Baade PD, Youl PH, Janda M, Whiteman DC, Del Mar CB, Aitken JF. Factors Associated With the Number of Lesions Excised for Each Skin CancerA Study of Primary Care Physicians in Queensland, Australia. Arch Dermatol. 2008;144(11):1468-1476. doi:10.1001/archderm.144.11.1468
Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008
To assess physician, patient, and skin lesion characteristics that affect the number of benign skin lesions excised by primary care physicians for each skin cancer.
Prospective study collecting clinical, patient, and histopathologic details of excisions or biopsies of skin lesions by random samples of primary care physicians.
Southeast Queensland involving traditional family medicine physicians (n = 104; response rate, 53.9%) and family medicine physicians working in 27 primary care skin cancer clinics (n = 50; response rate, 75.0%).
Of 28 755 skin examinations recorded during the study, 11 403 skin lesions were excised or biopsied; 97.5% of the excised lesions had clinical and histologic diagnoses recorded.
Main Outcome Measures
Number of lesions needed to excise or biopsy (NNE) for 1 melanoma (pigmented lesions only) and NNE for 1 nonmelanoma skin cancer (nonpigmented lesions only).
The NNE for nonpigmented lesions (n = 8139) was 1.5 (95% confidence interval, 1.4-1.6) and for pigmented lesions (n = 2977) was 19.6 (16.2-22.9). The NNE estimates were up to 8 times lower if the physician thought the lesion was likely to be malignant and up to 2.5 times higher if there was strong patient pressure to excise. The NNE estimates varied by other physician-, patient-, and lesion-related variables.
Clinical impressions of excised skin lesions were strongly associated with NNE estimates. By focusing on pigmented skin lesions and by addressing the physician- and patient-specific factors identified, the effectiveness of future training for primary care physicians in the clinical management of skin cancer could be improved.
Skin cancer is the most common cancer in countries with predominantly fair-skinned populations,1 and it is a significant and increasing public health issue.2- 5 Australia has the highest incidence rates of nonmelanoma skin cancer (NMSC), an order of magnitude greater than in other countries.3,6 In particular, Queensland, the northeastern state of Australia, has the highest incidence of melanoma in the world.6,7
Early diagnosis of skin cancer requires an ability to differentiate between benign and malignant lesions. In Australia, most skin cancer is diagnosed and treated by primary care physicians, so their ability to diagnose skin lesions is crucial.8
Physicians presented with a suspicious skin lesion will use clinical reasoning9,10 to decide the appropriate course of action. The decision to excise or biopsy a skin lesion is based on the perceived implications of treating vs not treating the lesion. Factors that affect this decision include the clinical diagnosis of the skin lesion; the degree of uncertainty about its malignant potential; the desire to not miss a malignant lesion11,12; patient-related issues, including insisting that the lesion be removed13; physician-related issues, such as experience12; and the economic consequences, including benefits to the physician of performing excisions.14
Mainly because of its high incidence, skin cancer was the single biggest contributor to cancer-related health costs in Australia in 2001, costing $262 million (in US currency), or approximately $14.60 per capita.15 It ranked fifth in the United States,16 the annual expenditure for NMSC being $650 million in 199517 and for melanoma being $560 million,18 which, when combined, equates to approximately $4.60 per capita. Skin cancer is also having increasing economic impact in Europe2; recent estimates for Germany suggest that hospitalizations due to skin cancer in 2003 cost $281 million, or approximately $3.40 per capita.19 Much of the economic burden is caused by excisions and biopsies of benign lesions. Between 2001 and 2005, the rate of diagnostic biopsies increased in Queensland by 130%, yet the rate of NMSC increased by only 28%.20 Published studies have reported that up to 31 pigmented lesions are excised by primary care physicians for 1 confirmed melanoma8,14,21 and up to 2 nonpigmented skin lesions for 1 confirmed NMSC.22
Therefore, there seems to be scope for reducing the economic impact associated with the management of skin cancer by reducing the excision of benign lesions. It is unclear what clinical-, lesion-, and patient-related factors are associated with higher rates of benign excisions. It was the aim of this large, prospective study of primary care physicians in the high incidence population of southeast Queensland to assess the impact that specific treating physician, patient, and lesion characteristics had on the number of lesions (treating pigmented and nonpigmented lesions separately) needed to excise or biopsy (NNE) to detect a malignant skin lesion.
This study, conducted in 2005, was designed to examine details of skin lesions excised or biopsied by family medicine physicians (FMPs) and those FMPs working solely in the field of skin cancer (hereafter “skin cancer physicians”) in primary care skin cancer clinics in southeast Queensland. These clinics are open access and are staffed by FMPs who have a specific interest in skin cancer medicine. The fieldwork methods used for this study are described in detail elsewhere,23 with a summary provided herein.
A University of Queensland ethics committee approved this study. Participating physicians prospectively recorded details of all consultations involving patients undergoing a skin examination. Because, by definition, a clinical and a histologic diagnosis are required on all lesions to estimate the NNE, this article focuses on consultations in which a lesion was excised or biopsied.
A sample of FMPs (n = 200) working in mainstream general practice were randomly selected and invited to participate. Nonrespondents were sent a reminder letter, and, if required, they were contacted by telephone at 2-week intervals. A total of 104 (53.9% of 193 eligible) FMPs participated in the study. Skin cancer physicians from 36 eligible primary care skin cancer clinics in the study area were similarly contacted. Twenty-seven skin cancer clinics (75.0%) participated in the study, representing 50 skin cancer physicians.
Each participating physician was asked to provide data on age, sex, year of graduation, location of training, years of experience working as a skin cancer physician or an FMP, type of fellowship, any additional training since graduation (including training in skin cancer diagnosis), and whether they use equipment to aid diagnosis in their day-to-day practice (response categories are given in Table 1).
Data collection for FMPs was split across two 8-week periods. Because the volume of skin examinations was known to be higher in skin clinics than in general practice,21,24 we split data collection for skin cancer physicians across two 4-week periods.
Physicians completed case report forms for all consultations involving a skin examination during the study period. Items on the case report forms (response categories are given in Table 1) for these consultations included type of skin examination, who initiated the examination, and basic demographics (age and sex) of the patient. When physicians excised or biopsied a lesion, they recorded the site of the lesion, whether the treatment was a planned surgical procedure or a reexcision, their provisional diagnosis, the likelihood of malignancy, and the degree of patient pressure to excise. Trained research staff extracted histologic diagnosis, body site, and whether the lesion was “excised” or “biopsied” from pathology reports at physicians' practices for each excised and biopsied skin lesion. Excisions were defined as all surgical procedures where the entire lesion was removed, regardless of the instrument used (eg, scalpel and curette); biopsies were defined as all surgical procedures in which part of the lesion was removed for the primary purpose of histologic diagnosis. These were predominantly “punch biopsies” or “shave biopsies.” Lesions that were biopsied and later excised were counted only once, based on the first surgical procedure (ie, biopsy).
The diagnosis of lesions was grouped into 9 categories: melanoma (including lentigo maligna), squamous cell carcinoma (including keratoacanthoma and intraepidermal carcinoma), basal cell carcinoma, solar keratosis, dysplastic nevus, benign nevus, other pigmented benign lesions (including seborrheic keratosis, lichenoid keratosis, and solar lentigo), other benign lesions (including cyst, dermatofibroma, and hemangioma), and other malignant lesions (including mycosis fungoides and sebaceous carcinoma).
The analysis focused on factors associated with the number of skin lesions excised or biopsied for 1 malignant lesion. The NNE14,25 is an epidemiologic measure of diagnostic accuracy calculated by dividing the total number of lesions excised or biopsied by the number of those lesions confirmed by histologic examination as being malignant. Estimates of NNE were calculated separately for pigmented lesions (melanoma, dysplastic nevi, benign nevi, and other pigmented benign lesions, with malignant lesions being melanoma) and nonpigmented lesions (NMSC, solar keratosis, and other benign nonpigmented lesions, with malignant lesions being NMSC, including squamous cell carcinoma and basal cell carcinoma). Analyses were conducted using Stata 9.2 (StataCorp LP, College Station, Texas).
All the analyses accounted for the multistage sampling design, including the sampling by physicians and the selection of multiple skin clinic physicians in the same clinic. We transformed the parameter estimates from a standard logistic regression model (using the formula eβo / [1 + eβo]) to calculate the subgroup-specific NNE, along with 95% confidence intervals using the nlcom command in Stata.
The NNE ratio represents the ratio of the NNE for 1 population subgroup divided by the NNE for the reference population subgroup. The NNE ratios were derived from a standard logistic regression model using the transformation (eβo + β1x1 / [1 + eβo + β1x1]) / (eβo / [1 + eβo]). These ratios, and their 95% confidence intervals, were obtained using the nlcom command in Stata.
The NNE ratios were adjusted for a range of clinical and patient-related variables (Table 1 and Table 2). We initially included all covariates in the logistic regression model and then manually refined the model by excluding nonsignificant variables based on a conservative significance level of P < .25. We assessed the significance of covariates in the multivariate model by testing the hypothesis that all the category-specific variables of the covariate were equal to zero. The choice of baseline reference categories for the NNE ratios was made on the basis of cell sizes (reference category had the largest cell size) or, in some cases, to facilitate easier interpretation.
For 3 variables—likelihood of malignancy, level of patient pressure, and who initiated the consultation—3% to 6% of the records had missing responses. To prevent these records from being deleted from the logistic models, we included a separate category (“no response”) for these variables. Significance of estimates did not include this “no response” category.
Of 28 755 skin examinations recorded during the study, 11 403 skin lesions were excised or biopsied. The mean age of patients who had an excision or a biopsy was 58 years compared with 51 years for patients who had a skin examination only. The proportion of male and female patients who had a skin excision or biopsy was 57.2% and 42.8%, respectively. Of these excised lesions, 11 116 (97.5%) had clinical and histologic diagnoses available for analysis. Approximately a third (33.6%) of the participating physicians were women, and the mean number of excisions per physician was 73 (median, 34; range, 1-741).
Almost half of the excised lesions (48.8%) were histologically confirmed as NMSC (squamous cell carcinoma, 17.8%; basal cell carcinoma, 31.0%), 12.9% were solar keratoses, 9.4% were benign nevi, 5.4% were dysplastic nevi, 10.6% were other pigmented benign lesions, 11.4% were other benign lesions, and 1.4% were melanomas. Only 9 lesions (<0.1%) were grouped under other malignant lesions. More than a quarter of the excised lesions (26.8%) were pigmented. There were 754 seborrheic keratoses (6.7%) included in the other pigmented benign lesions category.
There were 8139 nonpigmented lesions excised or biopsied during the study, of which 5431 were NMSC. The overall NNE for nonpigmented lesions was 1.5 (95% confidence interval, 1.4-1.6). That is, 1.5 nonpigmented skin lesions were excised or biopsied to identify 1 NMSC. Physicians who graduated overseas and those with no vocational registration were independently associated with higher NNE estimates (Table 1). Age of the physician was not consistently associated with higher or lower NNE for nonpigmented lesions. We found no significant independent association between NNE and any of the remaining characteristics of physicians.
Physicians initially diagnosed 34.0% of excised nonpigmented lesions as being very likely to be malignant, and these made up 46.2% of confirmed NMSCs, whereas 15.6% had low likelihood of malignancy, and these composed 6.8% of confirmed NMSCs. Initial diagnoses of very likely to be malignant were significantly associated with lower NNEs than were lesions classified as not at all likely to be malignant (Table 1). High levels of patient pressure were significantly associated with higher estimates of NNE (Table 1). Higher estimates of NNE for nonpigmented lesions were also associated with lesions initially biopsied rather than excised, whereas lower NNEs were observed for lesions excised from the back. Lesions from females and younger patients were associated with higher NNEs (Table 1).
There were 2977 pigmented lesions diagnosed, of which 152 were histologically confirmed to be melanoma. The overall NNE for pigmented lesions was 19.6 (95% confidence interval, 16.2-22.9) (Table 2). When we excluded the 55 in situ malignant lesions, the NNE for pigmented lesions increased to 30.7 (95% confidence interval, 24.1-37.3).
Female physicians had significantly higher NNEs than did male physicians, but there was no difference between overseas- and Australian-trained physicians. No other physician characteristics were retained in the final model (Table 2). The physician's assessment of the likelihood of malignancy was strongly associated with the NNE estimates for pigmented lesions: only 4% of pigmented lesions were initially diagnosed as being very likely to be malignant, and these composed 20% of the melanomas; in contrast, 53% of those initially diagnosed as not at all likely to be malignant (score of 1 or 2) composed 20%.
Increased patient pressure was significantly associated with higher NNEs, although the pattern was not consistent across categories. Higher NNEs for pigmented lesions were also associated with initial biopsy rather than excision, an excision site on the head or chest, those from female patients, and those from the young (Table 2).
This study found that, for more than 8000 nonpigmented skin lesions excised or biopsied by primary care physicians in Queensland, a mean of 1.5 nonpigmented lesions were excised or biopsied for 1 confirmed NMSC. For nearly 3000 pigmented lesions, primary care physicians excised or biopsied nearly 20 pigmented lesions for 1 confirmed melanoma. Estimates of NNE for pigmented and nonpigmented lesions varied substantially according to physician, patient, and lesion characteristics.
The overall NNE estimate of 20 for melanomas in this study of primary care physicians was higher than those reported for referral-based practices in Australia and internationally, with estimates of 16 from a referral-based pigmented lesion clinic in the United Kingdom,12 whereas estimates for dermatologists are lower, ranging from 7 to 13 in Australia8,26 and 6 in Italy.11 The NNE of 20 was, however, lower than that in most other Australian studies of primary care physicians, with NNE estimates of 29 to 31.8,14,21 A separate Queensland study27 of rural FMPs reported baseline NNE estimates for melanomas of 16, and it could be that these lower NNE estimates in Queensland reflect the high incidence of melanoma and the greater exposure physicians have to diagnosing pigmented lesions.
There are few other published estimates for nonpigmented excised lesions. However, the present estimate of 1.5 excisions of nonpigmented lesions for each confirmed NMSC was similar to that reported recently from a retrospective study22 of nonpigmented skin lesions in Tasmania, Australia (NNE of 1.6).
The lower NNE estimates for nonpigmented lesions vs pigmented lesions could reflect the higher prevalence of NMSC in the population compared with melanoma. It could also reflect the general difficulty of clinically diagnosing melanoma because many of the typical clinical features of melanoma also can be seen in benign pigmented lesions.5
The ideal NNE is difficult to quantify. From the point of view of fewer excisions of benign lesions and reducing unnecessary morbidity, complications, adverse effects, and cost, a lower NNE is preferred. However, a lower NNE risks malignant lesions being missed. This is of greater importance for melanoma, with its greater potential to cause mortality and associated medicolegal threats.11,12 Although the prevalence of missed melanomas from referral-based clinics is low,11,28 and medical inaction has been suggested to account for only a small proportion of the time until a melanoma is removed,29 we did not examine missed melanomas in this study. The desire to not miss melanomas has to be balanced by the need to reduce health care costs of unnecessary excisions and biopsies.9 Of course, NNE is not a measure of diagnostic accuracy alone; it also depends on the prevalence of the disease.8,14 Therefore, there can be no ideal standard for NNE30 without also assessing sensitivity.31
The finding that female physicians and those who were trained overseas had higher NNEs confirms reports by other researchers14 for pigmented lesions. Perhaps, given the lower skin cancer incidence in countries other than Australia, overseas-trained physicians are taught less about skin cancer during their initial training and obtain less experience with diagnosing skin cancer before moving to practice in Australia. The reasons for the female physician finding can also be speculative only: increased priority on not missing malignant neoplasms or being presented with fewer malignant lesions. We found that 49% of male physicians rated the likelihood of malignancy as high (score of 4 or 5) compared with 41% of female physicians.
Associations with likelihood of malignancy were the strongest observed in this study and suggest that clinical impressions play an important and effective role in the decision-making process. However, the finding that 20% of histologically confirmed melanomas were initially considered to have low likelihood of malignancy highlights the complexity of diagnosing pigmented skin lesions.
Increased levels of patient pressure (as perceived by the physician) resulted in higher NNEs. This suggests that patients have a generally lower ability to make the right diagnosis or are more anxious than their physician about missing a cancer. This highlights the importance of taking into account pressure from the patient to excise skin lesions when examining clinical diagnostic accuracy.
The higher proportion of benign lesions identified through biopsy compared with excision was to be expected and confirms the role of biopsy as an additional diagnostic tool. In contrast, we found that excisions were used more when the physician was very confident that the lesion was malignant. When we limited the analysis to excised lesions only (results not presented), we found similar associations to those already reported.
Higher NNEs in women and younger patients have been reported previously.8,14,30 This may relate to the lower incidence of skin cancer in these subgroups1,4 or to the increased effect of other unmeasured factors, such as cosmetic prompts.8We found that 25.9% of excisions in younger (<50 years) patients were histologically benign nevi compared with 3.6% in older patients. Although the data on the relative prevalence of benign nevi in young people compared with older people are inconclusive, the present data support the suggestion that nevi are at least equally, if not more, prevalent in young people compared with older people. Thus, the lower incidence of melanoma in young people means that more lesions need to be excised to find a melanoma than in older people.1,4
This large-scale prospective study was conducted in the state with the highest reported incidence of skin cancer in the world. We collected data on all skin lesions excised or biopsied by participating primary care physicians during the study period. The prospective study design enabled a wider range of covariates to be assessed than were available for previous studies using retrospective study designs.8,14,22,30,32,33 In particular, we measured physician perceptions of likelihood of malignancy and level of patient pressure to excise for each lesion, both of which were demonstrated to have strong associations with the main outcome measures in this study.
The high response rate obtained among skin clinic physicians reduces the risk of selection bias in that group. However, the lower response rate for FMPs, commensurate with the involvement required by FMPs who wanted to participate in this study,14,34 may indicate that FMPs with an interest in skin cancer were more likely to participate, which may have artificially reduced the NNE estimates. In the absence of information on the clinical interests of nonparticipating physicians, we cannot exclude this possibility. We found no evidence of selection bias on the basis of age, sex, or simple measures of clinical training. It is also possible that this prospective study found an artificially reduced NNE due to a Hawthorne effect (changes in behavior due to being a study participant).35 Histopathologic data were provided by several pathology laboratories, thus potentially increasing the generalizability of the estimates compared with other similar studies.8,32 We combined in situ and invasive melanomas in the estimates of NNE for pigmented lesions. Although we acknowledge that they might be different histopathologic entities, we suggest that the role of the primary care physician is to identify melanomas as early as possible, not to determine whether they are in situ or invasive. We did, however, find that excluding in situ melanomas from the set of malignant lesions increased the NNE estimate for pigmented lesions. Finally, because this study considered excised lesions only, we could not assess the sensitivity of diagnosis or the likelihood that participating physicians missed malignant lesions.
This study found that clinical impressions of excised skin lesions had a strong relationship with the number of benign lesions excised for every malignancy. However, it also demonstrated that many other factors contribute to the decision whether to excise or biopsy a lesion. It is likely that we could improve the effectiveness of future training for primary care physicians in the clinical management of skin cancer by focusing on pigmented skin lesions and addressing these nonclinical issues. Health care practitioners in many countries face pressures to reduce costs, and the increasing incidence of skin cancer in most fair-skinned populations means that there will continue to be a delicate balance between avoiding falsely negative treatment decisions and reducing costs to the health care system.
Correspondence: Peter D. Baade, PhD, Viertel Centre for Research in Cancer Control, Cancer Council Queensland, PO Box 201, Spring Hill, Queensland 4004, Australia (firstname.lastname@example.org).
Accepted for Publication: January 31, 2008.
Author Contributions: Dr Baade had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Baade, Youl, Janda, Whiteman, Del Mar, and Aitken. Acquisition of data: Baade, Youl, and Janda. Analysis and interpretation of data: Baade, Youl, Janda, Whiteman, and Aitken. Drafting of the manuscript: Baade and Janda. Critical revision of the manuscript for important intellectual content: Baade, Youl, Janda, Whiteman, Del Mar, and Aitken. Statistical analysis: Baade. Obtained funding: Baade, Youl, Janda, Whiteman, Del Mar, and Aitken. Administrative, technical, and material support: Youl. Study supervision: Aitken.
Financial Disclosure: None reported.
Funding/Support: This study was funded by the Australian National Health and Medical Research Council (NHMRC). In addition, Dr Janda is supported by a Training Fellowship from the NHMRC and Dr Whiteman by a Senior Research Fellowship from the NHMRC.
Role of the Sponsor: The authors' work was independent of the funding sources. The funding sources had no involvement with the collection, analysis, and interpretation of the data; the writing of the article; or the decision to submit the article for publication.