Poor Adherence to National Clinical Management Guidelines: A Population-Based, Cross-Sectional Study of the Surgical Management of Melanoma in New South Wales, Australia

The Melanoma Patterns of Care Study was a population-based, cross-sectional, observational study based on clinicians’ reported management of patients who had a histopathologically confirmed primary in situ or invasive cutaneous or unknown primary site melanoma [International Classification of Diseases for Oncology, Third Edition (ICD-O-3) site codes C44.0–C44.9 and C80.9; morphology codes 8720–8790/2 or/3] notified to the NSW cancer registry between 23 October 2006 and 22 October 2007. Ethics approval was granted by both The University of Sydney Human Research Ethics Committee and the Cancer Institute NSW Human Research Ethics Committee.

Since pathologists in NSW are required by law to notify new cancer diagnoses to the Central Cancer Registry, and laboratory information systems perform this automated task electronically whenever a cancer diagnosis is issued in a pathology report, the tumor capture rate was assumed to be 100%. The physician identified as the referring physician on the diagnostic pathology report on which the cancer registration was based was considered as the ‘physician providing initial care following diagnosis’ for this study. For each eligible patient, the physician providing initial care was contacted by the study team and asked to complete a questionnaire regarding the clinical management of that patient. If this physician referred the patient to other physicians, they were then also contacted by the study team and asked to complete a questionnaire. This process was followed for all notifications of eligible invasive melanomas, but for only the first 450 notifications of in situ melanomas due to limited resources. For physicians with large numbers of eligible patients, trained field workers with nursing experience completed the questionnaires from patient medical records, with recourse to the treating clinician as necessary. During the 12 month study period, questionnaires were completed for 2614 primary localized melanomas in 2590 patients, treated by 1084 physicians (78% questionnaire completion rate; 25% nurses, 53% physicians) from a range of primary and specialist care centers. There was a median of 385 days between diagnosis and questionnaire return.

Margins were calculated as the sum of both biopsy and wide local excision (WLE; narrowest) margins. The margins were intended to be the clinical margins as declared by the doctor, rather than those of the histopathological report(s), but only the words ‘excision margin’ were specified on the questionnaire. Therefore, when comparing the recorded excision margins with those recommended in the guidelines, a generous allowance was made for up to 30% shrinkage on pathological assessment following formalin fixation and processing compared with the clinical margin, as up to 20% shrinkage is common.10 For 34% of melanomas where only one excision (either biopsy or wide local) was reported, the other was then given a value of zero in the calculation.

In place in Australia at the time of the study in 2006/2007 were the 1999 evidence-based National Health and Medical Research Council (NHMRC) Australian Melanoma Management Guidelines,8 subsequently updated in 2008.4 The recommended management of a suspected primary melanoma in the 1999 guidelines was initial performance of an excision biopsy with 2 mm clinical margins (to confirm the diagnosis) followed by a WLE at a second operation. The recommended clinical surgical excision margins for the WLE varied according to the primary tumor’s histopathologic features, as summarized in Table 1.

Chi square tests were used to compare proportions between groups. To examine factors associated with appropriate surgical margins and with surgical complications, we constructed log binomial regression models to estimate prevalence ratios11 and 95% confidence intervals (CIs), using the variables of sex of the physician, sex of the patient, Breslow thickness, tumor histopathology, tumor site, tumor ulceration, evidence of tumor spread beyond the primary site, socioeconomic status of the physician’s practice and the patient’s residence (measured using the index for relative social disadvantage, which ranks social and economic disadvantage based on postcode, with the lowest quintile indicating the least advantaged),12 melanoma caseload of the physician over the 12 month period (low, 1–5 patients; medium, 6–30 patients; high, ≥31 patients), initial versus referral physician, physician’s and patient’s age, physician’s practice setting, physician’s specialty, and the rurality (using the Rural, Remote and Metropolitan Areas classification13) and accessibility (using the accessibility/remoteness index of Australia14) of both the physician’s practice and patient‘s residence. A stepwise forward inclusion threshold of p < 0.1 was used for variables in the logistic regression analyses. All analyses were conducted using SAS 9.3 (SAS Institute Inc., Cary, NC, USA), with statistical significance inferred at p < 0.05.

Of the 1745 lesions (67%) with complete margin data, only 35% were excised with margins concordant with the guidelines. In the 65% nonconcordant with guidelines, overtreatment was more common (45%) than undertreatment (21%) (Fig. 1a, b). Multivariate analysis, in which all patient, lesion, and doctor characteristics were included as candidate variables, identified five independent factors significantly associated with nonconcordance: lower Breslow thickness, lower socioeconomic status of the physician’s practice location, lower physician caseload, physician performing the initial biopsy did not do the wider excision, and physician age ≥ 65 years (Table 1). Highest concordance was for intermediate thickness tumors (63%; Breslow > 1.5 to 4.0 mm). The tendency to overtreat rather than undertreat when nonconcordant with the guidelines differed among the Breslow thickness groups (Fig. 1a, b). Overtreatment was most likely for in situ and thin (0–1.5 mm Breslow) invasive melanomas (54 and 51%, respectively), whereas thick melanomas (>4.0 mm Breslow) were the most likely to be undertreated (49%).

a Total Excision Margin by Breslow Thickness. Distribution of total excision margins according to tumour thickness, where recorded (missing in around 30% of each category). b Total Excision Margin Concordance. Differences in the rates of excision margin concordance with the 1999 Australian and New Zealand guidelines according to tumour thickness. Margins were calculated as the sum of both biopsy and wide local excision (narrowest) margins, with allowance for shrinkage on pathological assessment. Notably, only 35% overall compliance was achieved, with the best being 63%, which was for intermediate thickness (>1.5–4.0 mm) tumours. c Wide Local Excision Performed by Excision Biopsy Physician. Differences by specialty in whether or not the doctor who performed the initial diagnostic excision biopsy also performed the wide local excision. d Reconstruction method post resection of melanoma. Differences by specialty in the method of reconstruction utilized following wide local excision. GP general practitioner

Full size image

A WLE was performed in 78% of patients, by the same physician as did the excision biopsy in 50% of these cases. However, this differed according to the specialty of the initial physician (Fig. 1c). Only 27% of primary-care physicians who did an excision biopsy also did the WLE, compared with 55% of skin cancer clinic physicians, 59% of dermatologists, 80% of general surgeons, and 66% of plastic surgeons (p < 0.001).

Since surgical reconstruction training varies between specialties, we examined whether there were also differences in practice across the specialties (Fig. 1d). Primary-care physicians, skin cancer clinic physicians, and dermatologists had similar rates of performing primary closure (88, 86, and 90%, respectively), compared with 54% for surgeons and 42% for plastic surgeons (p < 0.001), with an overall rate of 66%. Complex reconstruction was more frequently performed with flaps (24%) than skin grafts (10%). Flap use was significantly greater among general surgeons (33%) and plastic surgeons (40%) than primary-care physicians (7%), skin cancer clinic physicians (14%), or dermatologists (6%). Similarly, skin grafts were seldom used by primary-care physicians, skin cancer clinic physicians, or dermatologists (5, 0, and 4%, respectively), compared with general surgeons (13%) and plastic surgeons (18%; p < 0.001). Multivariate analysis revealed that the factors associated with performing complex reconstructions were physicians with medium or high caseloads, general or plastic surgeons, older physicians, melanoma on the head and neck or lower limbs, and treatment in a rural setting (Table 2). Neither overtreatment nor undertreatment with respect to excision margins was significantly associated with the use of complex reconstruction (p = 0.12).

In this study 17% of patients with invasive melanoma underwent SLNB, including 45% (365/808) of those with a melanoma >0.75 mm Breslow thickness. SLNB was neither recommended nor discouraged by the 1999 NHMRC Australian Melanoma guidelines. A multivariate analysis was undertaken to determine which factors were associated with SLNB; they included Breslow thickness >0.75 mm (prevalence ratio 6.81, 95% CI 5.23–8.85 for Breslow 0.76–1.50 mm; and 9.88, 95% CI 7.64–12.78 for Breslow >1.50 mm), patient age < 50 years, primary tumor site on the upper limb (prevalence ratio 1.32, 95% CI 1.12–1.55), and treatment in an urban setting (Table 3).

Complication rates for patients in whom a WLE was performed were 8% overall; 3% developed wound infections. There were significant differences in complication rates between specialties, with the overall rate ranging from 15% in patients treated by surgeons to 2% in patients treated by dermatologists (p < 0.001). The complication rate in those who underwent SLNB was substantially greater than in those who did not (23 vs. 6%, p < 0.001). This excess was mainly due to higher rates of seroma or hematoma (10 vs. 1%, p < 0.001), wound breakdown (5 vs. 2%, p < 0.001), lymphedema (4 vs. 0%, p < 0.001), and other (nonspecified) complications (3 vs. 1%, p < 0.001). However, there was no significant difference in the rates of wound infection (5 vs. 3%, p = 0.22) or prolonged pain (1 vs. 0%, p = 0.16). Multivariate analysis identified independent risk factors for developing complications (Table 4) as undergoing SLNB, tumor on the lower limb, lower patient socioeconomic status, higher physician practice socioeconomic status, and treatment by a general or plastic surgeon. Nonconcordance with margin guidelines was not associated with an increased complication rate (p = 0.72).

The 35% overall rate of concordance with the recommended excision margin guidelines was surprisingly low, and of concern because undertreatment, with too narrow margins, increases the risk of local recurrence,15 ^, 16 while excessively wide margins are likely to result in increased morbidity to the patient and greater costs for the health service. The majority of patients whose treatment was not concordant were overtreated (68%), particularly those with in situ and thin invasive tumors, perhaps reflecting historical practice in which much greater excision margins, typically 3–5 cm, were recommended.17 However, physicians aged 65+ years had almost twice the proportion of undertreated tumors compared with those aged under 45 years. Importantly, physicians treating more than 30 patients a year were 27% more likely to comply with the guidelines than those who treated one to five patients a year. In comparison, the previously mentioned German study reported concordance with excision margin guidelines of 84% overall, with the greatest deviation being the excessive treatment of in situ disease (18.4%).9 Their margin guidelines were as follows: in situ disease, 0.5 cm; up to 2 mm Breslow, 1 cm; over 2 mm Breslow, 2 cm. The high overall concordance rate observed in the German study is despite no allowance being made for tumor shrinkage in their criteria of margin adequacy, compared with our 30% shrinkage allowance. There were a number of other differences in the results of the two studies. The German study found that head and neck location was independently associated with inadequate excision margins on multivariate analysis (31.7% for head and neck versus 5.0% for extremities, 5.2% for trunk, and 15.6% for others, p < 0.0001), whereas our study did not. It also found that patients from units with more patients registered in the study (presumably reflecting a higher caseload for treating physicians) and patients with tumors >1 mm Breslow thickness were more likely to be excessively treated, whereas we found the opposite. Furthermore, registration of patients was voluntary in the German study, thus there was potentially significant bias. Such bias may have favored inclusion of patients treated at centers that were more compliant with the national treatment guidelines, in which case the true situation in Germany may have been less favorable.

Our reported complication rate varied considerably between treating physicians’ specialties, but, on multivariate analysis, once SLNB was included as a risk factor, there was only a small effect due to specialty (p = 0.03). Furthermore, the clinical criteria used and how much effort was put into identifying complications is likely to have varied widely from physician to physician, and thus reporting bias is likely to exist. Performing SLNB increased the risk of complications significantly, in line with MSLT-1 trial results.18 However, this may also reflect a reporting bias since SLNB patients would nearly all have been seen by the operating surgeon within a few weeks following surgery, whereas those who did not have SLNB may well have had no early follow-up, or had it conducted by their family physician, with complication data not captured. Routine assessment of complications may also depend on patient volume, clinical trial participation, or physician specialty. Most of these complications are likely to have been short-lived, with the exception being the 4% lymphedema rate in patients who had an SLNB, although the severity of the lymphedema and criteria for its diagnosis were not recorded in this study and thus the morbidity is hard to quantify. In the control group of MSLT-1, i.e. in patients who had wide excision but no SLNB, the lymphedema rate was 0.75%, indicating that not all lymphedema occurring in patients who had SLNB can be attributed to the SLNB.18

Based on this first Australian population-based study of the surgical management of melanoma, there appears to have been poor adherence to the national guidelines in NSW during the 12 month period of the study, i.e. 2006/7. In the majority of nonconcordant cases, treatment was excessive and therefore likely to have increased the morbidity associated with treatment of the melanoma. However, there was also significant undertreatment of some patients, which has even greater implications since it is likely to increase recurrence rates and therefore significantly increase costs, morbidity, and potentially mortality.19 Overall, it appears the quality of surgical care for patients diagnosed with melanoma in NSW in 2006/2007 may have been significantly lower than that of Germany. However, it is possible that there has been significant improvement in guideline compliance within NSW, Australia, during the 10 years since the study period, although the recommendations for surgical margins remain similar, except for a reduction from 2–3 cm to 2 cm for tumors with Breslow thickness >4 mm, and there has been no specific intervention to improve compliance. Consequently, there needs to be a strategic change in the provision of melanoma treatment in Australia, perhaps encouraging or mandating treatment by those who treat more than 30 patients per year. High-level guideline concordance is required worldwide if patients are to be given high-quality care and the burden on health services is to be minimized.