Proportion of incomplete basal cell carcinoma (BCC) and squamous cell carcinoma (SCC) excisions across the 15 clinics by total number of excisions performed.
Proportion of incomplete basal cell carcinoma excisions by physician grouped by clinic, ordered by the total number of basal cell carcinoma excisions performed. At the top are the number of excisions performed per physician and clinic (the numbers going down match the physicians going left to right within each clinic).
Proportion of incomplete squamous cell carcinoma excisions by physician grouped by clinic, ordered by the total number of squamous cell carcinoma excisions performed. At the top are the number of excisions performed per physician and clinic (the numbers going down match the physicians going left to right within each clinic). The physician identification number matches up with the physician identification number assigned in Figure 2.
Craig Hansen, David Wilkinson, Mary Hansen, H. Peter Soyer. Factors Contributing to Incomplete Excision of Nonmelanoma Skin Cancer by Australian General Practitioners. Arch Dermatol. 2009;145(11):1253–1260. doi:10.1001/archdermatol.2009.270
To study rates of incomplete excision of basal (BCC) and squamous (SCC) cell cancer by Australian general practitioners with a special interest.
A network of 15 primary care skin cancer clinics across Australia.
Fifty-seven physicians performing excisions of 9417 BCCs and SCCs in a single network of 15 primary care skin cancer clinics across Australia between 2005 and 2007.
Main Outcome Measures
Rates of incomplete excision according to physician, clinic, anatomic location of the lesion, and whether a previous biopsy had been performed.
Four hundred forty-three of 6881 BCCs (6.4%) and 159 of 2536 SCCs (6.3%) were excised incompletely. Incomplete BCC and SCC excisions were more frequent on the head and neck (282 of 2872 excisions [9.8%] and 97 of 861 [11.3%], respectively) than elsewhere. Ears (74 of 388 excisions [19.1%]) and nose (78 of 546 [14.3%]) had the highest rates of incompletely excised BCCs, and ears (26 of 144 excisions [18.1%]) and forehead (20 of 157 [12.7%]) had the highest rates of incompletely excised SCCs. Of all BCC excisions, 67.3% were once-off excisions with no previous biopsy, and these excisions were more likely to be incomplete (odds ratio, 1.73; 95% confidence interval, 1.36-2.20) than those with a previous biopsy. There was, however, substantial variation in frequency of incomplete excision between clinics for BCC (ranging from 3.3% to 24.7%) and SCC (ranging from 0% to 17.2%) and between physicians within clinics (BCC ranging from 0% to 31.1%, and SCC ranging from 0% to 23.5%).
Overall frequency of incomplete excision is low and similar to that in other reports. However, high frequency in high-risk sites, low rates of previous biopsy, and substantial variation in performance between physicians and clinics suggests there is significant opportunity to further improve health outcomes.
Skin cancer is the most common cancer in Australia.1 Nonmelanoma skin cancer (NMSC) is the most commonly diagnosed cancer in Australia,2 with basal cell carcinoma (BCC) accounting for approximately 70% to 80% of malignant neoplasms and squamous cell carcinoma (SCC) accounting for approximately 15% to 30%.3,4 Recent surveys in Australia have shown that the incidence of BCC and SCC is on the rise3; despite the low mortality rate associated with these cancers,5 NMSC is the most costly cancer to treat in Australia, amounting to $264 million for the year 2001.6 Every year in Australia more than 1 million patient consultations for skin cancer are performed by general practitioners (GPs).5
A recent study reported that Australian GPs perform more excisions on NMSC than specialists, and the excision rate has increased significantly more among the GPs.7 This increase is largely due to GPs with a special interest (GPwSIs) becoming an established feature of the primary health care system, particularly in Australia and the United Kingdom. In the United Kingdom, GPwSIs have a specific role in the area of dermatology and skin cancer. Whereas in the United Kingdom GPwSIs are integrated into local care networks and are required to follow guidance provided by the National Institutes of Clinical Excellence, in Australia there is no such integration or guidance, and skin cancer clinics typically work in relative isolation.
Studies have shown that the rate of incomplete excision for NMSC is higher among GPs than specialist consultants.8,9 Therefore, concerns have been raised by specialist consultants (eg, dermatologists and plastic surgeons) about the quality of care in these skin cancer clinics largely serviced by GPwSIs; however, few data are available to validate these concerns. For example, although a handful of retrospective clinical audits have been carried out on the rate of incomplete excisions of NMSC, the sample size within these studies has been small.8,10- 12
It is vital to critically appraise the quality of care provided by novel models of health service as they emerge. Therefore, the purpose of our study was 2-fold. First, we sought to determine the overall and anatomic site–specific rates of incomplete excision of BCCs and SCCs among a single corporate network of skin cancer clinics. Second, we sought to determine how much variation in this measure there was between clinics and between physicians.
The clinics studied composed a single network of skin cancer clinics staffed by GPwSIs across most states in Australia. In this study, GPwSIs was defined as GPs who have chosen to spend all or most of their time working in primary care skin cancer. Their training consisted of a series of 10 modules that covered the basics of diagnosis and management of skin cancers, policies and procedures in the clinics, and how to bill correctly. The training was not developed by any academic institution but rather by a “registered training organization” with input from some experienced physicians in the clinics. All “certification” was internal to the clinics.
In contrast, in Australia the Dermatology College Training Programme (4 years, full time) is fully accredited by the Australian Medical Council. The program, outlined in detail in the College's Training Program Handbook, is very well structured and comprehensive, and malignant neoplasms of the skin represent an integral part. At the end of the training period, trainees are competent to accurately assess and diagnose all skin lesions with which they may be confronted in practice. They have gained experience in and become competent in dermoscopy, and they are competent in the performance of basic procedures to deal with the skin lesions encountered.
In Australia there are 4 or 5 relatively large chains of skin cancer clinics, with the largest having about 30 clinics. The chain we studied was the third largest at the time of the study.
Although owned by a single corporate structure, the clinics within the chain we studied were typically widely dispersed and the staff working within them had limited contact, other than by e-mail, telephone, and an annual meeting. There were no corporate instructions on how to deliver services or what treatments to use, and training was provided by informal orientation as well as a more formal in-house training program. Physicians worked as independent subcontractors in the clinics and, although there was no requirement to use a specific pathology provider, most physicians used a service that was partly owned by the company that owned the clinics. Since our data collection was completed, the clinics and pathology service have been sold to a larger national health care company.
We obtained 31 117 de-identified skin pathology reports for the period February 25, 2005, to March 30, 2007. Each pathology report included detailed information for each lesion in the form of statements of clinical, macroscopic, and microscopic findings and conclusion. Other information included the date of the clinical visit, patient sex, patient age, patient postal code, physician ID, clinic ID, and clinic postal code.
The electronic pathology reports came in the form of unstructured text, and several steps were carried out to establish categories for the final analyses. First, the text for each lesion within each report was extracted to create an individual record for each lesion. Then various categories were established by scanning the text within each record for terminology pertaining to the type of lesion (eg, final diagnosis), the type of procedure performed (eg, punch biopsy or excision), the anatomic site and orientation, and surgical margins. To identify the lesions that had a follow-up procedure (eg, final excision after an initial biopsy), we matched lesion reports by patient (including age and sex), diagnosis, anatomic site and orientation, type of procedure, and dates of visits. All BCC and SCC excisions were then selected for the final analyses. This research was approved by the University of Queensland Behavioral and Social Sciences Ethical Review Committee.
We report the proportions of incomplete margins for BCC and SCC separately. They are first reported as percentages for each specific anatomic site and then as percentages within the aggregated anatomic sites of “head and neck,” “trunk,” “arms,” and “legs.” We then assessed whether patient age, patient sex, anatomic site, and type of excision (once-off vs final excision after biopsy) were predictors of incomplete margins. Odds ratios and 95% confidence intervals (CIs) for an incomplete margin were calculated by means of a generalized estimating equation. The generalized estimating equation was used to control for the repeated data (eg, multiple lesions per patient) during analysis with PROC GENMOD in SAS Version 9.1 (SAS Institute Inc, Cary, North Carolina) with an exchangeable correlation matrix. The models included incomplete margins (yes/no) as the dichotomous dependent variable and patient sex, grouped patient age based on quartiles (≤50, 51-60, 61-70, and >70 years), anatomic site, and type of excision (initial vs follow-up) as the independent variables.
To assess for any variation across clinics, further models included an indicator variable for clinic. Arbitrarily, the clinic with the most excisions performed was chosen as the reference category in the models. In addition, the clinic and patient postal codes were linked to an index of socioeconomic status created by the Australian Bureau of Statistics at a postal code level.13 This index was then grouped into quartiles and also entered into the regression models. However, when patient and clinic socioeconomic status was included in the regression model, neither was associated with incomplete margins of BCC and SCC, and therefore they were not included in the final models.
We identified 7058 BCC excisions. In 177 (2.5%) there was no report on margin completeness, leaving 6881 BCC excisions for analysis. Of these, 2248 (32.7%) were the final excision that followed an initial biopsy, and 4633 (67.3%) were once-off excisions that did not have a previous biopsy. The 6881 BCCs came from 4428 patients (2809 of them male [63.4%]) examined by 52 different physicians from 15 different clinics. Mean patient age was 60 years (range, 11-98 years), and females were 2 years younger than males (P < .001). Overall (Table 1), 67.7% of the BCCs were excised from men, with most from the head and neck (41.7%) followed by the trunk (35.6%).
In all, 443 of 6881 BCC excisions were incomplete (6.4%), with the head and neck having a significantly larger percentage incomplete (282 of 2872 excisions [9.8%]) than other anatomic areas. More specifically (Table 1), the ears and nose had the highest rate of incomplete BCC excisions (74 of 388 [19.1%]).
After controlling for all other factors, results from the regression analyses (Table 2) showed that women had a 28% (odds ratio [OR], 1.28; 95% CI, 1.03-1.59) increased risk of having an incomplete excision compared with males, patients older than 70 years had a 52% (OR, 1.52; 95% CI, 1.13-2.04) increased risk of having an incomplete excision compared with those 50 years or younger, and, overall, if the excision was not preceded by a biopsy, there was a 73% (OR, 1.73; 95% CI, 1.36-2.20) increased risk that it would be incomplete. When stratified by anatomic site, patient sex was no longer significantly associated with incomplete margins. In other stratified analyses, patient age was significant only among the head and neck excisions and, if the excision was performed on the arms, the once-off excisions with no previous biopsy were 3 times more likely to be incomplete than were final excisions that had a previous biopsy (OR, 3.13; 95% CI, 1.07-9.16).
There were 2639 SCC excisions identified, of which 103 (3.9%) had no report on margin completeness, leaving 2536 SCC excisions for analysis. Of these, 545 (21.5%) were the final excision that followed an initial biopsy, and 1991 (78.5%) were once-off excisions that did not have a previous biopsy. These lesions came from 1881 patients (1209 of them male [64.3%]) treated by 55 different physicians from 15 different clinics. Mean patient age was 65 years (range, 23-95 years), and females were 3 years older than males (P < .001).
Table 3 shows the anatomic distribution of the SCC excisions; 34.0% were from the head and neck, 34.0% from the arms, and 21.6% from the legs. This is much different from the distribution of BCCs (Table 1). The proportion of SCC excisions reported as incomplete was 6.3% (the same as for BCCs), and the proportion was highest for the head and neck (11%). Specifically, ears and eyes/brow had the highest rates of incomplete SCC excisions (Table 3).
In contrast to BCC excisions, results from the regression model showed that, except for the head and neck site, which was almost 3 times more likely than the trunk to have incomplete SCC excisions (OR, 2.89; 95% CI, 1.52-5.51), none of the other variables studied were associated with incomplete margins among SCC excisions (Table 2).
Figure 1 shows the proportion of incomplete excisions across the 15 clinics ordered by the total number of excisions performed at each clinic. There was considerable variation across the 15 clinics (3.3%-24.7% for BCCs and 0%-17.2% for SCCs). The same clinic (clinic F) was responsible for the highest rate of incomplete BCC and SCC excisions, and in the regression model (results not shown) this clinic was 8 times more likely than the referent clinic to have incomplete margins.
Figure 2 and Figure 3 show the proportion of incomplete margins for BCC and SCC, respectively, per physician within each clinic. There was considerable variation in the rate of incomplete margins for BCC (ranging from 0% to 31.3%) and SCC (ranging from 0% to 23.5%) for physicians within clinics. This variation was not accounted for by differences in the number of tumors excised, or the proportion excised from different anatomic sites. These data demonstrate clearly that there was a considerable problem with physician 2 in clinic F and physician 5 in clinic L, as both had among the highest rates of incomplete BCC and SCC margins.
Our study has 2 key messages. The first relates to the rate of incomplete excision of BCCs and SCCs by GPwSIs. This is by far the largest series of cases we could find in the literature, and the overall rate of incomplete margins within our study (6%) is similar to, or lower than, that reported in other studies. For example, a recent study showed an overall rate of 14% incomplete margins for BCC and SCC excisions combined. Within that study, the rate was 26% among GPs, 10% among consultants (eg, general surgeons, plastic surgeons, and dermatologists), and 10% among registrars.8 A similar study also reported an overall rate of 14% incomplete margins for both BCCs and SCCs combined, and the rate among GPs was 16%; consultants, 12%; and registrars, 8%.9 Studies focusing on incomplete BCC excisions among plastic surgeons have reported rates ranging from 2% to 11.2%,10,11,14- 17 whereas a rate of 10.3% has been reported among dermatologists.18 Studies focusing on incomplete SCC excisions have reported rates of around 6% among specialist consultants.4,19 It should be noted, however, that high rates of incomplete excisions by consultant surgeons may largely be due to the selection of more difficult cases.10
As expected, we also show that a much higher rate of incomplete excision occurs in tumors in high-risk sites such as those around the face, and this is in agreement with the literature.4,8,14,18- 21 The current study showed that for BCCs the highest rates of incomplete excisions were on the ears (19%) and nose (14%) areas. Similarly, the highest rates of incomplete excisions for SCCs were also on the ears (18%), with the rates for eyes/brow, cheek/mouth/chin, nose, and forehead ranging between 10% and 13%. This is because, largely for cosmetic and functional reasons, there is a tendency to use narrower margins in the head and neck regions, specifically the cheek, nose, forehead, and ear.21
In the current study, female patients also had a small increased risk (28%) of an incomplete excision, but only for BCCs, whereas other studies have failed to find that patient sex is a predictor of incomplete excisions for BCCs18 and SCCs.4,19 After controlling for age, duration of tumor, anatomic site, histologic subtype, horizontal diameter of the tumor, and ulceration, Takenouchi and colleagues20 reported that BCCs in males have a propensity for deeper invasion than those in females; however, that does not explain why in our study females with BCCs had a higher risk of incomplete margins.
After controlling for all other factors, our results also showed that the risk of incomplete BCC excisions was significantly higher among patients older than 70 years (when compared with patients ≤50 years), but this was not the case for SCC excisions. In the few studies that have investigated various predictors of incomplete margins for BCCs, mixed results have been reported for age,11,18 whereas other studies, similar to ours, have found that age is not a significant predictor of incomplete margins for SCC.4,19
When the analyses were stratified by anatomic site, the increased risk of incomplete excision among elderly patients occurred only in the head and neck area. Although BCC lesions may be more prevalent among the elderly and this may vary by subtype,22 it does not explain the higher rate of incomplete excisions within this age group, particularly on the head and neck areas. However, this finding could be attributed to a longer time for cumulative sun damage and tumor growth among the elderly and also may be due to lack of medical attention.22 Hence, there is a likelihood that larger BCC lesions are found among the elderly and, for cosmetic reasons, lesions on the head and neck area are more difficult to excise completely. For example, Bhatti and colleagues8 found that the larger the lesion, the higher the rate of having an incomplete margin, many of which were found in the inner canthus and nasal bridge area. Furthermore, BCCs and SCCs can invade deeply on the facial areas and can contribute to re-excision and recurrence of lesions in these areas.14,20
We found an increased risk (73%) of an incomplete excision among BCCs without a previous biopsy compared with excisions that had a previous biopsy of some sort (eg, shave or punch biopsy or curettage). This is interesting because Chiller et al23 found that preoperative curettage decreases the frequency of incomplete margins by 26% for BCC excisions but not for SCC excisions. This is similar to our finding, as we found an association only with BCC excisions and not with SCCs. It is therefore possible that, without a biopsy and a histopathologic diagnosis, planning for full excision is insufficient. Specifically, it may be that the necessary margins for a particular histopathological subtype cannot be determined without a biopsy. National guidelines24 recommend biopsy in most cases of suspected BCC and also recommend wider margins for more aggressive types of BCC.
In addition, on the basis of stratified analyses, there was still an increased risk of an incomplete margin for BCC excisions without a previous biopsy (compared with excisions that had a previous biopsy) on the head and neck, trunk, and arms, with a much higher risk (3-fold) for the last of these. The much higher risk on the arms is interesting; we are unable to find any reference to such a finding in the literature, and no such association was found for SCC in our series.
The second key message from this study is that the services provided by GPwSIs may be highly variable in their outcomes. Most physicians in our series had rates of incomplete excision similar to the series average, although a small number had much higher rates. There also was substantial variation in incomplete excision rates between and within clinics. As shown in Figures 2 and 3, in several of the clinics this variation was due mostly to 1 or 2 physicians within each clinic. While studies have compared the incomplete excision rates for NMSC among different specialty professions,8,9 to our knowledge there is no literature comparing incomplete excision rates among GPwSIs within and across a large number of skin cancer clinics. Although we observed substantial variation in the rates of incomplete BCC and SCC lesions between and within clinics, further research is required to corroborate our findings.
The strengths of our data include the very large number of tumors, the substantial number of physicians and clinics included, and the wide geographic scope. This compares with other reports published on this topic, which typically include no more than a few hundred cases, often from a small number of clinicians.
Our data do have limitations. First, we cannot be sure that all tumors excised by all physicians in all of these clinics were examined by the same pathology provider, and hence our data set may be incomplete. Although physicians have no direct financial incentive to use the same provider, we know (from personal communications with them) that almost all do. Also, because this was not a longitudinal study, data on subsequent excisions after the time of data collection were not available. It also does not capture cases of incomplete excision in which pathological margins may have been clear but subsequent clinical tumor recurrence was not.
Second, because of the structure of the pathology reports, we were unable to reliably extract data on histopathological subtypes of BCCs and SCCs, so we were unable to explore the association between aggressive subtypes and incomplete margins. It is therefore possible that the high rates of incomplete margins among some physicians is due, at least in part, to their treating larger numbers of patients with aggressive cancers and tumors in high-risk sites (head and neck). However, we were able to control for anatomic site and have no reason to believe that case selection occurred to any significant extent. These are primary care clinics treating an unselected and nonreferred patient population, although it is possible that some intraclinic referral of difficult cases occurs. Furthermore, we were unable to extract information on the lesion size, and various analyses among different subgroups of patients and anatomic sites would have added important information to our findings.
Finally, in our data there was no personal information about the physicians, such as age, sex, and medical training, and it would have been interesting to include these factors in our analyses.
In Australia there are several corporate chains of skin cancer clinics. Our study occurred in only 1 of these chains, and generalization to the others should be done with caution. The clinics we studied had a companywide training program and an annual conference for physicians. However, there were no formal treatment guidelines, no audit, and no associated quality assurance processes, with no formal referral or specialist support networks. This contrasts with the arrangements in the United Kingdom, where the National Institutes of Clinical Excellence have established national guidelines and treatment pathway standards, together with formal networks and training for GPwSIs. However, despite this, recent evidence demonstrates that outcomes in the United Kingdom are less than ideal.25
In Australia, where the burden of skin cancer is the highest in the world,3 we believe that there is a pressing need for much more formal organization and oversight of the management of skin cancer in primary care. Although national guidelines do exist,24 they may not be widely used. Despite the overall incomplete rate in our study being within a reasonable standard, the large variation among physicians leads us to conclude that there remains a significant need for accredited training among the primary care workforce, and, perhaps even more importantly, there needs to be national audit and performance reporting.
Correspondence: Craig Hansen, PhD, School of Medicine, University of Queensland, Herston Road, Brisbane, Queensland 4005, Australia (firstname.lastname@example.org).
Accepted for Publication: April 10, 2009.
Author Contributions: Drs Hansen and Wilkinson and Ms Hansen had full access to all 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: C. Hansen and Wilkinson. Acquisition of data: C. Hansen and Wilkinson. Analysis and interpretation of data: C. Hansen, Wilkinson, M. Hansen, and Soyer. Drafting of the manuscript: C. Hansen, Wilkinson, and M. Hansen. Critical revision of the manuscript for important intellectual content: C. Hansen, Wilkinson, and Soyer. Statistical analysis: C. Hansen. Administrative, technical, and material support: M. Hansen. Study supervision: Wilkinson.
Financial Disclosure: None reported.