Prediction and monitoring the therapeutic response of chronic dermal wounds

Abstract

A significant proportion of chronic wounds fail to heal in response to treatment of underlying pathologies combined with good wound care practice. Current prognostic tests to identify these wounds rely on the use of algorithms based on clinically measurable parameters such as wound dimensions and wound duration. Venous leg ulcers may be stratified into healing/non healing at 24 weeks of compression therapy and diabetic foot ulcer treatment outcome assessed using a 3‐parameter algorithm. Accurate and reproducible measurement of wound area is required for these algorithms to have clinical utility. Whilst a number of attempts have been made to develop computerised wound‐assessment techniques, wound tracing by clinicians combined with planimetry remains the standard methodology. Once treatment has been initiated, it is important to continuously monitor the wound to assess efficacy of treatment. This can be achieved by measuring wound area change over the first weeks of treatment to identify whether re‐assessment of treatment strategy is required. A number of algorithms for assessing rate of wound area change have been evaluated to determine a surrogate endpoint for healing. Retrospective analysis of large patient groups indicates that approximately 75% correct prediction of healing outcome can be achieved.

Introduction

For the healthy individual, dermal healing is a structured process leading to rapid re‐establishment of skin barrier function. It follows a defined temporal sequence of haemostasis, early and late inflammation, granulation tissue formation, extracellular matrix synthesis, re‐modelling and scar formation (1). This generates a healing trajectory that can be used to predict time to closure of wounds healing by secondary intention after measurement of initial wound dimensions (2). However, many factors such as infection, comorbidities and ageing may exert a negative influence on healing to induce deviation from an optimal trajectory. This can lead to impaired healing resulting in delay or even prevention of wound closure. Chronic wounds with impaired healing include venous leg ulcers, diabetic foot ulcers and pressure ulcers. They occur at a relatively high frequency estimated at >1% of an aged population for venous leg ulcer (VLU) (3), with 15% of diabetics likely to develop foot ulceration (4), and the prevalence of pressure ulcers can be up to 15% in an acute care setting (5).

Many chronic wounds respond to a combination of treatment of underlying pathologies (compression therapy for VLU, pressure relief for DFU and PU) and good wound care practice by using wound contact dressings that generate a moist wound environment. However, a substantial minority remains refractory to treatment (6) and may require adjunctive therapies to achieve healing. An expanding body of knowledge characterising the differences between healing and non healing wounds has identified potential therapeutic targets and has led to the development of a number of treatments that are designed to modulate healing by interacting with key aspects of biological events that regulate the healing process. Examples include recombinant growth factors such as platelet‐derived growth factor (7), keratinocyte growth factor (8), granulocyte‐monocyte colony‐stimulating factor (9), tissue‐engineered dermal replacements (10), wound dressings with potential bioactivity (11, 12) or synthetic protease inhibitors (13).

The increasing number of therapeutic targets and treatments for chronic wounds generates a need to make early and rational therapeutic choices by identifying those wounds requiring advanced therapies and to monitor treatment efficacy. The demand for validated objective wound‐monitoring assessment systems at all levels of care was emphasised by a recent publication entitled ‘Why won’t this wound heal and what should I do about it?' (14). An ideal solution would be a minimally invasive diagnostic/prognostic monitoring system analogous to those used in other pathologies such as diabetes where the concentration of an analyte, glucose, is known to be closely linked to disease progression and response to treatment. However, the chronic wound may have a number of contributing factors causing multiple pathological differences compared with a healing wound. Our current understanding of the physiological events associated with healing does not allow development of a single parameter‐based assay, and it is probable that a successful wound‐monitoring system will utilise multiparametric assessment possibly incorporating as yet unidentified biomarkers.

The absence of such a system necessitates reliance on existing clinical and laboratory measurements, and it is the objective of this review to focus on clinically measurable wound parameters that may be used for predicting outcome and monitoring treatment efficacy in patients with chronic wounds.

Wound Prognostic Indicators

If available, a validated prognostic indicator would be used to stratify wounds into those requiring advanced therapies and those potentially responsive to standard treatment, and would confer quality of life and cost benefits by accelerating healing and reducing usage of inappropriate treatment. Although many parameters are known to be associated with a poor outcome for chronic wounds (Table 1), only initial wound area and wound duration have been incorporated into statistically evaluated prognostic algorithms.

Table 1.

Risk factors for delayed healing of venous leg ulcers

Deep vein incompetence (55)

Popliteal vein reflux (56)

Socio‐economic status (57)

History of previous ulceration (16)

Highly exuding ulcers (16)

Increasing age (16)

Male gender (58)

High body mass index (59)

Low serum zinc (60)

Low albumin levels (61)

Anxiety and depression (62)

Long duration (16)

Large area (17)

Venous leg ulcers of long duration (15, 16) and with a larger area at initiation of treatment have a decreased chance of healing in response to compression therapy (17). An ulcer length greater than 10 cm (18), increased width, length, length × width and area all correlate with failure to heal, but when the ulcer area is >40 cm² the correlation between area and healing/non healing decreases (19). A simple model for VLU prognosis based on scoring 1 point for ulcers >5 cm² and 1 point for >6 months' duration was developed to discriminate between those wounds that will heal after 24 weeks compression bandaging (95% of wounds with a score of 0) and those with a decreased chance of healing (37% of wounds with a score of 2) (20). Following analysis of a larger patient cohort, this model was subsequently refined so that a wound with an initial area <10 cm² of <12 months' duration will have a 81% chance of healing by 24 weeks of compression bandaging, whilst one, that is, area of >10 cm² and >12 months' duration has only a 22% chance of healing (21). It is important to recognise that these data were determined for VLU undergoing compression therapy. The authors introduced the important caveat that the prognostic indicators may not be valid for predicting outcomes of new therapies untested in this model.

The same relationship between healing, size and duration applies to diabetic neuropathic foot ulcers in that ulcers healing within 20 weeks of standard care are more likely to be smaller and of shorter duration. Incorporation of ulcer grade allows a 3‐parameter model as superficial diabetic wounds heal more effectively than deeper wounds and those with abscess and osteomyelitis (22). The model is based on a scoring system in which a point is scored if the wound is older than 2 months, larger than 2 cm², or has a grade ≥3 on a 6‐point scale. An increased score indicates a greater chance of non healing so that a score of 3 indicates an 81% chance of non healing compared to 35% for a score of 0 (23). Large diabetic ulcer size also correlates with a risk of subsequent limb amputation (24).

Monitoring Response to Treatment

Prognostic algorithms are intended to assess the patient and wound at a single time point before treatment as an aid to the development of a treatment strategy. Once treatment has been initiated, a dynamic model is required to monitor change in wound status and to evaluate therapeutic efficacy with sufficient accuracy that it can be used as a decision‐making tool. Wound closure is the obvious endpoint of healing, and the logical and simple measure of progress towards that endpoint is to monitor response to treatment by assessing change in wound size over time. It is accepted clinically that measurements taken at regular intervals give a good indication of a wound's healing progress and allow for a timely recognition of improvement or deterioration of the wound condition (25). Using this parameter allows monitoring algorithms to be devised using historical data and then to be evaluated in prospective studies.

Wound area change over 4 26, 27, 28 or 3 weeks (15) has been identified as the best indicator for use as a surrogate endpoint to monitor healing response and predict outcome. A complicating factor in evaluating healing is the consideration of which parameter to use in calculating change in wound area. In attempting to use absolute wound area change after 4 weeks of treatment to predict outcome at 24 weeks, the rate of healing or area healed per week did not differentiate between healing and non healing (19). Percent area reduction has been recommended as the best way of predicting healing rates (29), although this has been challenged on the basis that it biases wound closure rate in favour of smaller wounds (30). Expression of the data as percentage change in area from baseline to 4 weeks provides the best combination of positive and negative predictive values (68·2 and 74·7%, respectively) and the largest area (0·75) under the Receiver Operator Characteristis (ROC) curve (Box 1, 31). A retrospective study of 56 488 wounds (32) found that the log healing rate [(Log area at time 0 — Log area at time + t weeks later)/t], and percentage change in wound area measured over a 4‐week period could also be used as a surrogate marker of healing at 12 or 24 weeks. The ROC value found was 0·72–0·80 for change in wound area over 4 weeks.

In other studies, estimating change in area at 3 weeks by comparison with initial measurement, the so‐called baseline adjusted healing rate, was found to be unsatisfactory in predicting outcome (33). However, by taking the mean of healing rates between each visit to determine a mean‐adjusted healing rate, it did prove possible at 3 weeks after starting treatment to predict eventual healing outcome. By comparison with absolute area change or percentage area change, it has been suggested that use of the rate of change of wound perimeter may provide a more accurate reflection of healing rates when comparing ulcers of different size, as this parameter is independent of wound geometry (34). Although wound perimeter is rarely calculated in clinical practice, it is readily available if the area is measured by wound tracing.

The utility of measuring early area change on a single patient basis (as opposed to comparison of patient groups) has recently been questioned. A study of 17 patients indicated that the ability of initial healing rate to predict healing time for patients undergoing compression therapy was poor (35). Although an increase in wound area over 4 weeks consistently correlated with non healing, it did not prove possible to relate rate of area decrease and outcome with any accuracy. This may have been a consequence of the heterogeneity of the wounds included and the low healing rates (29%) achieved in this study compared with the higher rates (66%) in others where a positive correlation was demonstrated (32). For a prognostic test to be of value, it must have utility when applied to a patient in a clinical setting. On the basis of Hill's results (35), the value of this approach may be restricted to identifying those patients who require a re‐assessment of their treatment regime after 4 weeks of treatment. In contrast, a study of PU healing (36) concluded that if stage IV ulcers and wounds with initial size <2 cm² are assessed by initial area change, then use of a Gompertzian statistical model provides a relevant method to evaluate therapeutic interventions. Clearly, further development work is required before a reliable and universally applicable wound monitoring method is available.

Measurement of Wound Area

Assessment of wound appearance has traditionally been evaluated visually by clinicians administering patient care. Such assessment is highly dependant on individual proficiency (37). The obvious problems of lack of objectivity and requirement for skill and experience are compounded if different clinicians make assessments at different times on the same wound. There is thus a need for objective and documented information to monitor wound progress and to determine the effect of therapeutic interventions. Instrumentation developed for this purpose needs to duplicate the information derived by human assessment and, if possible, derive additional information using techniques such as colour or surface texture analysis.

Manual techniques

A crude assessment of wound area may be derived by measuring length and breadth with a ruler. Chronic wounds are rarely symmetrical, and this approach does not generate the desired accuracy (38). The widely used alternative method of tracing the wound margin onto a transparent film is more accurate and simple to use. It requires skill in use with complex or circumferential wounds and is subjective in that the clinician has to decide where the actual wound margin is when tracing onto the film. The tracing forms a permanent record, and area can be calculated by counting squares if a grid is placed over the tracing or by mechanical or computerised planimetry. A hand‐held device, Visitrak Digital (Smith & Nephew Medical, Hull, UK), has recently been introduced to reduce the time involved in calculation of area from wound tracings and improve inter‐ and intraoperator accuracy so that area change over time can be more accurately monitored and recorded (39).

Automated techniques

Identification of the wound margin to allow automated definition of the wound periphery and calculation of the wound area remains a major challenge. This can be achieved in a research setting using deformation of structured light to define the wound surface. Laser systems have been demonstrated to be faster and of equivalent accuracy to tracing for measurement of simulated wounds in vitro (40). Use of a laser mounted on a motorised X‐Y table located above the wound surface gives less than 1% error in measuring the volume of such model wounds (41).

Two systems using structured light techniques are currently available to measure wound parameters directly in a clinical situation. Patterns of laser‐generated lines or dots are projected onto the area around the wound. The measurement of area and volume instrument (MAVIS) system generates a calculated volume from the degree of observed distortion of parallel lines (42) but still requires the intervention of a clinician to define the wound margin on a computer display. The requirement for full automation is addressed by an alternative system utilising a pattern of dots projected over the wound surface. After image capture, the distortion of each dot is assessed, and where this falls outside the predicted shape, a computerised decision is made so that the dot falls on the wound margin to allow perimeter definition by interpolating dots at the margin (43).

These techniques require sophisticated, expensive and relatively cumbersome instrumentation at the patient's side. With wide availability of digital cameras, a more attractive option is to capture a digital image of the wound for subsequent computerised analysis (44). In effect, this converts a 3D image into 2 dimensions, and single plane photographs are inherently limited by this transformation. If a wound extends around the curvature of the leg then a 2D image under‐represents the true wound area. A further limitation is that the 2D image takes no account of the depth of the wound. Commercial products have been developed for photographic image analysis such as the VERGE Videometer system (45). A digital image of the wound is captured for computerised analysis, but the system suffers from the same limitation as the MAVIS system in that the clinician has to delineate the wound margin.

Progress has been made towards computerised delineation of the wound margin. Using an adaptive spline technique (46), it is possible to define the wound margin by performing analysis of image features in the region of the wound margin. Currently, this is a semi‐automatic technique as a sequence of initial control points have to be defined on the image at the wound margin. This technique has been incorporated into a system for colour analysis of venous leg ulcers (47).

Colour analysis can be of value in assessing a wound's progress towards healing. Depending upon severity and aetiology, wounds may to some extent be covered with black eschar and epithelialisation will fail until it is removed by debridement. As treatment succeeds and healing progresses, the wound colour changes from black to yellow to red and finally granular red. Less severe wounds may appear a healthy red colour on presentation. Wound appearance has been widely documented on a red/yellow/black system (48) and although easy to use has been criticised as difficult to quantify manually (49). Development of computerised image analysis for measurement of wound area brings with it the possibility of simultaneous colour analysis to impose objectivity to an assessment that has relied heavily on clinical expertise.

Conclusion

Following an initial diagnostic assessment, current treatment strategies follow a trial‐and‐error approach. For a venous leg ulcer, the first line approach is compression bandaging that will induce healing in a proportion of patients. Those non responsive to compression therapy alone may require adjunctive therapies such as tissue‐engineered dermal equivalents that provide temporary wound cover or growth factors that may stimulate healing (50). The remaining non responsive wounds may then be treated with aggressive surgical debridement or venous surgery. There will remain a residual population of wounds that are refractory to any treatment. These wounds will require management to maintain patient quality of life. As a non healing wound passes through each treatment iteration, the total treatment costs and impact on patient quality of life increase. These costs might be minimised if a prognostic indicator was available to identify at treatment initiation those wounds potentially non responsive to compression therapy. Additional clinical benefit may be gained by the use of a monitoring system that would give an early warning of poor treatment response and an indication that a change to the treatment regime was required to stimulate healing.

The biological complexity of the healing process and the interrelatedness of the many processes involved in healing present many opportunities for underlying pathologies to lead to impaired healing. It is unsurprising therefore that many risk factors for non healing have been identified (Table 1). Patient and wound assessment form the basis for chronic wound management, and consideration of risk factors at initial sassessment can provide an experienced clinician with an indication of the potential outcome of treatment. However, the only non subjective prognostic indicator currently available for the less experienced practitioner is an algorithm based on wound area and wound duration. Measurement of area by tracing is relatively simple and combined with wound duration can give an estimate of healing potential of venous ulcers in response to compression therapy with a reasonable level of confidence. Whilst not giving an absolute prediction of outcome, this algorithm may be used to identify those wounds that require more rigorous evaluation of response to treatment.

As well as depending on successful treatment of the underlying pathologies, wound management outcome is dependent on exogenous factors such as bacterial proliferation in wound tissue (51). This requires continuous monitoring of the wound to ensure that appropriate interventions are implemented. Whilst spreading infection is identified by the clinical symptoms of odour, pain and cellulitis, a lower level of bacterial bioburden resulting in critical colonisation or local infection may cause inhibition of healing without the associated symptoms of infection (52). Critical colonisation may manifest in impaired healing and may be observed as a non initiation or a cessation of healing in response to compression therapy. Whilst laboratory tests to quantify bacterial bioburden are available on biopsy tissue, these are not routinely performed clinically on a prospective basis.

Monitoring wound area change on a regular basis remains the main method for evaluation of treatment response. Whilst this can be performed on an ad hoc basis, it requires accurate and reproducible measurement at weekly intervals to accurately monitor and document wound progress. Additional to monitoring response, wound area change over time can be used as a predictor of outcome with an approximate 75% accuracy as shown by the area under the ROC curve. It has been suggested that this level of discrimination between healing and non healing is sufficient for the rate of wound area change over an initial 4 weeks of treatment to be used as a surrogate endpoint for clinical trials, and to identify patients unlikely to heal early in the course of treatment (32). The latter concept is supported by prospective data where non responding wounds were accurately identified by increasing wound area (35). However, this study did not find a good correlation between rate of wound area decrease and subsequent healing. These data strike a note of caution because the clinical value in this method lies in its use as an early prospective test to evaluate the success of a particular treatment regime. Validation in clinical use for the prediction of time to healing will have to await the outcome of further prospective studies.

Whilst the measurement of clinically measurable wound parameters has value in predicting and monitoring healing outcome, these are at best 80% accurate. A large body of data is available to characterise the molecular environment of the chronic wound (53), but no analytes have been identified for incorporation into a test of wound prognosis and monitoring. The possible combination of biomarkers characterising the wound microenvironment (54) with dynamic clinically measurable wound parameters may yield an accurate wound assessment and monitoring system.

Box 1.


 The Receiver Operator Characteristic 


The ROC curve is often used to evaluate diagnostic tests. It provides a measure of the ability of a test to properly detect an event that actually occurs, called sensitivity or the true‐positive fraction, relative to false detection in the absence of the event, called the false‐positive fraction (62). The closer the area under the ROC curve approaches a value of 1, the greater the utility of the test.