OUP user menu

Predicting foot ulcers in patients with diabetes: a systematic review and meta-analysis

F. Crawford , M. Inkster , J. Kleijnen , T. Fahey
DOI: http://dx.doi.org/10.1093/qjmed/hcl140 65-86 First published online: 3 February 2007

Abstract

Clinical guidelines recommend that all patients with diabetes should be screened annually to establish their risk of foot ulceration. The aim of this systematic review was to quantify the predictive value of diagnostic tests, physical signs and elements from the patient's history in relation to diabetic foot ulcers. Observational studies were identified from: electronic databases (MEDLINE, EMBASE and CINAHL); bibliographies of studies meeting the inclusion criteria; review articles and clinical guidelines; direct contact with authors. Published reports of cohort and case-control studies were considered for inclusion. Pooled estimates were calculated from absolute numbers as weighted mean differences, standard mean differences or odds ratios. Adjusted odds ratios from published reports were also extracted. We identified five case-control and 11 cohort studies. The incidence of foot ulcers ranged from 8% to 17% in the cohort studies, with varying lengths of follow-up. Diagnostic tests and physical signs that detect peripheral neuropathy (biothesiometry, monofilaments and absent ankle reflexes), and those that detect excessive plantar pressure (peak plantar pressure and joint deformity) were all significantly associated with future diabetic foot ulceration. However, there was a paucity of evidence concerning the predictive value of symptoms and signs. Further research is needed to establish the independent factors associated with diabetic foot ulceration, particularly elements from a patient's history and physical examination.

Introduction

The prevalence of foot ulceration among patients with diabetes mellitus ranges from 1.3% to 4.8% in the community, to as high as 12% in hospital.1 This represents considerable patient morbidity, and is associated with substantial health-care costs. The pathophysiology of diabetic foot ulceration is multifactorial, but peripheral neuropathy is thought to be responsible for most cases.

To prevent foot ulceration and amputation, clinical guidelines recommend early identification of risk, based on annual foot screening of all diabetic patients, with targeting of preventive and treatment interventions to ‘high risk’ individuals.2–4 Key to this preventive strategy is a structured clinical assessment that incorporates diagnostic tests alongside a thorough history and examination.

Current guidelines have not integrated data from primary studies that relate to the prognostic importance of diagnostic tests, physical signs and patient history (alone or in combination), the indicators that underlie any structured approach to preventive risk stratification in diabetic patients. We therefore undertook a systematic review to determine the predictive values of such features in estimating the risk of diabetic foot ulceration.

Methods

We followed recommended guidance concerning the conduct of systematic reviews.5

Search strategy

Electronic search strategies were used to identify studies which assessed the predictive value of diagnostic tests, signs and symptoms using MEDLINE (1966–February 2005), EMBASE (1980–March 2005), CINAHL1982–February 2005. The electronic search strategy was developed from clinical MeSH headings and text words. The search strategy is available from authors. We searched the bibliographies of included studies, review articles and national clinical guidelines.

Inclusion criteria

(i) Published reports of cohort or case-control studies that evaluated the factors used to predict diabetic foot ulceration. (ii) All study participants free of active foot ulceration at the time of study entry. (iii) All study participants in either study design had a diagnosis of diabetes (either type I or type II). The outcome (reference standard) was foot ulceration.

Definitions and explanations for the predictive factors assessed in the review are presented in Box 1.

Box 1 Descriptions of the index tests used to predict those at risk of diabetic foot ulceration

Graphic

Quality assessment

Assessment of methodological quality was used items adapted from the QUADAS tool, recommendations for methodological standards for clinical prediction rules and a checklist for assessment of the methodological quality both of randomized and non-randomized studies of health-care interventions.6–8 Quality assessment was done independently by two reviewers (FC, MI), and the results were used for descriptive purposes to provide an overall evaluation of the included studies. Disagreements were resolved by discussion, and information not available in the reports was sought from the corresponding authors of the primary study.

Data extraction

Data were extracted from the studies as absolute numbers and as means with SDs to permit the re-calculation of data as weighted or standardized mean differences and 95%CIs. Where data were available, they were re-calculated as pooled estimates of the effect of the predictive factors. All odds ratios and risk ratios presented in the included studies were also extracted from the published reports.

Statistical analyses

Data from the two different study designs are presented separately. We only present estimates of effectiveness where there were two or more reports for individual predictive factors. A complete list of estimates of all predictive factors is available from the authors.

As the review focused on a single outcome (diabetic foot ulceration) groups of patients were categorized into those who ulcerated and those who did not. Continuous outcomes, expressed as means and SDs, were pooled as weighted mean differences (WMD). Peak plantar pressure was measured using different dynamic platform-based equipment systems, and consequently a standardized mean difference (SMD) was used to pool data. Tests for heterogeneity were performed, and where heterogeneity was evident, a random effects model was used.9

Findings are presented using the following structure for potential predictive factors; diagnostic tests, physical signs and patient history.

Results

Characteristics of included studies

After independent assessment by two reviewers, 16 studies were judged to have met all inclusion criteria (Figure 1), with disagreements being resolved by discussion. Details of the study population and aspects of diagnostic tests, physical signs and patient history are presented in Tables 1 and 2. Five studies used a case-control design,10–14 and eleven a cohort design,15–25 (Tables 1 and 2). A conference abstract of unpublished data was also identified by the search.26

View this table:
Table 1

Study interventions and outcomes: case-control studies

View this table:
Table 2

Study interventions and outcomes: cohort studies

Figure 1.

Flow diagram of studies in the review.

Data from nine studies were available to calculate pooled estimates.10–14,16–18,,22 The incidence of foot ulceration developed by patients in cohort studies ranged from 8% to 17%, but with lengths of follow-up varying from 12 weeks to 4 years (Table 2).

Quality assessment (Tables 3 and 4)

Case-control studies

All five studies used statistical methods to adjust for confounding factors in the analysis (Table 1).10–14 In two studies, insufficient detail was given about the index tests to permit the presentation of data.12,,13

View this table:
Table 3

Case-control studies: quality assessment

View this table:
Table 4

Cohort studies: quality assessment

Cohort studies

Seven studies reported adjusted estimates for potential confounding factors (Table 2).15–17,19,21,22,,25 Patients received treatment between the index tests and the outcome (assessment of foot ulceration) in all except one study.23

Quantitative estimates concerning the predictive value of diagnostic tests, patient history, symptoms and signs

Unadjusted and adjusted estimates of effect of all predictive factors—diagnostic tests, physical signs and patient history—are summarized in Table 5. Pooled estimates (weighted and standardized mean differences) concerning the predictive value of diagnostic tests (peak plantar pressures, vibration perception threshold and HbA1c) and the duration of diabetes are presented in Figure 2.

View this table:
Table 5

Summary of pooled estimates for predictive value of diagnostic tests, physical signs and patient history in relation to diabetic foot ulceration

Figure 2.

Forest plots of pooled data for the predictive value of a peak plantar pressure, b vibration perception threshold, c HbA1c and d duration of diabetes, for foot ulceration in diabetes. Continues overleaf.

Diagnostic tests

Peak plantar pressures (Figure 2a, Table 5)

Two case-control studies and four cohort studies measured peak plantar pressure, using four different dynamic measuring systems (Musgrave,10 F-scan,22 EMED12,18,,21 and a pedobarograph20,,24). High plantar pressures constitute a risk of ulceration: SMD 0.98 N/cm2 (95%CI 0.63–1.33)10,,12 for case-control studies, and SMD 0.47 N/cm2 (95%CI 0.24–0.70) for cohort studies.18,,22

Vibration perception threshold (VPT) (Figure 2b, Table 5)

Four case-control10–13 and six cohort studies16,17,20–22,,25 found patients with foot ulcerations to have significantly higher VPT than those who did not: WMD 20.00 V (95%CI 11.81–28.20)10,,12 for case-control studies and WMD 17.07 V (95%CI 13.89–20.26)17,,22 for cohort studies.

Transcutaneous oxygen tension (Table 5)

Two case-control studies and one cohort study categorized patient measurements into groups (⩽30 mmHg and 31–60 mmHg). Transcutaneous pO2 ⩽ 30 mmHg was more strongly associated with the development of a foot ulcer, compared with pO2 31–60 mmHg.12,,13,16

HbA1c (Figure 2c, Table 5)

In pooled results from four case-control studies, patients who developed foot ulcers had higher levels of HbA1c than those who did not, but the effect did not reach statistical significance: WMD 0.95% (95%CI −0.33 to 2.23).10–12,,14 Data from one cohort study did demonstrate a statistically significant effect: WMD 1.1% (95%CI 0.57–1.61).16

Ankle brachial indices (ABI) (Table 5)

Four cohort16,17,21,,23 and four case control studies11–14 measured blood pressure at the ankle and arm. Only one cohort study found an effect which remained evident after an adjustment for confounding.16

Fasting blood glucose and serum creatinine (Table 5)

There was inconsistent evidence that increasing levels of blood sugar (mmol/l)14,,19 and creatinine (μmol/l) were associated with increased risk of ulceration.16,,19

Physical signs

Cutaneous sensation (monofilaments) (Table 5)

One case-control study and five cohort studies all found statistically significant differences in the rate of foot ulceration between people whose feet were insensate to ⩽5.07 monofilaments (⩽10 g pressure) and those who were not, with ORs ranging from 2 to 10.13,16,19,21–23

Absent ankle reflexes (Table 5)

Absent ankle reflexes were predictive of a higher risk of foot ulceration in one case-control13 and one cohort study,16 with unadjusted ORs of 4.9 and 1.4, respectively.

Visual acuity (Table 5)

Patients with lower mean visual acuity were at greater risk of foot ulceration in one case-control and one cohort study (adjusted RR 1.9).14,,16

Patient history

Duration of diabetes (Figure 2d, Table 5)

In five case-control studies, patients who developed foot ulcers had diabetes for longer than those who did not, but this effect was not statistically significant: WMD 2.62 (95%CI −0.75 to 5.99).10–14 Combined data from two cohort studies did find a statistically significant effect: WMD 1.88 (95%CI 0.48–3.28).16,,22

History of foot ulceration, amputation and history of lower limb bypass (Table 5)

In four cohort studies investigating the risk associated with a history of foot ulceration,16,20–22 patients who had previous ulceration were more likely to develop diabetic foot ulcers (adjusted ORs ranging from 1.6 to 4.2). One case-control study and one cohort study found a history of amputation to be a risk factor for foot ulceration.12,,16 These two studies also found that a history of lower limb bypass operation predicted future foot ulceration.

Discussion

Summary of findings

We found evidence to support the use of diagnostic tests and physical signs that detect peripheral neuropathy, the principal cause of diabetic foot ulceration. High vibration perception thresholds (VPTs) using a biothesiometer or a tuning fork, high plantar pressure and 10 g monofilaments appear reliable methods to identify those at risk of future ulceration. Absent ankle reflexes, and limited joint motion at both the first metatarsal-phalangeal joint and the subtalar joint were also found to increase the risk of foot ulceration. These findings were evident across different study designs, pooled, unadjusted and adjusted estimates of effect. Established vascular disease, in the form of a history of previous amputation, ulceration or lower limb bypass procedures, was also consistently associated with risk of future ulceration.

None of the published studies reported on the predictive value of signs associated with foot trauma, such as inappropriate footwear and improperly cut toenails.

Evidence concerning the predictive value of ‘contributory’ factors in diabetic foot ulceration, such as some physical signs and elements from the patient's history, was less clear. For example, HbA1c and ankle brachial indices (ABI, ABPI, or AAI) produced inconsistent and contradictory findings (Table 5). The length of time that a person had diabetes was marginally predictive in two cohort studies,16,,22 although in five methodologically weaker case-control studies, the association was not statistically significant.10–14

Shortcomings of this review

Synthesized evidence from this review does support the predictive value of most conventional diagnostic tests used to assess the risk of foot ulceration in people with diabetes. However, only a minority of primary studies assessed the independent predictive value of diagnostic tests in addition to physical signs and elements from a patient's history. Furthermore, different cut-points have been used for many of the diagnostic tests, making comparisons between studies difficult. Some diagnostic tests require a standardized procedure when being carried out. For example, ankle brachial index studies (ABI, ABPI, or AAI) were difficult to interpret because of the lack of detail disclosed as to the position of the patient when blood pressure was measured. Ankle systolic pressure is affected by posture; 1 mmHg higher for each inch the ankle is below the heart.27 This detail was missing from three case-control and four cohort studies, and prevented us from pooling data. Only one cohort study found <0.8 ABI to be predictive of future ulcer risk.16 The value of this procedure in that ulceration risk assessment is yet to be established.

The review included studies that assessed multiple potential predictive factors, and there is a risk of false positive findings in the estimates reported from the primary studies.28 The pooled estimates of nearly all predictive factors showed evidence of significant heterogeneity. This is a consequence of the different and varied definitions for some of the predictive factors, that different cut-points were used, different methods with ascertaining diabetic ulceration and different lengths of follow-up (Tables 1 and 2). Standardizing diagnostic tests and the cut-offs used would be helpful for both research and practice.

Context of other studies

A previous systematic review assessing some of the methods advocated for preventing diabetic foot ulceration suggested that monofilaments, biothesiometer, tuning fork and peak plantar pressure were useful screening tests.29 Our results are consistent with these findings. We also identified eight additional studies that have provided more evidence of the predictive value of tests, and a more integrated approach has permitted data to be pooled.

National and international clinical guidelines also suggest that previous foot amputation and foot ulceration are useful criteria for the identification of patients at ‘high risk’ of foot ulceration.2–4 These guidelines are supported by the findings from this systematic review. However, some recommendations in clinical guidelines do not appear to be based on any firm evidence. For example, we could find no convincing data to support that ‘absent leg or pedal pulses’ is a risk factor for diabetic foot ulceration, despite the suggestion that this clinical sign is a key indicator of risk.3,,4 Few studies adopted a comprehensive approach to the evaluation of predictive factors, and some variables that could predispose to foot ulceration, such as levels of exercise, the presence of callus, Charcot deformity or adequate footwear, have not been subject to extensive evaluation. There is a clear need for further research to address these clinical uncertainties.

Generalizability of findings

The incidence of foot ulcerations in the cohort studies varied from 8% to 17% (Table 2). These are much higher levels of ulceration than those cited in UK national clinical guidelines (5–7%), diabetes text books (7%) and in a national survey (5%).3,.4,.30,.31 This observed difference in incidence may imply that patients included in the studies had more severe disease than those in the general diabetic population. Most of these patients were recruited from hospital diabetes clinics and dedicated foot clinics.

Future studies

Future observational studies need to include the characteristics of patients who were lost to follow-up, and should also attempt to explain the reasons for patients’ withdrawal from studies. None of the sixteen studies included in the review measured adverse events from any of the diagnostic tests evaluated.

The predictive value of relatively simple clinical signs such as the presence or absence of leg and pedal pulses, skin colour, skin texture, hairlessness of the lower legs and condition of the toenails are not known. Clinical signs are potentially more cost-effective than more complex diagnostic tests, and are more feasible in community settings. Given that there are quite marked differences in cost between different tests, any new evidence about cost-effectiveness would deserve consideration. A diagnostic rule, based on elements of the clinical history, examination and available diagnostic tests needs to be developed, validated and tested to establish the effectiveness and cost-effectiveness of using such an approach when assessing the risk of diabetic foot ulceration in community settings.32

Conclusions

Diagnostic tests and clinical signs are helpful in predicting the risk of diabetic foot ulceration. Evidence concerning the predictive value of simpler elements from the clinical history and examination is less clear. Future studies should assess the independent predictive value of all elements of patient history, physical signs and diagnostic tests when assessing the risk of diabetic foot ulceration.

Acknowledgements

The authors thank Ms Julie Glanville, the Centre for Reviews and Dissemination at the University of York and Mr Donald Orrock, Ninewells Hospital Library at the University of Dundee, for their help with the development of the electronic search strategies. Also Professors Armstrong, Boyko, Boulton, Kastenbauer, Rith Najarian and Veves, who provided additional information about the primary studies. Fay Crawford and Melanie Inkster are funded by the Chief Scientist Office, The Scottish Executive, Edinburgh.

References

View Abstract