Healing Rate as a Prognostic Indicator of Complete Healing: A Reappraisal

Disclosure: This work was Supported by NIH Grants AR42936 and AR46557 Introduction Chronic wounds affect millions of people in the United States each year and are a tremendous financial drain to our healthcare system, accounting for billions of dollars annually.[1,2] In Western medicine, venous ulcers are the most common type and diabetic ulcers are a major cause of lower-extremity amputations.[2–4] All together, wounds result in substantial morbidity causing significant reductions in patients’ quality of life.[3] Being able to tell early in the treatment of a wound whether a therapy is working is of extreme importance and can offer both medical and economical benefit to clinical practice as well as clinical research.[5] In clinical practice, this prognostic ability would enable the early identification of poorly healing wounds, allowing for re-evaluation to establish if factors beyond the therapy are causing delayed healing (e.g., incorrect diagnosis) or if the current therapy should be altered.[5] Advanced therapeutics, such as cultured skin equivalents and growth factors, are expensive and are generally reserved for patients who have failed standard therapy. A dependable early determinant of healing will aid in management of these resources by identifying those who will benefit the most.[5] More importantly, by preventing a patient from continuing on in an ineffective therapy, it will decrease the number of days a patient’s ulcer persists, thus lessening the adverse effect on his or her quality of life and the opportunity for complications, which in the case of diabetic ulcers could mean amputation.[6] Most clinical studies of wound therapeutics use primary endpoints that are assessed over a lengthy period of time, such as 24 weeks. Having a reliable prognostic index that may be instituted early in treatment not only will decrease the resources required for a clinical study but will also allow study participants who would not have healed to be placed on an alternate therapy far sooner. Over the last 15 years, a number of studies have considered the ability of initial wound healing rates to predict complete healing. There are three parameters for measuring healing that have shown promise in assessing initial healing rate for reliable prediction of future healing: linear inward progression of the wound edge, change in wound area, and percent change in wound area. Change in Area and Percent Change in Area In 1987, Bulstrode, Goode, and Scott found that percent change in area in the third week of a new therapy gave the earliest close correlation to time to complete healing, predicting time to complete healing within one week of actual outcome in 47 percent of the ulcers.[7] Predicting time to complete healing is much more specific and challenging than predicting complete healing in a particular time frame, such as 24 weeks, which may be just as helpful clinically and was something for which they did not assess. They did, however, recognize the importance of their findings to clinical studies, in that if the time required to follow a wound therapy during a study could be reduced from 24 weeks to four weeks this would be of great benefit. In 2000, Phillips and colleagues retrospectively assessed 165 patients with venous ulcers for a number of prognostic indicators.[6] Using computer-based planimetry to measure changes in wound area, they determined that 44-percent healing at week 3 correctly predicted 77 percent of outcomes. They did not report the sensitivity or specificity of this. In 2000, Kantor and Margolis evaluated change in wound area and percent change in wound area as measures of initial healing suitable for use as a prognostic index.[8] Choosing failure to heal at 24 weeks as the primary endpoint, they assessed 104 patients with venous leg ulcers. They determined change in area had no predictive value, but percent change in area at weeks 2, 3, 4, and 5 did. Percent change in area at four weeks had the best combination of positive and negative predictive values (68.2%, 74.7% respectively) and the largest area under the receiver-operating curve (ROC) (0.75). They concluded that a wound that increases by three percent or more at week 4 has a 68-percent probability of not healing and that a wound that increases by less than three percent after four weeks has a 75-percent chance of healing by 24 weeks. As they pointed out, with an area under the ROC of 0.75, it is considered acceptable but not good.[8] In late 2002, Gelfand, Hoffstad, and Margolis published a very large cohort study assessing wound healing characteristics as candidate surrogate markers.[9] Using the data base of Curative Health Services, a company that operates more than 150 wound clinics, they identified 56,488 wounds in 29,189 patients for analysis. They found that absolute change in area (Area0 – Areat) and healing rate [(Area0 – Areat)/t] at four weeks of therapy were poor discriminators of eventual healing, but when these parameters were log transformed, their reliability in predicting eventual healing very much improved. Of more importance was the mathematically simpler parameter of percent change in area, which produced relatively equal results. Using a percent change in area of 28.79 at four weeks as a cut point, complete healing of the primary ulcer by 24 weeks could be predicted with a sensitivity of 0.67, a specificity of 0.69, a positive predictive value of 0.80, and a negative predictive value of 0.52. The area under the ROC was 0.80 for predicting healing by 12 weeks and 0.73 for predicting healing by 24 weeks. All together this could correctly predict 68 percent of final outcomes. Only the log ratio was better, and not by much, predicting 69 percent of outcomes. All of these measurements were based on the maximum length and width of the wound. No computer-based planimetric devices were used. Wound Edge Migration In 1987, Bulstrode, Goode, and Scott studied 48 chronic leg ulcers treated with 6 to 14 weeks of bed rest, evaluating a number of different wound parameters for correlation with time to complete healing.[7] To assess epithelial migration, they divided the change in wound area by the change in wound perimeter (?A/?P, where A = area in square centimeters and P = perimeter in centimeters). They reported the rate of epithelial migration did not correlate well with time to complete healing. They did not, however, assess the ability for epithelial migration to predict the more easily assessed, and possibly equally useful, endpoint of complete healing by a particular time, such as 24 weeks. Several years later, Gilman suggested that a similar equation could be used to assess the prognostic power of initial healing rate.[10] He theorized that differences in the sizes and shapes of wounds being compared created distortion in the rates of healing: reduction in area tended to exaggerate the progress of larger wounds, and percent area reduction exaggerated the progress of smaller wounds. Considering that shallow wounds, particularly venous ulcers, primarily heal from the edges inward as a result of both epithelialization and contraction,[11] Gilman chose to use change in wound perimeter to account for the differences in size and shape. He incorporated an equation for changes in wound edge length into the equation for change in area to create a transformed equation that measured the linear advance of the wounds edge (Table 1). Gilman did not apply his equation to an actual set of wounds, but it was not long before others did. In 1991, Pecoraro, et al., was the first to report use of Gilman’s transformed equation. They assessed wound healing rates in 46 diabetic foot ulcers and found that 85 percent of ulcers with a positive healing rate over the first four weeks of treatment eventually healed (p = 0.013).[12] In 1993, Margolis, et al., used Gilman’s equation in healing rates of 27 venous leg ulcers being treated with a standard regimen.[13] They confirmed the validity of Gilman’s equation to correct for differences in size and shape. During the first four weeks of treatment, the average rate of healing for the ulcers that healed was 0.087cm/week versus -0.005cm/wk in those that did not heal. Statistical significance was not reported. All of the ulcers that eventually healed had positive initial healing rates. Of the ulcers that did not heal, there was a wide distribution in initial healing rates, including positive and negative rates. Therefore, a negative initial rate guaranteed a bad outcome and a positive initial rate suggested a good outcome. They suggested that using Gilman’s equation during an observation period encompassing the initial four weeks could predict healing or not with conventional therapy. In 1996, Gorin, et al., used Gilman’s equation to retrospectively assess 49 venous leg ulcers from 39 nonstudy patients in a vascular surgery clinic.[14] They reported a linear healing rate of 0.011cm/d, similar to that determined by Margolis and colleagues. They also assessed wound geometry for correlation with Gilman’s linear transform and found that measurements of linear healing rates of the wound edge were independent of any geometric variable, including length, width, and area. They concluded that linear healing rate, as opposed to measurements of change in wound area, was preferable for comparing wound healing rate in clinical trials. In 1997, Gilman’s transformed equation was further evolved.[5] While using Gilman’s equation to predict the outcome of 15 venous ulcers being treated with standard therapy, Tallman and colleagues observed healing rate instability from week to week. They suspected this might be due to any variety of variables that could destabilize healing (i.e., inadequate compression, concomitant disease) and weaken the prognostic ability of Gilman’s equation. To compensate for this variation, they adjusted Gilman’s equation. Instead of measuring the mean healing rate at each visit in comparison to the initial ulcer size on day 0, they measured it in comparison to the size at the most previous visit; then they calculated the average of all of these measures taken at each visit, creating a “mean-adjusted healing rate.” This method continued to account for variation in perimeter but also now factored in week-to-week fluctuations in healing rate. With this new assessment tool, they determined that at weeks 3 and 4 all ulcers that healed within 24 weeks had healing rates of greater than 0.02cm/wk, whereas those that did not heal by 24 weeks had rates of less than 0.005cm/wk. When using the method by Gilman, a positive healing rate was not as accurate in predicting final healing as with the mean-adjusted healing rate. With Gilman’s equation, p-values for Spearman’s correlation coefficient were >=0.20, but with the mean-adjusted healing rate, they were 0.11cm/wk (p = 0.11cm/wk will, in most cases, predict healing, and rates of =0.075cm/wk makes a good predictor of eventual healing in both types of ulcers and possibly in all types of wounds. This is just for ?A/?P; both Gilman’s equation and the mean-adjusted healing rate, the latter in particular, may actually be better than this. Although initial healing rates measured by these parameters can predict complete healing and thus efficacy, the short time period they are measured in is likely not adequate to demonstrate safety. One would expect it necessary to follow a therapy to complete healing to assuredly confirm safety in a new product. Also, for new therapies, particularly unique ones (e.g., growth factors), it is possible that intermediate markers may not capture the full effect of treatment.[9] Some therapies may prove to give their greatest benefit early or late in the healing process, thus destined to be misrepresented by a surrogate marker. At the least, initial healing rates that reliably predict healing will lend themselves well to pilot studies and phase II trials of therapies being considered for large clinical trials.[5,9] All of the studies the authors reviewed, except for one, used computer-based techniques, mostly planimetry, to make measurements. Even though planimetry is practically universal in wound research, it does prove to be an obstacle for clinical practice because planimetry is not readily transferred to the bedside.[18] As far back as 1994, Margolis was sounding the need for the use of simple rate data in clinical studies that would not just facilitate comparison of outcomes between studies but would also allow easy transfer of findings from clinical research to actual clinical practice. In 1998, Kantor and Margolis determined that for wounds