Chen T, Bertenthal D, Sahay A, Sen S, Chren M. Predictors of Skin-Related Quality of Life After Treatment of Cutaneous Basal Cell Carcinoma and Squamous Cell Carcinoma. Arch Dermatol. 2007;143(11):1386-1392. doi:10.1001/archderm.143.11.1386
To identify predictors of skin-related quality of life (QOL) after treatment of nonmelanoma skin cancer (NMSC).
Prospective cohort study of consecutive patients with NMSC diagnosed in 1999 and 2000.
University-affiliated private practice and a Veterans Affairs clinic.
A total of 633 patients who responded to a questionnaire before treatment.
Main Outcome Measure
Skin-related QOL, measured with the 16-item version of Skindex-16, a validated measure. Skindex-16 scores vary from 0 (best QOL) to 100 (worst QOL) and are reported in 3 domains: symptoms, emotional effects, and effects on functioning.
Controlling for treatment group, the strongest independent predictor of skin-related QOL after treatment of NMSC was pretreatment skin-related QOL. Other patient characteristics that predicted better QOL included less comorbidity and better mental health status. No tumor or care characteristic (including location of tumor, size of tumor, site of therapy, or training level of treating clinician [attending physician, resident, or nurse practitioner]) was found to predict better skin-related QOL after treatment of NMSC.
Patients with better pretreatment skin-related QOL, less comorbidity, and better mental health status had better skin-related QOL after treatment of NMSC. These findings may be useful for pretreatment assessment and counseling.
Health-related quality of life (QOL)—the impact of illness on a patient's functioning and physical, psychological, emotional, and social well-being—is a crucial outcome in cancer research and clinical care, particularly for typically nonfatal cancers for which alternative therapies exist. Understanding QOL effects of nonfatal cancers is important for informing clinician-patient interactions and decisions about choice of therapy (hereinafter, “clinician” refers to attending physicians, residents, and nurse practitioners).
Cutaneous basal cell carcinoma (BCC) and squamous cell carcinoma (SCC) are typically nonfatal, but lesions grow locally and can jeopardize adjacent structures such as eyes or ears. Treatment for these cancers is common and expensive, among the top 5 most costly cancers to treat within the Medicare population.1 Moreover, the risk of subsequent skin cancer approaches 50% in the 5 years after a tumor is diagnosed,2 highlighting that the care of patients with skin cancer is often lifelong. However, despite the prevalence and chronic nature of these cancers, their effects on patients' experiences are poorly understood.
We have previously shown that skin-related QOL was similarly improved after Mohs surgery and excision and was unchanged after tumor destruction.3 We used the skin-related QOL measure Skindex-16 (hereinafter, Skindex) and determined that patients treated with excision or Mohs surgery improved in all QOL domains, with approximately 10-point and 20-point improvements in symptoms and emotional effects, respectively (improvements that are clinically meaningful).3 The purpose of this study was to determine pretreatment patient, tumor, and care characteristics that are associated with better skin-related QOL after treatment for nonmelanoma skin cancer (NMSC), controlling for treatment type. Based on our clinical impressions and previous studies, we hypothesized that skin-related QOL after treatment would be related to pretreatment QOL and would be better in men, in patients with smaller tumors, in patients with tumors that were not located on the head or neck, and in patients treated in a private setting compared with those treated at a Veterans Affairs (VA) Medical Center.
This study was part of an ongoing prospective study of consecutive patients with cutaneous BCC or SCC diagnosed in 1999 and 2000 and treated at a university-affiliated private dermatology practice or the dermatology clinic at the VA Medical Center affiliated with the university.4 The study was approved by the institutional review boards of both institutions.
All treatments were available at both sites. Decisions regarding treatment choice were made by the clinician who performed the biopsy after histopathological results were known. The most common treatments were electrodesiccation and curettage (EDC), excision, and Mohs surgery. Residents and attending physicians performed biopsies at both sites; in addition, nurse practitioners also performed biopsies at the VA Medical Center. Most often, the treating clinician was not the same clinician who performed the biopsy. All clinicians performed EDC; residents and attending physicians performed tumor excision. Mohs surgery was performed only by attending physicians.
Patients with BCC or SCC (including SCC in situ) were identified by daily review of dermatopathology records at both hospitals. All patients were considered eligible for inclusion in the study; however, individuals were excluded if they were younger than 18 years, if the medical record was protected because the patient was an employee, or if the patient had already been diagnosed with a prior BCC or SCC during the study period. Because recurrent cancers are believed to be at higher risk for further recurrence, we eliminated from the sample 222 patients whose tumors were described in the medical record as “recurrent” or “possibly recurrent.”
Patients with tumors were enrolled if they had a current mailing address and if they had responded to a pretreatment questionnaire about their health and QOL. If a patient had multiple tumors, he or she was asked to respond only about the effects related to the most bothersome lesion. The 633 enrolled patients were similar to all patients with BCC or SCC diagnosed during the study period in terms of age; tumor histological type, diameter, and location on the body; and type of treatment received. Compared with those not enrolled, however, enrolled patients were more likely to be male (77% vs 69% of those not enrolled; P = .002), to have had a history of previous BCC or SCC (54% vs 47%; P = .02), to have been treated at the VA Medical Center (49% vs 31%; P < .001), and to have had the tumor biopsy performed by a nurse practitioner (19% vs 12%; P = .001).3
Data from medical records and from mailed patient surveys were gathered by trained research staff. Follow-up surveys were responses at 12, 18, or 24 months after therapy. We defined skin-related QOL after treatment as the mean of the subscale scores at 12, 18, or 24 months because QOL stabilizes at 12 months after therapy.3 The numbers of patients who completed the questionnaire at 12, 18, and 24 months after therapy were 447, 422, and 386, respectively.
The age and sex of patients were recorded from medical records. Data about QOL, health status, comorbid illnesses, and sociodemographic variables such as marital status, race, education, employment, and income were obtained from patient surveys. For this study, we explicitly defined QOL as skin-related QOL measured with Skindex, a valid and reliable tool.5 The version used in this study was adapted to inquire about QOL effects of the skin cancer and its treatment. Skindex scores range from 0 to 100 (worst effect on QOL) and are reported as 3 subscores: symptoms, emotional effects, and effects on functioning. We also calculated a composite QOL score as the mean of the 3 Skindex subscale scores. Based on our previous work with patients with BCC or SCC, the minimal clinically meaningful difference in all Skindex subscales is 10 points.3 Health status was measured with an adapted version of the 12-Item Short-Form Health Survey (SF-12) instrument.6 Scores on the SF-12 are reported as a Physical Component Summary (PCS) score and a Mental Component Summary (MCS) score; a higher score reflects a better health status, and the median value is designed to be 50 points. Comorbid illnesses were recorded using an adapted version of the Charlson Comorbidity Index.7,8
Data obtained from the medical records included the histologic type, location (specifically, whether the tumor involved the head and neck and areas that are more difficult to treat, such as the area of midface and ear known as the H-zone9), and the diameter of the tumor, as well as the presence of histologic risk factors for recurrence (ie, whether the tumor was infiltrative, morpheaform, sclerosing, adenoid, poorly differentiated, acantholytic, or neurotropic).10,11
Data included the site (VA Medical Center or private hospital), type of treatment, and the training level of the treating clinician (attending physician, resident, or nurse practitioner).
We first described patient, tumor, and care characteristics in patients grouped according to tertiles of pretreatment composite QOL scores. Differences among the groups were evaluated using χ2 or Fisher exact tests if the variable was dichotomous and with analysis of variance if the variable was continuous.
Next, we determined the bivariable association of pretreatment patient, tumor, and care characteristics with follow-up skin-related QOL, while adjusting for pretreatment Skindex scores. Because Skindex scores are not normally distributed, conventionally calculated P values are not appropriate to assess statistical significance. Instead, we calculated P values with permutation tests,12 using 40 000 random permutations to ensure that the P values are correct to at least 2 decimal places. (For the bivariable analyses only, we divided age and diameter into categorical variables in order to have groups for comparison. We chose the age of 65 years because it is the typical age for Medicare eligibility, and chose a tumor diameter of 10 mm because tumors with diameters greater than 10 mm are conventionally regarded as larger.) We performed these calculations for composite Skindex scores, as well as each of the 3 Skindex subscales.
Furthermore, to determine the independent association of pretreatment characteristics with QOL after therapy, we performed a sequential regression analysis by creating multivariable linear regression models in a stepwise fashion, including first the baseline QOL, followed by patient, tumor, and care characteristics in subsequent steps. In these models, we adjusted for treatment type and incorporated a random effect for treating clinician. Dependent variables were the composite and subscale Skindex scores. Independent variables were those significant (P ≤ .10) characteristics in bivariable analyses, type of treatment, and other characteristics that we reasoned may be important: pretreatment QOL, age, sex, histologic type of tumor, and tumor diameter. To accommodate the nonstandard distribution of the outcome, Skindex scores, we computed confidence intervals for the regression coefficients using nonparametric bootstrap analyses with 40 000 repetitions.12P values for the hypothesis tests were obtained by inverting the confidence interval obtained from the bootstrap procedure.12 We used Stata statistical software (release 9.2; StataCorp, College Station, Texas) for all statistical analyses.
Pretreatment QOL was associated with several patient, tumor, and care characteristics (Table 1). Patients who were married, who had completed at least some graduate school or professional school education, who had an annual income of greater than $30000, who had fewer comorbidities, or who had better physical and mental health status had better skin-related QOL. Before treatment, QOL was not found to correlate with age, race, or history of a previous BCC or SCC.
Regarding characteristics of the tumor itself, patients with BCCs, tumors not located on the head or neck, and smaller tumors had notably better pretreatment QOL. With respect to care characteristics, patients who were treated at the private hospital had better QOL before treatment vs those treated at the VA Medical Center.
Therapies used to treat skin cancers of patients in the cohort were EDC in 132 cases (21%), surgical excision in 245 (39%), Mohs surgery in 238 (37%), and a variety of other, less common therapies in 18 (3%). Of the 633 enrolled patients, 514 (81%) responded at 12, 18, or 24 months after therapy. Patients who responded were similar to those who did not respond following treatment in most patient, tumor, and care characteristics. Compared with those who responded, those who did not respond were, however, more likely to have been separated, divorced, or widowed (42.4% vs 31.5% of those who responded; P = .02), to have more comorbidities (Charlson Comorbidity Index, 2.8 vs 2.0; P = .01), to have smaller tumors (9.6 vs 10.3 mm; P < .001), to have been treated with therapy other than EDC, excision, or Mohs surgery (5.8% vs 2.3%; P = .04), and to have been treated at the private hospital (59.7% vs 50.7%; P = .04).
After adjusting for pretreatment QOL, relatively few characteristics were related to QOL after treatment (Table 2). With respect to patient characteristics, those who considered themselves “white (not of Hispanic origin),” with annual incomes greater than $30 000, with fewer comorbidities, or with better physical or mental health had better QOL after therapy. Tumors located on the head and neck were associated with better QOL after therapy. No other tumor characteristic, however, was associated with better QOL after therapy. Finally, the treatment site (VA Medical Center or private hospital) was not associated with QOL after treatment. Similar results were found in analyses of each Skindex subscale.
Adjusting for treatment, better pretreatment QOL, white race, fewer comorbidities, and better mental health status independently correlated with better QOL after therapy. No other patient, tumor, or care characteristic was independently associated with QOL after treatment. Similar results were found in multivariable analyses of each Skindex subscale.
Table 3 provides the coefficients and probabilities from the sequential regression analysis and describes the overall relationship of pretreatment characteristics and QOL after therapy in terms of change in the level of each characteristic. For an improvement in pretreatment Skindex score of 10 points, the Skindex score after therapy improved roughly 5 points. For every 10-point improvement in pretreatment SF-12 MCS scores, the Skindex score after therapy improved approximately 3 points, and for every improvement in the pretreatment Charlson Comorbidity Index of 1 point, the Skindex score after therapy improved approximately 1 point. Also, white race was associated with an improvement in Skindex score after therapy of roughly 7 points (P = .05), although the confidence limits around this estimate were wide and the variable was statistically significant only when the model included all patient, tumor, and care characteristics.
We found that, controlling for treatment group, the strongest independent predictor of skin-related QOL after therapy for BCC and SCC was skin-related QOL before therapy. Mental health status, comorbidity, and race were the only other independent predictors of QOL. Given that the minimal clinically meaningful change in Skindex scores is 10 points,3 the actual predictive strength of changes in pretreatment mental health status and comorbidity is quite low. For skin-related QOL, a 20-point difference before treatment would predict a clinically significant improvement after treatment.
Correlates of pretreatment skin-related QOL were consistent with previous work about QOL in general.13 For example, patients who were married, more educated, and with more financial advantages presented with better skin-related QOL than others, as did patients with fewer comorbidities and better physical and mental health status. Similarly, patients with tumors that were smaller or not on the head and neck had better pretreatment QOL. These findings were consistent with our clinical impressions and also illustrate the validity of the QOL measurement because patients responded to the instrument in clinically sensible ways.14
Surprisingly, however, few characteristics were found to predict QOL after treatment. Sex, for example, was not related to QOL, and, similarly, although they were related to QOL before treatment, marital status, education, and income were not independent predictors of better skin-related QOL after treatment. Similarly, despite several important correlates in bivariable analyses, no tumor characteristic was an independent predictor of skin-related QOL after treatment.
An early prospective study15 using the Dermatology Life Quality Index reported that BCC had little effect on QOL scores and found no association between QOL and tumor size after treatment. Our previous work3 has shown that QOL after treatment of BCC or SCC improves in most patients. Compared with the study that found no change in QOL,15 our studies have a larger sample size and a longer duration of follow-up and used a more comprehensive measure of QOL. Our current findings are similar, however, in that tumor size was unrelated to QOL after treatment in both studies.
In a more recent prospective study of 121 patients with cervicofacial skin cancer, Rhee et al16 noted little variation in QOL scores 4 months after treatment of NMSC using the Medical Outcomes Study Short Form 36-item Health Survey (SF-36) and the Functional Assessment of Cancer Therapy-General (FACT-G) measures. However, they found an improvement in the SF-36 mental health subscale as well as in the FACT-G emotional well-being subscale, particularly for employed patients younger than 65 years.16 In our larger study with longer follow-up and an instrument that specifically measures skin-related QOL, however, we found that age and employment did not predict skin-related QOL after treatment.
These investigators also found no demographic or tumor characteristics to be associated with change in QOL.17 Our results are similar in that, with the exception of race, we found that no demographic or tumor characteristic independently predicted better QOL. Furthermore, our finding that QOL after treatment is most strongly associated with a patient's well-being and health status before therapy and not with features of the skin cancer itself is supported by studies of other conditions.18- 20
The finding that white race was associated with better QOL after treatment of BCC and SCC is difficult to interpret. Most of the sample was white (94%), and the total number of nonwhite patients was very small (n = 37). The confidence interval around the point estimate in the multivariable model is also wide, indicating substantial uncertainty about the value, and the variable did not reach statistical significance (P = .05) until the final model in sequential regression analyses. Although nonwhite patients have been shown to have worse QOL in other conditions,21,22 our study does not provide enough information to validate or understand the effect of race on skin-related QOL after treatment of NMSC.
Because this study was conducted at 2 hospitals within the same city, the results may not be generalizable to all patient populations. However, the participants are typical of patients with NMSC nationwide.23 In addition, the study was limited only to those who had responded to a pretreatment questionnaire. These patients were more likely to have been male, treated at the VA Medical Center, and to have a history of previous BCC or SCC. None of these variables, however, was related to QOL after treatment in bivariable and multivariable analyses.
In summary, adjusting for treatment type, better skin-related QOL before treatment was the most consistent and strongest independent predictor of better skin-related QOL after therapy for NMSC. Other independent predictors of better QOL after treatment for NMSC were better mental health status, less comorbidity, and white race. Tumor characteristics were not related to QOL outcomes. These results may be useful for pretreatment assessment and counseling because they help clinicians identify patients who are at higher risk for poor QOL outcomes. For example, clinicians may focus on improving mental health status (ie, treating depression) or improving the status of comorbidities prior to therapy. Further studies should address whether QOL outcomes can be improved by interventions that target remediable pretreatment characteristics.
Correspondence: Mary-Margaret Chren, MD, San Francisco VAMC 151R, 4150 Clement St, San Francisco, CA 94121 (firstname.lastname@example.org).
Financial Disclosure None reported.
Accepted for Publication: April 5, 2007.
Author Contributions:Study concept and design: Chen, Sahay, and Chren. Acquisition of data: Bertenthal, Sahay, and Chren. Analysis and interpretation of data: Chen, Bertenthal, Sen, and Chren. Drafting of the manuscript: Chen and Chren. Critical revision of the manuscript for important intellectual content: Bertenthal, Sahay, Sen, and Chren. Statistical analysis: Bertenthal, Sen, and Chren. Obtained funding: Chren. Administrative, technical, and material support: Bertenthal. Study supervision: Sahay and Chren.
Additional Contributions: Leah Maddock, MPH, contributed data collection and management.