Flowchart describing the sample derivation. All nonmelanoma skin cancer (NMSC) diagnosed by biopsy at the 2 study sites in 1999 and 2000 were identified by daily review of pathology records at both hospitals. Tumors were not included if the patient was younger than 18 years, if the medical record was protected because the patient was an employee, or if the patient already had had a tumor diagnosed on another date at another time during the study period.
Clark FL, Sahay A, Bertenthal D, Maddock L, Lindquist K, Grekin R, Chren M. Variation in Care for Recurrent Nonmelanoma Skin Cancer in a University-Based Practice and a Veterans Affairs Clinic. Arch Dermatol. 2008;144(9):1148-1152. doi:10.1001/archderm.144.9.1148
To learn if treatment of recurrent nonmelanoma skin cancer (NMSC) varied in different practice settings.
Prospective cohort study of consecutive patients with recurrent NMSC.
A university-based dermatology practice and the dermatology clinic at the affiliated Veterans Affairs Medical Center (VAMC). Conventional therapies for NMSC were available at both sites.
All 191 patients diagnosed as having recurrent NMSC in 1999 and 2000 were included in the study. Data were collected from medical record review and surveys mailed to patients.
Main Outcome Measure
Performance of Mohs micrographic surgery (Mohs).
Patients at the VAMC were older, less educated, poorer, and had more comorbid illnesses, but their tumors were similar to those of patients at the university-based practice. Treatment choices differed at the 2 sites: the proportions of tumors treated in the VAMC and university sites were 60% and 14%, respectively, for excisional surgery; and 24% and 61%, respectively, for Mohs (P < .001). In multivariate analyses adjusting for patient, tumor, and physician features that may have affected treatment choice, tumors treated at the university-based site remained significantly more likely to be treated with Mohs (odds ratio, 8.68 [95% confidence interval, 3.66-20.55]; P < .001).
Substantial variation existed in the treatment of recurrent NMSC in different practice settings. This variation was not explained by measured clinical characteristics of the patients or the tumors.
For most nonmelanoma skin cancers (NMSCs), many therapies can prevent recurrence. We have previously found that for common, lower-risk, primary (ie, nonrecurrent) tumors, treatments varied substantially between university-based and Veterans Affairs (VA) clinical sites that shared many practitioners.1 Even after controlling for multiple variables that might have affected treatment choices, primary tumors at the university site were more than twice as likely as those at the VA site to be treated with Mohs micrographic surgery (Mohs). This variation in care was largely unexplained and may have been due to unmeasured clinical variables, patient preferences, educational incentives, or differences in the practical availability of the therapies at each site. Nonetheless, the variation highlights that insufficient data exist to support evidence-based choices among therapies for most primary tumors and that a clear consensus about optimal therapy in all situations does not exist.2
Recurrent NMSCs pose unique and substantial treatment challenges. Not only is the treatment itself for a recurrent tumor more difficult,3 but treated recurrent tumors have a higher rate of subsequent recurrence than primary tumors, and serious sequelae such as metastasis are more frequent.4 Formal recommendations for treatment of recurrent NMSC include both surgical excision and Mohs as first-line therapy,5 although some authors6,7 believe that existing data justify the recommendation that Mohs be the treatment of choice.
Little is known about how physicians choose different therapies for recurrent NMSCs, and the purpose of this study was to learn if variations exist in the treatment of recurrent NMSCs in different practice settings. Because of the explicit recommendations in favor of Mohs5- 7 we hypothesized that most recurrent tumors would be treated by Mohs and that any variation in therapy between different clinical sites would be less than that found for primary tumors.
We performed a prospective cohort study of consecutive patients with recurrent NMSC diagnosed in 1999 and 2000 and treated at 2 sites, a university-based dermatology practice and the dermatology clinic at the nearby VA medical center (VAMC) affiliated with the university. Details of the study, settings, and data collection have been described elsewhere.1 This study was approved by the institutional review boards of both institutions.
Nonmelanoma skin cancers were defined as basal cell or squamous cell carcinomas and were identified by daily review of pathology records at both hospitals. Only tumors diagnosed at 1 of the 2 sites were included; tumors in patients who had been referred from other institutions solely for therapy were excluded. A dermatopathologist was available at both sites. Herein, the word “clinician” refers to attending physicians, resident physicians, and nurse practitioners. A biopsy-proven NMSC was considered recurrent if it was described in the medical record by the clinician who performed the biopsy as recurrent or previously treated. The flowchart describing the sample derivation is shown in the Figure.
Both clinical sites were staffed by dermatology attending physicians and dermatology resident physicians; at the VAMC site only, dermatologic nurse practitioners also practiced. Most physicians at the VAMC also practiced at the university site.
In general, decisions about therapy were made by the clinician who performed the biopsy. Conventional therapies for NMSC were available at both sites. Of the therapies, tumor destruction by electrodesiccation and curettage (ED&C) was performed by any of the 3 types of clinicians, excisions were generally performed by residents or attending physicians, and Mohs was performed only by attending physicians, all of whom were accredited by the American College of Mohs Micrographic Surgery and Cutaneous Oncology. No additional approvals were required to schedule any of the therapies at either site.
The clinical environment for therapies varied somewhat between sites. Although both sites had designated Mohs suites and histologic technicians, the university site had 3 Mohs surgeons on staff, and 2 to 3 days per week devoted to Mohs; at the VAMC, a single day per week was devoted to Mohs, which was performed by 1 surgeon. At the VAMC site, the Mohs surgeon was an assistant professor who had been on the faculty less than 5 years; this Mohs surgeon also practiced at the university site, along with 2 more senior surgeons who were professors.
Data were obtained from medical records and patient survey. Patients were surveyed by mail before therapy. Baseline health status was measured with an adapted version of the Medical Outcomes Study 12-Item Short Form (SF-12) instrument.8 Comorbidity was measured with a modified version of the Charlson Comorbidity Index.9
We compared patients, tumors, and time interval between biopsy and treatment at the 2 clinical sites. Because delays longer than 6 months were unusual at both sites, for determination of treatment delay we eliminated the 3 tumors with delays of that length or longer. We compared rates of therapies performed at the university and VAMC sites in all patients and in important clinical subgroups. Differences between groups were evaluated using t test or χ2 test analyses; Fisher exact test analyses were used for all 2 × 2 analyses and whenever expected cell sizes were less than 5. To examine differences in therapies between the 2 sites, we used logistic regression to model the performance of Mohs. Independent variables included those that we reasoned might affect choice of therapy based on conventional clinical practices and the published literature. Because the response rate for survey data was approximately 60% (thus limiting the effective sample size), we did not include patient-reported variables in the multivariate analyses. Patient features in the model included age and sex. Tumor characteristics included the tumor type, location in the H zone of the face, location on the head or neck, tumor diameter (>10 mm), and the presence of histologic risk factors for recurrence.1 Care features in the model included the practice site and level of physician training (attending vs resident). Although interaction terms, usually including site of treatment, were included in the initial model, none was preserved after using a backward elimination strategy to develop the final model.
Over the 2 years of the study, 210 recurrent NMSCs were diagnosed in 191 consecutive patients. A total of 91 recurrent tumors in 81 patients were diagnosed at the VAMC site, and 119 recurrent tumors in 110 patients were diagnosed at the university site. No tumor was evaluated and treated at different sites.
The 210 recurrent tumors were diagnosed by 38 different clinicians. Five clinicians practiced exclusively at the VAMC, 15 clinicians practiced exclusively at the university site, and 18 clinicians practiced at both sites. At the VAMC site, 31 tumors (34%) were biopsied by an attending physician, 38 tumors (42%) by a resident physician, and 21 (23%) by a nurse practitioner (n = 90; data for 1 patient were missing). At the university site, 101 tumors (86%) were biopsied by the attending physician, and 17 (14%) by a resident physician (n = 118; data for 1 patient were missing).
Compared with patients treated at the university site, those treated at the VAMC were older, more likely to be male, and more likely to be poor and less educated. They also reported more comorbid illnesses and worse physical health status. Tumor characteristics were similar between the 2 sites (Table 1 and Table 2).
Compared with recurrent tumors treated at the VAMC, recurrent tumors treated at the university site were more likely to be treated with Mohs (60.5% vs 24.2%; P < .001) and less likely to be treated with excision (14.3% vs 60.4%; P < .001) (Table 3). This variation in treatment existed in multiple clinically important subgroups. For example, tumors located in or outside of the H zone of the face and both basal cell and squamous cell carcinomas were all more likely to be treated with Mohs at the university site than at the VAMC site (Table 4).
Overall, the median time interval from biopsy to treatment was shortest for ED&C (29 days), followed by excision (35 days), and longest for Mohs (50 days). The median time interval was longer at the VAMC than at the university site for ED&C (47 days vs 21 days, respectively; P = .07) and for excision (50 days vs 16 days, respectively; P = .01) but was similar for Mohs (51 days vs 50 days, respectively; P = .79).
In multivariate logistic regression models controlling for patient age, sex, tumor histologic type, size, location on the head or neck, location in the H zone of the face, presence of histologic risk factors for recurrence, and level of physician training (attending vs resident), tumors treated at the university site remained much more likely to be treated with Mohs than tumors treated at the VAMC (odds ratio [OR], 8.68; 95% confidence interval [CI], 3.66-20.55). Important explanatory variables for the performance of Mohs were age (OR for a 10-year increase, 1.44; 95% CI, 1.06-1.95), basal cell carcinoma (OR, 5.40; 95% CI, 2.11-13.77), tumor diameter greater than 10 mm (OR, 3.23; 95% CI, 1.33-7.86), and tumor location on the head or neck (OR, 17.47; 95% CI, 6.14- 49.73). Sex, location in the H zone of the face, presence of histologic risk factors for recurrence, level of physician training, and interaction terms were not independently related to the choice of Mohs.
Because recurrent NMSCs are more difficult to treat and more likely to recur than primary tumors, many experts have recommended that Mohs be the treatment of choice for recurrent tumors,6,7 and we predicted that therapy selection would be similar at different sites. Instead, we found substantial and consistent differences in the treatment of recurrent NMSC at 2 affiliated academic clinics. Recurrent tumors treated at the university site were significantly more likely to receive Mohs than those at the VA site even after adjusting for many patient, tumor, and physician characteristics that might influence treatment (P <.001).
Our data do not fully explain this finding. Patients at the 2 sites differed markedly, and unmeasured patient characteristics may have contributed to the variation in therapy. For example, patient preferences may have been important in affecting treatment choice. Patients at one site may have believed that therapies differ in costs, benefits, and/or effects on many health outcomes and may have requested one of the therapies based on these beliefs. We did not measure patient preferences or the extent to which clinicians included patient preferences in their treatment choices. Thus, we do not know whether university and VAMC patients differed systematically in their preferences for treatments and whether any differences may have contributed to the variation in treatments. Similarly, although the multivariate models adjusted for a variety of potentially important tumor features, unmeasured tumor characteristics may also have contributed to the variation in treatment.
Clinicians at the 2 sites had different incentives to choose different therapies, which may have contributed to the variation in care. More clinical encounters at the university site than at the VAMC were reimbursed on a fee-for-service basis, but clinicians at both sites were salaried, so their incomes were not directly related to their treatment choices. Moreover, for most tumors, the diagnosing and treating clinicians were different, so that treatment decisions were not made by the clinician who ultimately performed the procedure. Thus, direct financial incentives seem unlikely to have accounted for the difference. Some of the patients seen at the university site may have been referred to the practice for diagnosis and therapy, and there may have been an expectation by the patient or referring physician that Mohs was to be performed. Finally, residents (who treated proportionately more tumors at the VAMC site than at the university site) may have had educational incentives to recommend excision (which they could perform, and thus obtain experience with the procedure).
Finally, we do not have a specific measure of access to care at the 2 sites. On the one hand, all therapies were available at both sites, and the fact that there was no difference in median time interval between biopsy and Mohs at the 2 sites does not suggest that Mohs was less available at the VAMC. On the other hand, even though the 2 sites had similar time intervals to treatment, the university site had 2 Mohs surgeons, whereas the VA had only 1 part-time Mohs surgeon. In addition, a third Mohs surgeon was recruited to the university site soon after the study period to improve the availability of Mohs at this site. Thus, there may have been reduced access to Mohs at the VAMC in a practical sense because of the striking difference in clinician availability at the 2 sites, which may have resulted in fewer attempted referrals for Mohs at the VAMC.
These and other potential limitations to this study are listed in Table 5. The study was limited to 1 city and 1 academic program, which may not be typical of other locations. Also, the sample size is relatively small, which limited the adjustments that could be made in the multivariate analyses. The major finding is consistent in multiple clinically important subgroups, however, suggesting that the basis for the difference in care is probably not entirely clinical and may be related to systematic differences in patient preferences, educational incentives, or availability of Mohs.
We have no evidence that the quality of care varied at the 2 sites. Regardless of the explanation for the variation in care, these data further highlight a lack of consensus about a single preferred treatment for recurrent NMSC2 and the fact that rigorous data do not exist to inform such a consensus. Early results of a recent randomized trial10 for recurrent basal cell carcinomas on the face demonstrated no statistically significant difference in tumor recurrence at 18 months after excision or Mohs. These findings have been controversial,11 however, and are too preliminary to change practice guidelines. Mohs was the more costly therapy in that European study,10 which illustrates that determining the comparative effectiveness of the treatments is important. Overall, our findings emphasize that the care of this highly prevalent condition of older patients warrants increased scrutiny. Determining optimal treatment for these tumors will require a longitudinal study of outcomes of different therapies to provide a basis for formal treatment guidelines that can be adopted more broadly.
Correspondence: Mary-Margaret Chren, MD, San Francisco VAMC 151R, 4150 Clement St, San Francisco, CA 94121 (firstname.lastname@example.org).
Accepted for Publication: September 20, 2007.
Author Contributions: All authors had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Chren. Acquisition of data: Sahay, Maddock, and Chren. Analysis and interpretation of data: Clark, Bertenthal, Lindquist, Grekin, and Chren. Drafting of the manuscript: Clark and Chren. Critical revision of the manuscript for important intellectual content: Clark, Sahay, Bertenthal, Maddock, Lindquist, Grekin, and Chren. Statistical analysis: Clark, Bertenthal, Lindquist, and Chren. Obtained funding: Chren. Administrative, technical, and material support: Clark, Sahay, and Maddock. Study supervision: Chren.
Financial Disclosure: None reported.
Funding/Support: This work was supported by Investigator-Initiated Research (IIR) grants 97010-2 and 04-043-3 from the Health Services Research and Development Service of the Department of Veterans Affairs, and by an Independent Scientist Award (K02 AR 02203) and Midcareer Investigator Award (K24- AR052667) from the National Institute of Arthritis and Musculoskeletal and Skin Diseases, National Institutes of Health.