Does dermal thermometry predict clinical outcome in diabetic foot infection? Analysis of data from the SIDESTEP ^trial

Abstract

The purpose of the study was to assess in patients with a diabetic foot infection (DFI), whether differences in skin temperature of the affected foot as compared to the corresponding site on the contralateral foot using dermal thermometry (DT) correlates with infection severity and clinical outcome. As part of the SIDESTEP DFI study, investigators took DT measurements at baseline and the discontinuation of intravenous therapy (DCIV) and performed a systematic evaluation of the infected limb to calculate a wound score. We compared the skin temperature differential between the limbs at the two assessments and determined the correlation between this value and surrogate markers of inflammation and the clinical response to treatment. Among patients enrolled in SIDESTEP, 332 were fully evaluable. The mean temperature differential between the limbs was 2·81 ± 5·75°F at baseline and 2·43 ± 4·84°F at DCIV (mean change: −0·37; 95% confidence interval (CI): −0·98, 0·23; P= 0·225). Skin temperature differential at baseline did not correlate with white blood cell count, level of C‐reactive protein or erythrocyte sedimentation rate or the infection severity score (r= 0·058, 0·148, −0·002, 0·067, respectively). We observed no overall trend between surface temperature differential at baseline and clinical outcome at DCIV, but patients with a skin temperature differential of ≥10°F at baseline had a significantly lower clinical response than those whose differential was <10°F (81·4% versus 94·3%; difference 12·9%; 95% CI: 3·5, 27·3%, P= 0·007). While there was no overall relationship between skin temperature and poor clinical outcome, there may be a threshold effect in DT (<10°F versus >10°F) between the limbs at baseline that predicts outcome of therapy.

Introduction

Foot wounds are a common and serious problem among persons with diabetes (1). More than half of diabetic foot wounds will ultimately become infected and one in five of these will lead to lower extremity amputation 2, 3. The severity of diabetic foot infection (DFI) is largely determined by clinical parameters, mostly related to local signs of inflammation of the wound combined with evidence of systemic illness and metabolic aberration. Determining the severity of an infection is a key determinant in selecting the site of patient treatment (hospital versus outpatient setting), the route (parenteral versus oral), spectrum (narrow versus broad) and duration (long versus short) of antibiotic 4, 5, 6. The clinical parameters used to assess the severity of a DFI are largely subjective, however, and there are few reliable clinical markers to assist the clinician in this regard.

Inflammation is one of the earliest signs of impending foot ulceration in the diabetic patient. While many of the characteristic signs of inflammation (i.e. redness, pain, swelling and loss of function) are difficult to assess objectively, warmth can be measured. Local skin warmth has been found to be a relatively consistent surrogate marker for inflammation 7, 8, 9, 10. Skin surface temperature can be readily measured in both the clinical and home setting using a method called dermal thermometry (DT) 11, 12, 13, 14, 15, 16. When used for evaluating an extremity, investigators have not generally relied on a specific temperature, but rather the difference in temperature between the affected foot and a corresponding point on the unaffected contralateral foot, which serves as a physiological control. In this context, DT can provide quantitative information that has been shown to be predictive of ulceration in high‐risk patients and a useful tool to diagnose and monitor healing of diabetic neuropathic osteoarthropathy (Charcot’s arthropathy) 12, 17, 18. A pilot study in this area suggests that persons at high risk for foot complications can effectively use an infrared thermometer as a home monitoring tool to prevent foot ulceration (13).

We have postulated that measuring skin temperature by DT might be useful in determining the severity of a DFI and in predicting the clinical outcome with treatment. While there have been reports of the potential utility of DT in monitoring progress of infected joint replacements and post‐sternotomy infections 19, 20, we are unaware of any previous reports in the medical literature that have addressed the use of DT in DFI. We sought to assess whether differences in skin temperature of the affected foot as compared to the corresponding site on the contralateral foot correlates with infection severity, certain laboratory markers for inflammation or clinical response to antimicrobial therapy.

Methods

SIDESTEP was an 89‐centre, randomised, double‐blinded comparison of intravenous (IV) ertapenem (1 g once a day) and piperacillin/tazobactam (3·375 g every 6 hours) for 5–28 days for treating moderate to severe DFI. We have previously described the methods and results of this, the largest published trial of treating DFIs (5). As part of SIDESTEP, the investigators performed a thorough and systematic evaluation of the patients’ infected wound, which included conducting DT on the infected wound site.

Before enrolling any patients, all investigators and study coordinators attended a centralised training meeting (or viewed videotapes of the lectures) to teach them about DFIs and to ensure they understood the protocol and verify that they could properly conduct all required procedures. Enrolled patients were instructed to remove their shoes and rest supine in a chair or on an examination table or bed for 3–5 minutes. The investigator held an infrared thermometer (Thermo‐Trace™; Deltatrak, Pleasanton, CA, USA) approximately 1 inch above the skin. Temperature measurements on the affected foot were taken as close as possible to the area of the DFI. Temperature was also measured in the corresponding area on the unaffected contralateral foot. For patients who had undergone an amputation on the contralateral limb, the temperature was measured on the distal aspect of the residual limb. Skin temperatures were determined at baseline (the day of patient enrolment) and at discontinuation of IV therapy (DCIV). Evaluations conducted on each enrolled patient included determining their baseline white blood cell (WBC) count, C‐reactive protein and erythrocyte sedimentation rate, as well as determining their University of Texas Wound Classification System stage and grade (12). We designated patients as stratum I if they had an infection with a ‘moderate’ wound (Grade 0 or 1, Stages B or D) and stratum II if they had an infection with a ‘severe’ wound (Grades 2 or 3, Stages B or D).

We calculated the difference in surface skin temperatures by subtracting the temperature measured on the unaffected foot from that of the affected foot. We compared the mean changes in skin temperature from baseline to DCIV using the paired t‐test. Corresponding 95% confidence intervals (CIs) were calculated for the mean changes. We used an analysis of covariance model with a factor for treatment group and covariate for baseline severity (moderate or severe) to compare treatment groups with regard to the mean change in skin temperature difference from baseline to DCIV. To compare the means of skin temperature difference between patients with moderate versus severe baseline wounds we used analysis of variance. We calculated 95% CI for each severity group mean and for the difference in severity group means.

We used Pearson’s correlation coefficient, with the corresponding 95% CI and P‐values, to assess the relationship between baseline WBC, C‐reactive protein, sedimentation rate and wound score with baseline skin temperature difference. We calculated favourable clinical response rates for each of the following skin temperature difference categories: less than −3, −3 to 1, 0 to 1, 2, 3 to 4, 5 to 6, 7 to 9, <10, and ≥10°F. For each favourable clinical response rate, we used the exact methods to calculate 95% CI. We used exact methods to calculate a 95% CI about the difference in favourable clinical response rates for patients with baseline skin temperature differences of <10°F versus ≥10°F. All tests and CIs were two‐sided with an alpha of 0·05.

Results

Among the evaluable patients in SIDESTEP, 64% were male, and their mean age was 59 years. Investigators obtained skin temperatures measurements by DT at the baseline assessment in 362 patients. The skin temperature differential between limbs was similar in patients with baseline diabetic foot wounds classified as moderate (3·04 ± 5·88°F) compared to severe (3·09 ± 5·38°F). A total of 332 (92%) of the patients with a baseline DT measurement for both the affected and unaffected contralateral foot also had one at the DCIV assessment. These data are illustrated based on treatment group assignment and severity in 1, 2, respectively.

Table 1.

Difference in skin temperature between the affected foot and unaffected contralateral extremity

Treatment group	N	Baseline		DCIV		Change
						Within‐group
		Mean	SD	Mean	SD	mean	SD	95% CI	P‐value
Ertapenem	167	3·01	5·82	2·61	5·05	−0·40	5·34	(−1·22, 0·41)	0·330
Piperacillin/tazobactam	165	2·60	5·68	2·25	4·63	−0·35	5·89	(−1·25, 0·56)	0·452
Overall	332	2·81	5·75	2·43	4·84	−0·37	5·61	(−0·98, 0·23)	0·225
Between‐group difference in change from baseline (ertapenem–piperacillin/tazobactam)
Mean	95% CI	P‐value
−0·05	(−1·26, 1·17)	0·937

CI, confidence interval; SD, standard deviation; DCIV, discontinuation of intravenous antibiotics. N is the number of follow‐up assessment clinically evaluable patients with both baseline and DCIV temperature data. Between‐group estimates are derived from an analysis of covariance model, including treatment group and baseline severity (moderate, severe) as a covariate.

Table 2.

Difference in skin temperature (affected foot minus unaffected contralateral extremity) at baseline, by baseline severity

Baseline severity	N	Mean	SD	95% CI
Moderate	250	3·04	5·88	(2·31, 3·77)
Severe	112	3·09	5·38	(2·08, 410)
Overall	362	3·05	5·72	(2·46, 3.64)
Between‐group difference in change from baseline (severe–moderate)
Mean	95% CI	P‐value
0·05	(−1·23, 1·33)	0·938

CI, confidence interval; SD, standard deviation. N is the number of follow‐up assessment clinically evaluable patients with baseline temperature data. Between‐group estimates are derived from an analysis of variance model with a factor for baseline severity (moderate, severe).

At the baseline assessment, the mean temperature differential between limbs in these patients was 2·81 ± 5·75°F. At the DCIV assessment (a mean of 11·7 ± 7·5 days after the baseline assessment) the mean temperature differential was 2·43 ± 4·84°F. Thus, the mean change between these two visits was only −0·37°F (95% CI: −0.98, 0·23; P= 0·225). There was no correlation between skin temperature differential at the baseline assessment and the systemic markers of inflammation, i.e. the patient’s WBC, C‐reactive protein, erythrocyte sedimentation rate or the DFI wound score (r= 0·058, 0·148, −0.002, 0·067, respectively). Overall, we observed no trend between skin temperature differential at baseline and the clinical success rate for treatment of the DFI at the DCIV assessment. In a subset of patients with a temperature differential ≥10°F at baseline assessment, however, a significantly smaller proportion of patients had a favourable clinical response, compared to patients with a temperature differential <10°F (81·4% versus 94·3%; difference between groups: 12·9%; 95% CI: 3·5%, 27·3%; P= 0·007). These data are illustrated in Figure 1.

Figure 1.

Proportion of patients with favourable clinical response atdiscontinuation of intravenous therapy (DCIV) by difference in skin temperature at baseline.

Discussion

This study was designed to see if measuring the skin temperature near a DFI and comparing it to both the temperature on the unaffected limb and the temperature at a later follow‐up time, would be useful in objectively determining the severity of an infection and the likely clinical outcome of treatment. Our results suggest that the mean temperature differential between the affected and unaffected limbs in patients with a DFI was not very helpful in predicting the severity of an infection or the outcome of treatment. There was no significant difference in the skin temperature measurements at the baseline and DCIV assessments. Additionally, there was no significant difference between subjects based on other potential markers of systemic infection or severity of inflammation, such as the WBC, C‐reactive protein, or erythrocyte sedimentation rate. Finally, there did not appear to be a difference in skin temperature differential based on whether the wound severity by the University of Texas classification was moderate or severe, There did appear to be some difference, however, between the degree of skin temperature differential and a poor clinical outcome.

The lack of difference in skin temperature differential between the baseline and the DCIV assessments in this study may reflect the fact that just under 12 days is an insufficient time to allow ‘cooling’ of the signs of inflammation. Unfortunately, we did not collect DT data at the end of the entire course of antibiotic therapy or the final follow‐up study visit at ∼2 weeks after finishing the treatment for infection. It is feasible that there would have been a steady reduction in the temperature difference between the DFI and the contralateral site when taken out to a longer clinical duration. Our results strongly suggest that a rapid reduction in temperature (baseline to DCIV assessments) was not obvious.

The results also show that there was no difference in the mean temperature differential between patients classified as having moderate compared with severe wound severity, as described by the University of Texas Wound Classification system. This is among the most commonly used diabetic foot classification systems because it includes the main variables known to be associated with poor outcome, i.e. depth of the wound and presence of infection and ischaemia., This system is not, however, designed for grading infection severity. Therefore, a system designed specifically for grading DFIs, such as the Infectious Disease Society of America or International Working Group on the Diabetic Foot system 4, 5, 6 might prove more useful for gauging infection severity and predicting infection outcome. At the time we designed this study, the University of Texas Wound Classification System was the only validated wound classification system for the diabetic foot. It is also possible that either of the infection classification systems might not have led to a different result, as there was no correlation observed between skin temperature differential and baseline disease severity as measured by WBC count, erythrocyte sedimentation rate or C‐reactive protein.

A higher favourable clinical response rate (improvement or cure) was observed at the DCIV assessment for patients with a baseline difference in skin temperature <10°F compared to patients with a baseline temperature difference of ≥10°F. Conversely, those patients with a markedly elevated temperature differential had worse clinical outcomes than those that did not. This is not surprising, as we have found that among patients enrolled in SIDESTEP, an elevated WBC (above one standard deviation over the mean) was associated with a nearly twofold increased risk for a poor outcome (21). It is possible that the WBC and skin temperature reflect a patient’s systemic and local response to infection, respectively. To our knowledge, this is the first study in the literature to evaluate these potential associations in a large population.

In conclusion, it appears that the difference in DT measurements between the infected and uninfected limb did not change significantly between baseline and the DCIV assessments just over 11 days later, nor was there any correlation between the skin temperature differential at baseline and laboratory markers of inflammation. Of note is that there was evidence to suggest a threshold effect (<10°F versus >10°F) between baseline DT differential and clinical outcome. This may assist in the local evaluation and subsequent prediction of clinical disposition in this most challenging of patient populations.