Presented at the 4th World Conference on Melanoma, Sydney, 10–14 June 1997; Journées Dermatologiques de Paris, Paris, 3–6 December 1997; EDEN Congress, 2nd International Meeting on Epidemiology and Prevention of Skin Diseases, Bamberg, Germany, 2–4 May 1998.
Promoting early diagnosis is a widely accepted strategy to improve melanoma prognosis. It is thus important to identify all potential causes for a long delay before diagnosis and a high tumor thickness at diagnosis. All over the world, education campaigns have been conducted to increase awareness about the early signs of melanoma (Marks, 1995, Rhodes, 1995). Few education programs were targeted to doctors (Laidlaw et al.,1996; MacKie, 1995, MacKie, 1997). However, the organization of the health-care system, the quality of medical training, the accuracy of melanoma diagnosis in routine practice, and the quality of the information provided by physicians to their patients are likely to have a major influence on the delays before tumor resection. It is thus crucial to assess the medical intervention in primary melanoma and to identify all possible problems that could be corrected by training and information. However, little is known about the responsibility of doctors in the delays before melanoma resection (Temoshok et al.,1984; Cassileth et al.,1988; Rampen et al.,1989; Krige et al.,1991).
We designed a prospective multicenter study to assess separately patient and doctor responsibility in the delay before diagnosis and the prognosis of cutaneous melanoma. Doctor-related factors are presented herein.
MATERIAL AND METHODS
Patients and methods are described in detail elsewhere (data not shown) and are briefly summarized herein. Specific investigations of doctors' attitudes are described thoroughly.
Consecutive patients, referred after resection of primary cutaneous melanoma in 18 French dermatological departments of the Public Hospital System, entered the study. Patients referred for cutaneous melanoma in these centers entered the study when they fulfilled inclusion criteria, described elsewhere (data not shown).
General description of the survey
All patients were examined and interviewed by a specially trained dermatologist in each center following a precise questionnaire, described in detail elsewhere (data not shown). Briefly, 75 questions investigated patient characteristics and habits, tumor clinical features, circumstances of melanoma detection, causes of delay in diagnosis, and doctor attitudes before tumor removal. Each patient was asked to recall several dates: the date (d1) when he or she first noticed a lesion, the date (d2) when he or she first felt that this lesion could be curious or suspicious, the date (d3) when this lesion was first examined by a doctor who gave an opinion about it, and the date (d4) when a physician first proposed resection. The date d5 corresponded to the date of final melanoma removal. A list of questions also investigated what precisely happened during each time interval, d1–d2, d2–d3, d3–d4, and d4–d5 (data not shown). Data from clinical examinations and results of psychological evaluations of the patients using the Mini-Mult test (Faschingbauer, 1976) were recorded.
When patients reported that the melanoma was diagnosed by a physician during a routine clinical examination or during a visit for an unrelated problem, it was considered to be “coincidentally diagnosed”; the remaining patients were considered to be “self-detected”. From the medical point of view, coincidentally diagnosed melanomas are “doctor-detected melanomas”.
The histological characteristics of the melanomas were reviewed, including classification into superficial spreading melanoma (SSM), nodular melanoma (NM), acrolentiginous melanoma (ALM), and lentigo melanoma (LM) (data not shown).
All data were computerized and analyzed using the EPI-INFO program (version 6.04a; CDC, Atlanta, GA, and World Health Organization, Geneva) and the SPSS program (release 6.1; Chicago, IL). For calculation convenience, all in situ melanomas were standardized to 0.1 mm thickness. Because of skewed data, tumor thickness and delays were mainly expressed as medians and not as means.
The patient role in the delays and prognosis of melanoma is reported elsewhere (data not shown). The study of physician-related factors in melanoma diagnosis is detailed below.
Assessment of the role of physicians in melanoma diagnosis
Analysis of descriptive data.
The patient was interviewed separately about the attitudes and actions of all doctors who examined the lesion before resection was proposed.
Two time intervals, d3–d4 (delay to propose resection) and d4–d5 (delay to perform resection), were considered to depend on physicians. These medical delays were compared to patient delays (d1–d2, d2–d3) using Kruskal-Wallis tests. Medical delays and melanoma characteristics of doctor-detected cases were compared to those of self-detected cases using the Kruskal-Wallis test and the χ2 test, respectively. Similarly, tumor thickness of doctor-detected and self-detected melanomas were compared using the Kruskal-Wallis test. The reasons for a long d4–d5 interval (>1 month) were investigated, using a specific questionnaire which addressed not only patient-related factors (such as anxiety, negligence) but also medical responsibility (the way the patient was informed).
Univariate analysis of medical delays and tumor thickness.
The influence of the following factors on the duration of “medical” delays (d3–d4 and d4–d5) and tumor thickness was studied (Kruskal-Wallis test): clinical features, location, and histological characteristics of the melanoma, as well as the specialty and the attitude of the first physician seen.
Multivariate analysis of medical delays and tumor thickness.
To identify the factors which were most predictive of the duration of the d3–d4 interval, a stepwise multiple linear regression using BMDP (Los Angeles, CA) statistical software was performed. All variables influencing the d3–d4 delay were included in the model: melanoma characteristics, i.e., histological type, location, and color (pigmented or amelanotic melanoma); doctor's specialty; number of doctors who had seen the lesion before resection was proposed; and patient characteristics, i.e., age, residence, education level, medical habits, awareness, history of previously resected mole or previous melanoma, and number of nevi. The same variables were used in a logistic regression analysis using BMDP software to study the most predictive variables of a short d3–d4 interval, which was defined as <2 months. The same process was used to identify the most predictive of a short d4–d5 interval (rapid resection was defined as <7 days) and Breslow thickness.
Accuracy of medical examination.
We assumed that any melanoma which had not been diagnosed by the first doctor who examined the lesion would be removed sooner or later. Physician sensitivity for early diagnosis of melanoma was thus assessed by the percentage of tumors that had been diagnosed at the first medical examination.
Proportion of doctor-detected cases.
If we assume that any melanoma not detected early by a physician would be detected sooner or later by the patient himself, the proportion of doctor-detected cases compared to self-detected cases can be considered a surrogate to the ability of physicians to diagnose early. The higher the proportion of doctor-detected tumors in a given type of melanoma, the higher the ability of doctors to detect this type of melanoma. The ratio of doctor-detected cases to self-detected cases was thus studied in the different sites and types of melanoma.
Melanoma detected by physicians.
In 29.1% of cases (172/590), melanomas were detected by the physician, 51.2% (88/172) of which were found during a routine clinical examination unrelated to a skin disease, 12.2% (21/172) during a systematic skin examination for moles, 10.0% (17/172) during a visit for other skin disease, and 26.8% (46/172) in various other situations. The 70.8% remaining patients (418/590) detected their melanoma themselves.
Assessment of delays under medical control.
The median delay before the doctor proposed tumor resection (d3–d4) was 0, (mean 103, range 0–5,783) days. The median delay to perform removal (d4–d5) was 7 (mean 68, range 0–11,051) days. For comparison, the median delay under patient responsibility (d1–d3) was 912 (mean 3,829, range 0–25,261) days (data not shown).
Of the 172 doctor-detected melanomas, 45.3% (78/172) were detected by a dermatologist, 34.4% (59/172) by a general practitioner (GP), and the remaining 20.3% (35/172) by other physicians.
Of all patients with self-detected melanoma, 58.6% (245/418) first showed their lesion to a dermatologist, 33.1% (138/418) to a GP, and 8.3% (35/418) to another physician (χ2 test, p < 0.001).
In the whole population (n = 590), the “first” physician was thus a dermatologist in 54.7% of cases (323/590), a GP in 33.4% of cases (197/590), an occupational doctor in 5.2% of cases (31/590), a surgeon in 1.9% of cases (11/590), and another doctor in 4.8% of cases (28/590).
Attitude of the first physician.
The first advice from the first doctor was considered to be appropriate in 85.8% (506/590) of cases. He or she removed the tumor in 51.4% (303/590), referred the patient to a dermatologist in 26.4% (156/590), and referred the patient to a surgeon for tumor removal in 4.7% (28/590) of cases. The attitude was considered acceptable in the remaining 3.2% (19/590) of cases in whom the doctor performed a partial tumor biopsy. In the other 14.2% of cases (84/590), the first doctor's attitude toward the melanoma was inappropriate and dangerous: “wait and see” in 5.6% (33/590), “no danger” in 3.9% (23/590), or other inappropriate attitudes in 4.6% (28/590) of cases. In 60.7% of all inappropriate attitudes (51/84), several visits to other doctors (usually 1 or 2, up to 4 different doctors in 1 case) were needed before resection was proposed. In 39.3% (33/84), patients revisited the same doctor (usually 2 or 3 visits but up to 10 in 2 cases) until removal was finally proposed.
First action on the tumor.
The first intervention was a complete resection of the lesion in 90.0% of cases (531/590), partial biopsy in 8.6% (51/590) of cases, “shaving” of the superficial part of the tumor in 0.7% (4/590) of cases, cryotherapy in 0.3% of cases (2/590), and electrodissection in 0.3% of cases (2/590). In 32.7% of cases (193/590), the physician who removed the tumor was the same who first examined it. The first intervention (d5) was performed immediately during the first visit (d3 = d4 = d5) in 16.6% of cases (98/590). Resection was performed by a dermatologist in 58.8% of cases (347/590), a surgeon in 39.4% (232/590), another specialist in 0.3% (2/590), and a GP in 1.5% (9/590).
Medical reasons for delays before final removal.
The d4–d5 time interval was considered long (>1 month) in 12.2% (72/590) of patients. Among them, 19.5% (14/72) alleged that the doctor “did not inform them that removal was urgent”, and the remaining 80.5% (58/72) admitted that they were responsible for the delay: 9.7% (7/72) attributed this delay to “fear”, 38.8% (28/72) to being “too busy”, and 31.9 % (23/72) to “negligence”.
Doctors' accuracy in everyday practice
Among the doctors who first saw the melanoma, 14.2% did not propose resection (see below). Assuming that all melanomas would finally be detected, these data suggest that in routine practice in France, medical examination has a sensitivity of 86% for the diagnosis of melanoma.
Factors influencing medical delays (d3–d4 and d4–d5) and tumor thickness
As expected, the delay to propose resection (d3–d4) was much longer when the attitude of the first physician was inappropriate than when removal was proposed at the first visit (median 109 vs. 0 days, p < 0.001). Although there was a higher tumor thickness when the attitude was inappropriate (median 1.40 vs. 1.15 mm, mean 3.15 vs. 2.00 mm), the difference was not significant (p = 0.99).
Breslow thickness was not different when the melanoma was immediately removed or when it was treated by electrodissection or cryotherapy (mean 2.29 vs. 2. 41 mm, median 1. 16 vs. 2.65 mm, respectively; p = 0.17). However, it is difficult to compare the thickness of tumors previously submitted to incomplete destruction with unaltered tumors.
Tumor thickness was significantly lower when first seen by a dermatologist than by another physician (median 0.94 vs. 1.50 mm, mean 1.88 vs. 2.82 mm, respectively; p < 0.001). The delay to propose removal (d3–d4) was significantly shorter when the first physician was a dermatologist than when he or she was a GP or another specialist (median 0 vs. 25 days, mean 60 vs. 153 days, respectively; p < 0.001). The delay to perform removal (d4–d5), though significantly different (p = 0.011), was similar with dermatologists and other doctors (median 6 vs. 7.5 days, mean 68 vs. 67 days, respectively). This was true whatever the circumstances of diagnosis (i.e., self-detected or doctor-detected melanoma), the histological type, and the clinical presentation of the tumor (pigmented lesion or amelanotic) (Table I). First medical advice was more often appropriate when patients with a suspicious lesion first visited a dermatologist than when they first visited another physician.
Table I. Tumor Thickness, Doctor Delays, and Attitudes According to Specialty and Tumor Characteristics
d3–d4, delay to propose resection; d4–d5, delay to remove lesion.
First advice was studied only in self-detected melanomas since, by definition, it cannot be incorrect in doctor-detected cases.
p < 0.001
p < 0.001
p < 0.001
p < 0.001
p = 0.051
p = 0.028
p < 0.001
p = 0.026
p = 0.011
p < 0.001
p = 0.40
In melanomas detected by physicians, tumor characteristics, i.e., size of the lesion, pigmentation, location, accessibility to view, histological type, regression, ulceration, and histological remnants of a nevus, did not influence significantly doctor-related delays (data not shown).
In tumors considered suspicious by the patients and submitted to the doctors, i.e., self-detected tumors, doctors proposed removal significantly later (d3–d4) for ALM, amelanotic melanomas, and melanomas of the hand and foot than for other tumors. The d3–d4 interval was significantly longer in de novo melanomas than in melanomas developed on nevi (as established by histological remnants of a nevus) (Table II). Unusual tumor type, i.e., ALM, or amelanotic melanomas were more often associated with a complicated and long medical tour. Indeed, 16.5% (68/412) of SSM, 26.2% (28/107) of NM, 19% (4/21) of LM, and 37.5% (9/24) of ALM were examined several times by a doctor before removal was proposed (p = 0.03). Similarly, 42.0% (21/50) of all amelanotic melanomas were first examined several times vs. 17.4% (94/540) of pigmented melanomas (p = 0.025).
Table II. Influence of Tumor Characteristics on Medical Delays (in Patients with Self-Detected Melanoma)
Number of patients
Delay to propose resection (d3–d4) (days) (mean/median)
Delay to remove lesion (d4–d5) (days) (mean/median)
p < 0.001
p = 0.34
p = 0.80
p = 0.81
p = 0.48
p = 0.61
Histological remnants of a nevus
p = 0.012
p = 0.05
p = 0.039
p = 0.60
Size of tumor according to doctor (mm)
p = 0.51
p = 0.60
p = 0.02
p = 0.30
In a stepwise multiple linear regression, the most predictive factors influencing d3–d4 were histoclinical type and the ability of the first physician seen to recognize melanoma. The shortest d3–d4 intervals were observed with LM and melanomas first seen by dermatologists (Table III).
Table III. Multivariate Analysis to Identify Factors Most Predictive of d3–d41 Delay (Stepwise Multiple Linear Regression)
F to enter
d3–d4, delay to propose resection.
Histoclinical type of melanoma
Specialty of first physician
n = 278, r2 = 0.03
In a stepwise logistic regression, the factor most predictive of a long d3–d4 interval (>30 days) remained the specialty of the first physician (other physicians vs. dermatologists; coefficient 2.27, SE 0.32, OR 9.7, 95% CI 5.16–18.2, p < 0.001).
Multivariate analysis to identify the factors most predictive of time to remove the lesion (d4–d5) and Breslow thickness did not retain any medical factors.
Comparison of doctor-detected and self-detected melanomas
Doctor-related delays (d3–d4 and d4–d5) were not statistically different between doctor-detected and self-detected melanomas. Median d3–d4 delay was 0 days (mean 160 days) among doctor-detected melanomas vs. 0 days (mean 79 days) in self-detected melanomas (p = 0.50). Median d4–d5 delay was 7 days (mean 80 days) in doctor-diagnosed melanomas vs. 7days (mean 62 days) in self-detected melanomas (p = 0.47). Tumor thickness was significantly lower in doctor-detected than in self-detected melanomas: median 0.93 (mean 1.54) vs. 1.30 (mean 2.61) mm (p < 0.001).
The distribution of histogenetic types was significantly different in the doctor-detected and the self-detected melanomas (χ2p < 0.001). SSM, NM, and ALM were mainly detected by patients: 72.1% of SSM (297/412), 81.3% of NM (87/107), and 62.5% (15/24) of ALM. Conversely, more LM were detected by doctors than by patients: 61.9% (13/21) vs. 38.1% (8/21).
The location of the melanoma also had an influence on circumstances of diagnosis (χ2p = 0.05). The proportions of melanomas detected by doctors were as follows: 43.5% (20/46) for melanomas located on the face, 12.5% (1/8) for melanomas on the soles, 21.5% (47/217) for melanomas on the lower limbs, 31.5% (24/76) for melanomas on the upper limbs, 25% (15/59) for melanomas on the anterior trunk, 34% (48/142) for melanomas on the posterior trunk. The distribution of other tumor characteristics (color, ulceration, histological regression, histological remnant of a pre-existing nevus) was not significantly different between doctor-detected and self-detected melanomas.
A few studies have focused on medical delays in melanoma diagnosis (Krige et al.,1991; Cassileth et al.,1988; Rampen et al.,1989; Temoshok et al.,1984). However, our prospective study, performed in a large sample of melanomas, assessed separately different components of these delays. Our data show that the prognosis of melanoma patients is generally little influenced by medical delays. However, a few physicians still have inappropriate attitudes toward melanoma, and melanoma with an unusual presentation or located in hidden areas are still diagnosed very late. Training is certainly an important issue since dermatologists tend to detect and manage more rapidly and more correctly early melanoma than other physicians.
Our data are certainly linked to the physician-training system and to the health-care system in France, where access to specialists is immediate and free. However, our results are in line with those of other publications from Europe (Rampen et al.,1989), South Africa (Krige et al.,1991), and the United States (Cassileth et al.,1988). They can thus be extrapolated to other developed countries in which the medical system is similar.
One-third of the melanomas (29.1%) were considered coincidentally detected by a doctor. In other words, the patient did not ask for any medical advice in one-third of melanomas. However, in 12.2% of these coincidentally diagnosed melanomas, detection of the tumor was not really “coincidental” since it occurred during a skin examination for moles. As previously noticed (Cassileth et al.,1988; Rampen et al.,1989; Krige et al.,1991, Koh et al.,1992; Epstein et al.,1999), these melanomas were much thinner than those detected by patients. Our study showed that this benefit was certainly due to an earlier detection (time interval before d3) since time interval after first medical examination (d3–d5) was not different for doctor-detected and self-detected melanomas. Although a self-examination can be done every day and a medical examination is necessarily more unusual, a skin examination by a doctor is still a major way to detect early melanoma.
As previously reported (Cassileth et al.,1988; Krige et al.,1991), delays in the diagnosis of melanoma under the responsibility of physicians are apparently much shorter than those due to patients. However, medical delays and patient delays may not have the same impact. Indeed, one may suspect that the kinetics of tumor growth increase with time. In other words, a 3-month patient delay before the first visit to a doctor may be associated with a lower increase in thickness than a 1-month medical delay in the diagnosis and removal of the tumor. Whatever the respective impact of medical- and patient-related delays, we have shown that tumor thickness is poorly correlated with medical delays (Richard et al.,1999), suggesting that the impact of medical delays on prognosis is weak. Furthermore, although d3–d4 and d4–d5 time intervals were considered medical delays, the d4–d5 interval was also probably influenced by patients. Once a physician has proposed tumor removal, the patient may escape and delay the resection for different reasons. Indeed, patients were more frequently responsible (80.5%) for a long d4–d5 delay than doctors. Taken together, these data show that, in a developed country with easy access to medical care and a good awareness of doctors, physician responsibility in melanoma prognosis has become very limited.
However, doctor delays are not always short. All doctors do not have the same ability to diagnose. The information they give to patients is not always complete, and their attitudes are not always relevant. All of these points deserve discussion.
Our study showed that 14.2% of doctors did not propose removal of the melanoma when they first saw it. This proportion is in line with other studies in Europe and South Africa (Rampen et al.,1989; Krige et al.,1991). Our data suggest that the medical examination in routine practice in France has a sensitivity of 86% for the diagnosis of melanoma. Medical accuracy for the clinical diagnosis of melanoma has been estimated to be 50% to 97% (Whited and Grichnik, 1998) by testing physicians with pictures of pigmented lesions (Cassileth et al.,1986) or with a panel of patients (Lindelöf et al.,1998; Burton et al.,1998). However our estimation is probably more representative of the actual sensitivity for melanoma diagnosis in daily practice than these estimations in experimental situations.
Doctors do not have the same ability to diagnose all types of melanoma. Tumor characteristics strongly influence medical delays. These delays tend to be longer and the number of physicians seen before final diagnosis tends to be higher in melanomas with unusual presentation (such as amelanotic melanoma), melanomas that did not seem to appear from a pre-existing nevus, and ALM than in more common types of melanoma. In the study of Rampen et al. (1989), melanoma suspicion by the physician was poorer and tumor thickness higher for the 26% amelanotic melanomas than for pigmented melanomas. Similarly, Cassileth et al. (1988) reported that time from initial visit to diagnosis was significantly shorter for pigmented lesions than for amelanotic tumors. In the study by Krige et al. (1991), medical delays tended to be short for NM and LM and long for ALM. As previously reported (Temoshok et al.,1984; Rampen et al.,1989), we found that medical delays were usually shorter when tumors were easily visible than when they were located in areas hardly seen in a quick examination. Using the ratio of doctor-detected to self-detected melanomas as a surrogate to the ability of physicians to diagnose a given type of melanoma, the high ratio in LM and the low ratio in ALM confirm that physicians detect easily visible tumors and tend to miss tumors that are not easily visible at first glance. Medical delays also tended to be longer for de novo melanomas than for melanomas on nevi. One possible explanation is that many physicians consider that a melanoma always grows from the transformation of a mole. Taken together, these results suggest that early diagnosis of melanoma could be further improved by shortening medical delays in particular subtypes or locations of the tumor. Training of physicians should alert them to the possibility of unusual presentation of melanoma and the need for a complete skin examination including hidden areas in any medical examination. Indeed, in 1988, half of the physicians in the United States reported that they did not examine the entire cutaneous surface (Cassileth et al.,1988).
Doctors can also be responsible for delays before melanoma removal, when they do not correctly inform patients or do not provide the possibility of rapid resection. Once removal of the melanoma is proposed by the physician, 12.2% of patients wait for more than 1 month before resection. In one-fifth of these cases, the patients alleged that the physician did not correctly inform them about the need for a rapid resection. This is, however, the point of view of the patient, who may tend to alleviate his own responsibility. In patients who waited because of anxiety or for other reasons which may hide anxiety, a better psychological approach or better information may have also reduced the delay. The management of the psychological impact of cancer suspicion should also be addressed in doctor training.
In our study, only 0.6% of tumors were not initially submitted to histological control and 10% were not treated in an appropriate way (partial biopsy, cryotherapy, cauterization, shaving). This is probably an under-estimation since we included only primary melanomas in this study and thus do not take into account melanomas discovered at recurrent or metastatic disease after several medical errors. This rate of inadequate attitude is much lower than that presented by Rampen et al. (1989). They reported a first treatment without histopathological examination in 13.5% of melanomas. Although results in the Netherlands and France are not necessarily comparable, knowledge of physicians may have improved in recent years, as suggested by Salopek et al. (1995).
The first physician who discovered the lesion or to whom the lesion was submitted was a dermatologist in 54.7% and a GP in 33.4% of cases. These results are comparable to those obtained in the United States by Cassileth et al. (1988). They showed that 50% of patients sought the advice of a dermatologist and 20% the advice of a GP. The proportion of patients whose melanoma was detected by a dermatologist is similar in the French and the American health systems, though the density of dermatologists and the access to medical resources are quite different in the 2 countries.
In our study, tumor thickness was lower in melanomas diagnosed by a dermatologist than by a GP or another type of physician. Whether this is due to better training of dermatologists or to a difference in the population of patients which usually visit dermatologists or GPs is unclear. However, the role of training is probably the most relevant hypothesis. First, several studies have shown that dermatologists are more accurate in diagnosing pigmented lesions than GPs (Ramsay and Fox, 1981; Cassileth et al.,1986, 1988; Rampen et al.,1989; Langley and Sober; 1997). Second, our data showed that initial attitude was more often appropriate and medical delays were significantly shorter when a melanoma was first seen by a dermatologist than when it was seen by a GP. This was true regardless of the histoclinical type and the clinical characteristics of the melanoma. In a multivariate analysis, the most predictive factor of a short d3–d4 interval was first examination by a dermatologist. Taken together, these results not only show that dermatologists are the most competent physicians for melanoma detection but also confirm that medical training is actually converted into a benefit in normal practice.
We are grateful to C. Proust (Service de Dermatologie, Institut Gustave Roussy, Villejuif, France), A. Spatz (Département d'Histopathology, Institut Gustave Roussy), H. Benchiki (Service de Dermatologie, Hôpital Henri-Mondor, Créteil, France), B. Legoux (Service de Dermatologie, Hôpital Hôtel-Dieu, Nantes, France), M. Balutet (Service de Dermatologiel, Hôpital Pellegrin, Bordeaux, France), M.F. Ferrando and I. Couffignal (Service de Dermatologie, Hôpital Haut-Lévèque, Pessac, France), I. Le Hir (Service de Dermatologie, Hôpital Hôtel-Dieu, Rennes, France), R. Choux (Laboratoire d'Anatomie-Pathologique, Hôpital Sainte Marguerite, Marseille, France), and J. Pellegrini and C. Bertorello (Service d'Informatique Médicale, Hôpital Sainte Marguerite) for their contribution to this study.