcers, and the numbers of patients included in the other two studies (22,24) were relatively small (n = 100 and 165) compared with our cohort (n = 575). There were few patients with significant renal insufficiency in our cohort, which may explain why we did not find an association between renal insufficiency and amputation. Periwound and pretibial edema as well as the presence of increased exudate have not previously been found to be associated with amputation. Periwound edema may be related to more severe infection, and pretibial edema may be associated with worse overall health status, leading to a greater risk of amputation. The presence of increased exudate and foul smell predicted any amputation, but not amputations excluding the lesser toes. It is possible that clinicians would consider an amputation of one of the digits 2–5 in the presence of these signs but would not consider it such a poor prognostic sign as to consider a larger amputation.

Previous studies have shown that with increasing grades of IWGDF infection severity, the risk of amputation increases (13–16,20,21). Also, in our cohort the number of amputations increased with increasing IWGDF risk score. The new risk scores we developed for any amputation, and amputations excluding the lesser toes had higher prognostic capability, based on the area under the ROC curve (0.80 and 0.78, respectively), than the IWGDF system (0.67). However, it should be noted that these Eurodiale scores were developed based on the available data of our cohort, and they will need to be validated in other populations before any firm conclusions can be drawn. The advantage of these newly developed scores is that they are easier for clinicians to perform, as they do not require additional laboratory tests, as does the IWGDF system. There are relatively few signs and symptoms of infection in the Eurodiale compared with the IWGDF risk scores, and several items of the new scores are known risk factors for poor outcome (25–27). Theoretically, these new scores might predict the amputation risk of any ulcer, infected or not. In line with Louis Pasteur (“the terrain is everything, the bacteria is nothing”), the characteristics of the individual in whom an infection develops might be more important than the infection itself.

Sex, PAD, and ulcer depth are included in the risk scores for both any amputation and amputations excluding the lesser toes. Ulcer size was not a risk factor for any amputation in our cohort, probably because ulcers on digits 2–5 are per definition small ulcers, diluting the effect of ulcer size on amputation risk when all amputations were analyzed. It is unclear why periwound edema is a risk factor for any amputation but not for amputations excluding the lesser toes and why pain on palpation is a risk factor for amputations excluding the lesser toes but not for any amputation.

Some experts prefer an early toe amputation to avoid long-term antibiotic treatment, and we therefore expected a large variation between centers in the selection of patients for toe amputations. However, when we added center to the risk score for any amputation, the area under the ROC curve only marginally improved, from 0.80 to 0.84, contradicting our assumption.

Our study has several limitations. In 37% of cases, patients had been treated by a medical specialist for their ulcer before they were enrolled. As these patients likely suffered from severe or complex diabetic foot disease, our results may not be applicable to patients with milder disease. While all patients included in the Eurodiale cohort were treated according to IWGDF protocols, it is possible that treatment differed in the various centers. Thus, we corrected all the analyses for center. As mentioned previously, the risk score we created needs to be validated in other cohorts of patients with DFI. Our data were obtained in 2003–2004, and the incidence of lower-extremity amputation in patients with diabetic foot ulcers has decreased since then in several countries (28). However, studies have observed that the relative risk of amputation for these patients has not changed (29,30), which is why we believe our results are still valid and applicable.

Of the several clinical signs that we studied, deep ulcers, the presence of increased exudate, foul smell, periwound and pretibial edema, and fever and elevated CRP levels were identified as independent risk factors for amputation in patients with infected diabetic foot ulcers. We confirmed the prognostic value of the IWGDF system for predicting amputation, and we developed two new risk scores, one for any amputation and one for amputations excluding the lesser toes, with better prognostic accuracy than this system, which is currently the only one in use for infected diabetic foot ulcers. These newly developed risk scores can be readily used in daily clinical practice without the necessity of obtaining additional laboratory testing. Future studies should be aimed at validating our findings in other cohorts of patients with an infected diabetic foot ulcer.

Acknowledgments. The authors thank Prof. B. Lipsky, Emeritus Professor of Medicine, University of Washington; Visiting Professor, Infectious Diseases, University of Geneva; and Teaching Associate, Green Templeton College, University of Oxford, for critically reviewing the manuscript.

Funding. The Eurodiale study was supported by the Fifth Framework Programme of the European Commission.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. K.P. and V.S. researched data and wrote the manuscript. M.K. reviewed and edited the manuscript and contributed to the discussion. J.A., K.B., M.E., P.H., A.J., E.J., D.M., A.P., G.R.T., H.R., M.S., L.U., V.U., K.v.A., and J.v.B. planned the study, collected data, and reviewed and edited the manuscript. N.S. planned the study, collected data, and wrote the manuscript. K.P., V.S., and N.S. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.