Aid to diagnosis of melanoma in primary medical care

Introduction

Although several new techniques for helping in the diagnosis of cutaneous melanoma are being developed—for example, epiluminescence microscopy1 and computer image analysis2—visual inspection is still the essential part of the diagnosis. Diagnosis is difficult because early melanomas share many of the clinical characteristics of other pigmented skin lesions, especially melanocytic naevi, which are much more common.3 4 5 Evidence from New Zealand suggests that 4% of the white population annually consult a primary care physician about a melanocytic lesion for assessment of malignancy; around 20% of these consultations result in excision of the skin lesion.6 Excision rates of melanocytic lesions are high in Australia7 8 and result in substantial annual health costs—only 1% of melanocytic lesions excised from the under 40 age group are malignant.8 While there are several possible reasons for removing benign skin lesions, including problems with discomfort and unsightliness, it is thought that most are removed to exclude malignancy as in New Zealand. These patterns of consultation and excision are likely to reflect those in similar white populations in the United States and Europe.

When a patient presents with a pigmented skin lesion the outcome may be straightforward because the lesion is manifestly benign, and simple reassurance of the patient is all that is required or the lesion may be suspicious enough to justify its surgical removal to exclude malignancy. The similar clinical appearance of benign naevi and early melanomas, however, means that it is often impossible to exclude malignancy at one visit.9 In these instances, recommended management is to record the appearance of the lesion and review it after an interval of time to determine if any of the changes associated with malignancy have occurred.10 Behaviour of a pigmented lesion over time is believed to be more important than its immediate appearance at time of presentation.11

Given the critical importance of the documentation of suspicious change in a pigmented lesion on the one hand and the extremely high excision rates of benign lesions we have observed locally8 on the other, we explored the possibility of decreasing the number of false positive diagnoses without increasing the number of false negatives. To that end we designed a randomised controlled field trial to investigate whether providing a structured diagnostic approach would help in the differential diagnosis of melanoma. An essential component of this was an objective means of recording change in clinical appearance of a suspicious lesion. We hypothesised that providing primary care doctors with a simple diagnostic algorithm and a camera would lead to a reduction in the proportion of benign melanocytic lesions (mostly naevi) excised without reducing the rate of invasive melanomas and potentially invasive melanocytic lesions excised.

Methods

Two provincial cities in central Queensland were selected for the trial on the basis of their similarity: both are relatively isolated tropical cities near the coast with populations of around 55000 and 65000 people working in industries with substantial agricultural and tourist components. The cities are sufficiently far apart that an intervention in one city was unlikely to affect clinical behaviour in the other. The city to receive the active intervention was chosen at random. Each city is served by about 50 general practitioners with additional junior hospital doctors who also are involved to a limited extent in primary care. At the outset we visited each doctor who practised within a 15 km radius of the central post office in both cities and whose practice included the management of skin cancer as a primary responsibility to explain their possible role in the study and invite them to participate. Thus, as well as general practitioners and junior hospital doctors we visited surgeons and dermatologists, asking them to give permission for a copy of the histology report of every melanocytic skin lesion they excised over the next two years to be sent to us by the pathology services concerned. In addition, reports from the previous six months were collected as a baseline and to check that excision rates of benign and malignant melanocytic lesions were comparable between the two cities.

For each melanocytic skin lesion that was excised the relevant doctor received a questionnaire asking about reasons for excision; the level of pressure (if any) exerted by the patient to have the lesion removed; and who (if anyone) referred the patient. If the patient had been referred another questionanire was sent to the referring doctor. Doctors in the city randomised to receive the intervention were offered a protocol to assist in the management of any melanocytic lesion for which a diagnosis of malignancy was entertained. The protocol consisted of an algorithm (figure 1) which was a modified and simplified summary of “best clinical practice,”10 and incorporated the discretionary use of a camera. This was a Polaroid instant developing camera and came attached to a plastic enclosure which fixed the lens distance from the skin, thus requiring no adjustment by the user for either focus or the flash exposure. One camera was left in each participating practice and hospital after a demonstration of its use. One minute after a lesion was photographed a lifesize colour print was available to be filed with the patient's notes as an objective record of the lesion's clinical characteristics. It was made clear that the responsibility for managing pigmented skin lesions lay with each doctor and that the protocol and camera were offered only as an aid to diagnosis and management.

We pilot tested the protocol including the method of collecting histology reports with eight general practitioners and three different pathology services in a separate city and showed it to be acceptable and feasible. At the end of the study a self administered questionnaire to evaluate the use of the camera was posted to all the doctors in the intervention city who were still living there.

We assumed that every melanocytic lesion excised by local practitioners in the study was submitted for histological examination, as in Queensland with its high annual incidence of melanoma12 this is standard medical practice. On the basis of the histology reports we classified excised melanocytic skin lesions as invasive melanoma (level II and above), potentially malignant (either premalignant—that is, in situ melanoma and lentigo maligna—or atypical or dysplastic naevi, whose presence may have clinical importance for future surveillance of a patient for melanoma), or benign—that is, halo, compound, dermal, intradermal, Spitz, balloon cell, blue, junctional, combined, or unclassifiable types of naevi, and lentigenes.

We obtained consent from all 53 medical practitioners approached (45 general practitioners, seven surgeons, and one dermatologist) in the control community and 52 out of 55 medical practitioners (48 out of 51 general practitioners and all four surgeons) in the intervention community. Over the two year intervention period nine new treating doctors entered and two left the control community, and seven new doctors entered and five left the intervention community. All new incoming doctors agreed to take part except for one general practitioner in the intervention city. The populations in each city had undergone similar growth (under 2% in each city between the census years of 1986 and 1991). Pathology laboratories supplied aggregate data on the melanocytic lesions excised by the medical practitioners who did not participate. There were no changes in reporting procedures of the pathology laboratories during the period of the trial.

All data analysis was performed before the code identifying the city was broken. Comparisons were made between the cities at baseline and between and within cities after intervention with respect to age, sex, classification of lesion, and anatomical site grouped as head and neck, trunk, arms, legs, and buttocks. Differences in proportions were analysed univariately by the χ² test and in median age and thickness of melanoma by the Wilcoxon two sample test. Estimates of the magnitude of the intervention effect were finally adjusted for the potential confounding effects of sex, age, and site of the lesion by using multivariate logistic regression analysis. Differences between cities with respect to changes from baseline to after intervention were examined by fitting an interaction term to these multivariate models.

Results

Discussion

We found that offering doctors a diagnostic algorithm and providing them with a camera seemed to reduce the relative proportion of benign naevi they removed especially from women and young patients, in whom relative rates of excisions of benign pigmented lesions are known to be high.8 The effectiveness of the intervention protocol was shown by the differences in changes in the excision patterns between the intervention city and the control city. Neither the proportion nor the absolute number of melanomas decreased; rather they doubled in the intervention city relative to the control city even though there was a decrease in the background incidence of excisions seen in both cities. That is, the rate of false postive diagnoses seems to have been reduced, and there was no indication that the rate of false negative diagnoses had risen. Our conclusions regarding the importance of the changes after the intervention must be tentative, however, given the sample size of one city in each arm. Ideally several cities would have been randomised to each arm of the trial, but logistic and financial considerations constrained the size of the present study.

The simple algorithm provided a structured approach to the management of pigmented lesions. With the algorithm clinicians could systematically sort those which were obviously benign and did not require excision; those whose excision was clearly indicated on the basis of clinical appearance or the patient's history; and those lesions for which the diagnosis of melanoma could not be definitely excluded. For the latter lesions a doctor can elect to follow up the patient rather than excise the lesion immediately because detectable change in its appearance over a relatively short period increases the probability that it is a malignant melanoma.11 We provided a camera as an aid to increasing the certainty that a change in appearance of the lesion in doubt had occurred, hypothesising that diagnostic specificity would be increased if the clinician had an accurate and objective record of the appearance of the pigmented lesion on initial presentation. A Polaroid photograph attached to a patient's records thus provided a valid baseline against which to compare the clinical appearance of the pigmented lesion on follow up.

Other aids to improving the differential diagnosis of melanoma do not entail monitoring the change in appearance of suspicious pigmented lesions. Techniques such as computer image analysis2 and epiluminescence microscopy13 rely on the clinical appearance of a suspicious pigmented lesion at the time of initial presentation. The aim of computer image analysis is to refine the physician's assessment of critical diagnostic features of early melanoma—namely, area, colour variegation, regularity of outline—whereas epiluminescence microscopy attempts to grade the patterns of epidermal and dermal pigmentary structures not visible to the naked eye. Appropriate computer hardware and software are required for the former technique and considerable skill to interpret the pigmentary patterns observed is required for the latter. Such diagnostic adjuncts cannot be absorbed simply into clinical practice, unlike a Polaroid camera which requires no special training to use proficiently. Moreover, the camera is relatively cheap and within reach of most clinicians.

It is difficult to speculate what cost-benefits might accrue from the widespread use of these adjuncts to diagnosis, and they would be greater the higher the prevalence of pigmented lesions in the community. In the present study the total number of lesions saved from excision was about four (out of eight) for each general practitioner a year or about 200 lesions annually in a community served by 50 general practitioners. The cost of operations entails not only the price of the surgical excision (about $A50-100) but also the cost of the histology (about double the cost of excisions), the cost of visits for care of wounds, and, in a proportion of patients, the additional costs of any complications such as infection that may occur after operations. Other costs to patients, such as the discomfort and unsightliness of surgical scars, are also a consideration. Against these must be weighed the costs of more follow up visits (of those not incorporated into other forms of episodic care) at about $A25 each visit, the camera itself (about the cost of a good stethoscope), and the film (about $A2 per photograph).

Information on acceptability and perceived usefulness of the intervention protocol obtained from the doctors in the intervention city indicated that the camera was a feasible addition to the normal apparatus of a medical practice. Most doctors were satisfied that the photographs helped not only by improving accuracy of follow up but also in reassuring patients.

Despite the large variation in clinical circumstances surrounding the diagnosis of pigmented skin lesions, careful photographic follow up of suspicious lesions about whose malignant state a clinician harbours genuine doubt seems to be an immediately accessible and effective adjunct to clinical diagnosis. Furthermore, if further research confirmed that this strategy helps to decrease the high incidence of excision of benign pigmented lesions8 its cost effectiveness would be substantial, especially among white populations whose background incidence of melanoma is high.

We thank the pathologists and all the doctors of both cities who participated in the study. Diana Battistutta for statistical help, Lyn Francey for research assistance, and the University of Queensland's Mayne Bequest Fund and the Queensland Cancer Fund for financial assistance.