A hypothesis generated a retrospective analysis of the improvement of survival over time in patients with cutaneous melanoma (CM) in order to identify factors contributing to this progress.
A hypothesis generated a retrospective analysis of the improvement of survival over time in patients with cutaneous melanoma (CM) in order to identify factors contributing to this progress.
A cohort of 4791 patients diagnosed with primary CM in southern Germany between 1976 and 2001 was analyzed. Kaplan-Meier analyses were performed to estimate and to compare overall survival (OS) and survival after first recurrence (SAR). The Cox proportional hazards model was used to evaluate the effect of multiple variables on OS and SAR.
Diagnosis of primary CM during 1990–2001 compared with 1976–1989 was associated with more favorable 10-year OS (88.6% vs 80.0%, P < .0001). The median tumor thickness at primary diagnosis was significantly lower during the second period (0.75 mm vs 1.07 mm, P < .0001). In the multivariate analysis, adjusted for tumor thickness, ulceration, age, gender, and anatomical site, the period of diagnosis retained its significance as a predictor of OS (P = .002). The SAR was more favorable during the second period in patients who developed their first metastasis in the regional lymph nodes (45.2% vs 32.9%, P = .017) or in distant sites (13.4% vs 2.0%, P = .0002) and this finding persisted in the multivariate analysis.
Improvement of survival of CM patients diagnosed 1990–2001 as compared with 1976–1989 may not be entirely attributable to factors associated with early diagnosis and more favorable primary tumors. Therefore, factors of melanoma management, which changed between the 2 time periods, such as sentinel node biopsy, adjuvant treatment, structured follow-up, and surgical interventions in distant metastasis, may have to be taken into account. Cancer 2007 © 2007 American Cancer Society.
Cutaneous melanoma (CM) is a potentially lethal neoplasm with a propensity for distant metastasis. The outcome of patients diagnosed with invasive CM is directly related to early detection. Whereas the overall prognosis of early diagnosed thin tumors is excellent, once metastasized melanoma is an incurable disease with high mortality.1–3 During the past 2 generations the incidence of CM has risen dramatically among Caucasian populations mainly due to diagnosis of thin, prognostically favorable lesions, whereas the incidence of advanced, thick tumors generally remained stable.4–6 Consequently, stabilization or even leveling off of the mortality from melanoma was reported in European, American, and Australian populations.4–10
During the same period a number of new diagnostic and therapeutic modalities emerged in the field of melanoma. Standard excision with safety margins was established and a new staging system was implemented so as to take into account the new prognostic information gained from lymphatic mapping and sentinel lymph node biopsy. In our institution, a new standardized surveillance program was introduced for patients with CM in the early 1990s. This follow-up scheme may have facilitated earlier detection of recurrences with the routine use of new diagnostic modalities like lymph node ultrasound.11 Additionally, various combinations of chemotherapeutics were studied in numerous trials and a large number of patients received interferon as adjuvant therapy for melanoma. So far, it has not been shown conclusively that patient management has contributed to any survival benefit. Herein, we analyzed the survival of a cohort of 4791 patients diagnosed with invasive CM in southern Germany during the years 1976–2001 in order to analyze the factors associated with improvement of survival.
We included in the analysis 4791 patients of German origin, according to their place of birth, who were diagnosed with primary invasive CM at the University Medical Center of Tuebingen in southern Germany during the years 1976–2001. Data were obtained from the electronic database of the Central Malignant Melanoma Registry, which is a hospital-based registry. Only patients without clinical evidence of metastasis after all investigations performed at the time of primary diagnosis were included in the study. Patients with positive sentinel lymph nodes were also included because this information was not available for the early years of the study. Histologic parameters were abstracted from the pathology reports around the time of diagnosis of the primary tumor. Abstraction was made by the specialized dermatologist who was in charge of the Central Malignant Melanoma Registry. The histologic diagnosis of a nodular melanoma was made by the dermatopathologists of our institution when the growth pattern was monophasic or pure vertical. This infiltration sometimes coexists with intraepidermal growth extending no more than 3 rete ridges from the nodule. Tumor characteristics and case history were recorded in a standardized manner and patients were examined regularly every 3 to 6 months for a period of 10 years. After 1990, the follow-up protocol was carried out according to the guidelines established by the German Society of Dermatology.8 Follow-up time was cut to a maximum of 10 years, as this is the maximum duration of the regular follow-up of CM patients. After 10 years only patients with progressive disease are examined and little information can be retrieved for patients without recurrences. Tumor thickness was analyzed in 4 subgroups (<1.00 mm, 1.01–2.00 mm, 2.01–4.00 mm, >4.00 mm). Age was classified into 4 subgroups (<35 years, 35–49 years, 50–65 years, >65 years).
Follow-up time was defined as the date of last follow-up or death minus date of operation (overall survival, OS) or the date of diagnosis of first recurrence (survival after recurrence, SAR). Recurrence-free survival time (RFST) was defined as the date of diagnosis of first recurrence minus the date of operation. Follow-up time was described as a median value with interquartile range (IQR). Only deaths caused by CM were considered in survival analysis. Melanoma-specific survival curves and estimated survival rates with relative standard errors (SE) were generated according to the Kaplan-Meier product-limit method and were compared by the 2-sided log-rank test. Variables proven significant in the univariate analyses were included in a Cox proportional-hazards model using both forward and backward stepwise selection methods that resulted in the same model. Age and RFST (years) were entered in the multivariate analysis as continuous variables. Categorical variables were dummy-coded. The Cox model was described by means of hazard ratios (HRs) together with 95% confidence intervals (95% CIs) and P-values were based on the Wald test. We examined the graphic plot of the log-log survivor function for the variables entered in the analysis. No violation of the proportional hazards assumption was found. Throughout the analysis 2 sided P-values less than 0.05 were considered statistically significant. All statistical analyses were performed with the Statistical Package for Social Sciences (SPSS) v. 13.0 (SPSS, Chicago, IL).
Overall there were 4791 patients with primary incident invasive (Clark's level of invasion ≥II) CM without clinical evidence of metastasis. The proportion of male patients increased from 40.7% during 1976–1989 to 45.3% in 1990–2001 (P = .003). The median age at the time of primary diagnosis did not differ significantly between the 2 periods (50.2 vs 51.6, P = .7). The most prominent anatomical site of CM overall was the trunk, followed by the lower limbs. The distribution of anatomical site of CM did not differ significantly between 1976–1989 and 1990–2001 (P = .068). The most common histopathologic type of CM was superficially spreading melanoma followed by nodular melanoma. The proportion of superficially spreading melanomas increased from 64.2% in 1976–1989 to 74.6% in 1990–2001 (P < .0001). During the same period, the proportion of nodular melanomas decreased from 20.2% to 11.4%. The median tumor thickness decreased from 1.07 mm in 1976–1989 to 0.75 mm in 1990–2001 (P < .0001). The median follow-up of uncensored cases was 3.8 years (IQR, 4.8) during 1976–1989 and 3.3 years (IQR, 3.0) during 1990–2001. The median follow-up of censored cases was 9.9 years (IQR, 1.6) during 1976–1989 and 6.1 years (IQR, 4.5) during 1990–2001 (Table 1). The disease-related mortality was 1.9/100 person-years during 1976–1989 (95% CI: 1.6–2.1) (n = 217) and 1.2/100 person-years during 1990–2001 (95% CI: 1.0–1.3) (n = 231).
|1976–1989 n = 1487||1990–2001 n = 3304||P*||Overall survival 10-Year||P‡|
|Median (IQR§)||50.2 (24.0)||51.6 (27.0)||.700|
|Upper limbs||16.9||14.8||91.5 (1.2)|
|Lower limbs||32.1||31.1||88.3 (1.0)|
|Breslow tumor thickness|
|<1.00 mm||47.9||64.8||.00001||96.4 (0.5)||.00001|
|1.01–2.00 mm||26.8||19.7||84.1 (1.4)|
|2.01–4.00 mm||18.6||11.2||65.4 (2.3)|
|>4.00 mm||6.7||4.3||50.4 (3.9)|
|Median (IQR)||1.07 (1.4)||0.75 (0.9)||—||—||—|
|Level of invasion|
|II and III||27.6||33.7||.00001||93.5 (0.6)||.00001|
|IV and V||72.4||66.3||75.8 (1.4)|
|Follow-up years (median, IQR)|
|Uncensored cases||3.8 (4.8)||3.3 (3.0)||.760‖||—||—|
|Censored cases||9.9 (1.6)||6.1 (4.5)||.00001‖||—||—|
|Time period of diagnosis|
Kaplan-Meier overall survival probabilities were significantly different for age (P < .0001), gender (P < .0001), anatomical site (P = .002), tumor thickness (P < .0001), level of invasion (P < .0001), ulceration (P < .0001), and histopathologic subtype (P < .0001). There was a significant improvement in the 10-year OS in patients who suffered from primary CM during 1990–2001 compared with 1976–1989 (88.6% vs 80.0%, P < .0001) (Fig. 1). Multivariate Cox analysis revealed that the independent predictors of OS were: tumor thickness (P < .0001), presence of histopathologic ulceration (P < .0001), age (P = .028), gender (P = .012), anatomical site (P = .0002), and period of primary diagnosis (P = .002).
There were 331 patients diagnosed with first CM recurrence in the regional lymph nodes. The distribution of gender, age, anatomical site of primary melanoma, level of invasion, and histopathologic ulceration did not differ significantly between the 2 periods (P ≥ .05 for each variable). In this group of patients the proportion of thin melanomas (≤1.00 mm) increased significantly from 5.5% during 1976–1989 to 19.9% during 1990–2001, coinciding with a decrease of thicker tumors (P = .001). Accordingly, the proportion of nodular melanoma during 1990–2001 was significantly lower compared with 1976–1989 (P = .003). The median follow-up of patients from the time of diagnosis of first recurrence (SAR) was 1.9 years (IQR, 4.9 years) during 1976–1989 and 2.7 years (IQR, 4.2 years) during 1990–2001 (P = .225) (Table 2). The disease-related mortality (assessed from diagnosis of primary melanoma) was 13.0/100 person-years during 1976–1989 (95% CI: 10.2–15.8) (n = 72) and 9.1/100 person-years during 1990–2001 (95% CI: 7.0–9.8) (n = 121).
|1976–1989n = 110||1990–2001n = 231||P||5-Year OS||5-Year SAR|
|Percentage (SE)*||P†||Percentage (SE)*||P†|
|Follow-up, y (median, IQR‡)|
|RFS||1.2 (2.1)||2.2 (3.1)||.020§||—||—||—||—|
|SAR||1.9 (4.9)||2.7 (4.2)||.225§||—||—||—||—|
|OS||3.9 (7.0)||5.4 (7.0)||.021§||—||—||—||—|
|Time period of diagnosis of recurrence|
|1976–1989||—||—||—||49.0 (4.9)||.008||32.9 (4.7)||.017|
|1990–2001||—||—||—||61.9 (3.3)||45.2 (3.5)|
Kaplan-Meier OS probabilities were significantly different for tumor thickness (P = .0001), level of invasion (P = .003), ulceration (P = .03), histopathologic subtype, and year of diagnosis of the recurrence (P = .008). Kaplan-Meier SAR probabilities were significantly different for Breslow thickness (P = .002) and year of diagnosis of the recurrence (P = .017). Patients diagnosed with regional lymph node recurrence during 1990–2001, as compared with 1976–1989, had significantly higher 5-year OS (P = .008) and SAR (P = .017) (Table 2, Fig. 2A,B). Multivariate Cox regression analysis revealed that the independent predictors of the OS were the RFST (P < .0001) and the time period of diagnosis of first recurrence (P = .015). The independent predictors of the SAR were the RFST (P < .0001) and the time period of diagnosis of first recurrence (P = .023).
There were 192 patients diagnosed with first CM recurrence in distant anatomical sites. The distribution of gender, age, anatomical site of primary melanoma, tumor thickness, level of invasion, histopathologic ulceration, and anatomical site of distant metastasis did not differ significantly between the 2 periods (P ≥ .05 for each variable). The median follow-up of patients from the time of diagnosis of first recurrence (SAR) was 0.4 years (IQR, 0.9 years) during 1976–1989 and 0.6 years (IQR, 1.2 years) during 1990–2001 (P = .002) (Table 3). The disease-related mortality (assessed from diagnosis of primary melanoma) was 25.5/100 person-years during 1976–1989 (95% CI: 19.7–31.3) (n = 54) and 13.3/100 person-years during 1990–2001 (95% CI: 10.8–15.8) (n = 91).
|1976–1989 n = 58||1990–2001 n = 134||P*||5-year OS||5-year SAR|
|Percentage||Percentage||Percentage (SE)†||P‡||Percentage (SE)†||P‡|
|Site of distant metastasis|
|Skin-lymph nodes||17.2||13.4||.492||43.0 (4.0)||.475||13.4 (3.1)||0.01|
|Visceral||82.8||86.6||40.9 (9.6)||8.5 (2.9)|
|Follow-up, y (median, IQR§)|
|RFS||2.0 (2.8)||3.3 (4.6)||.030‖||—||—||—||—|
|SAR||0.4 (0.9)||0.6 (1.2)||.002‖||—||—||—||—|
|OS||3.1 (2.7)||4.4 (5.1)||.120‖||—||—||—||—|
|Time period of diagnosis of recurrence|
|1976–1989||—||—||—||23.9 (5.7)||.00001||2.0 (2.0)||.0002|
|1990–2001||—||—||—||51.1 (4.5)||13.4 (4.0)|
Kaplan-Meier OS probabilities were significantly different for tumor thickness (P = .0003), level of invasion (P = .043), ulceration (P = .0005), histopathologic subtype (P = .043), and time period of diagnosis (P = .0001). Kaplan-Meier SAR probabilities were significantly different for site of distant metastasis (P = .01) and time period of diagnosis of the recurrence (P = .0002). Patients diagnosed with distant CM recurrence during 1990–2001 as compared with 1976– 1989 had significantly higher 5-year OS (P = .0001) and SAR (P = .002) (Table 3, Fig. 2C,D). Multivariate Cox analysis revealed that the independent predictors of the OS were the RFST (P = .017) and the time period of diagnosis of first recurrence (P = .0001). The independent predictors of the SAR were the site of distant metastasis (P = .045) and the time period of diagnosis of first recurrence (P = .001).
In this study we analyzed the survival of a cohort of 4791 CM patients diagnosed at the University Medical Center of Tuebingen/Stuttgart during a period of 26 years starting from 1976. The well-established prognostic factors of CM (tumor thickness, ulceration, gender, age, anatomical site) were confirmed in our sample and all revealed strong trends toward tumors with favorable prognosis. Consequently, the survival rates of CM patients increased by approximately 10% during the study period. These advances have also been described for other countries in Europe, Australia, and the US and have been mainly attributed to early diagnosis of primary tumors.2, 4, 5, 8–10, 12
The negligible proportion of “Not otherwise specified melanomas” or “missing cases” observed here is in contrast with reports from general cancer registries of Europe, the US, and Australia, where the proportions of histologically unspecified melanomas may reach up to 30% of the total.6, 13 However, in the University Department of Dermatology in Tuebingen all external histopathologic diagnoses are scrutinized by internal reference histopathology with a standardized reporting schedule. Thus, the correctness and the completeness of the data generated here may be better as compared with the normal population-based cancer registries. The variable ulceration of the primary tumor, however, may have been underreported during the first time period, because at that time its prognostic significance was not yet established.
Interestingly, we found that the survival of patients diagnosed since 1990 seemed to have improved beyond the effects of early diagnosis as characterized by established prognostic factors. This could be explained by changes in unmeasured biologic features of melanoma or by improvements in the management of this disease. The treatment of choice of primary CM is wide local excision. Standard excision with safety margins was established during the last 30 years as the therapeutic procedure for CM, but the only progress that has been made in the field of primary surgery for melanoma was a narrowing of the appropriate excision margins from 5 cm during the 1970s and 1980s to 1–2 cm since the early 1990s.14–16 It is unquestionable that this advancement significantly reduced morbidity associated with the treatment of melanoma but it is rather unlikely that it has contributed to the improvement of the survival found in this study.
Until the early 1990s, patients presenting with primary CM melanoma without clinical evidence of metastasis often underwent elective lymph node dissection as a staging and therapeutic procedure. This was based on the hypothesis that melanoma cells spread to regional nodal basins before metastasizing widely and that early removal of nodal deposits may prevent subsequent dissemination. Consequent trials designed to test this hypothesis failed to show a benefit of elective lymphadenectomy in OS and this practice was largely abandoned with the emergence of sentinel node biopsy as a staging and possibly therapeutic procedure in the management of melanoma.17–19 The introduction of sentinel lymph node biopsy into clinical practice after 1990 provided accurate assessment of the regional lymph node status, avoiding full regional lymphadenectomy in a large proportion of patients with negative sentinel nodes.20 The prognostic value of sentinel lymph node status has been confirmed in many studies but until now there is no evidence that removal of these nodes, with complete regional node field dissection if micrometastatic melanoma is found, improves survival. The results of large multicenter trials that will assess this question are expected with great interest.21 Until then, it cannot be excluded that the improvement of the outcome in patients with melanoma reported here is, at least in part, due to the therapeutic impact of the sentinel lymph node biopsy and consequent elective lymph node dissection. Additionally, a significant number of patients without clinical evidence of metastasis at the time of primary diagnosis after 1990 were upstaged utilizing information gained from the sentinel lymph node biopsy, and many of these patients were identified as candidates for systemic adjuvant therapy with interferon. Whether interferon therapy can improve the outcome of patients with microscopic nodal disease remains to be clarified.21, 22
Approximately 7% of patients diagnosed with primary melanoma experienced tumor progression to the regional lymph nodes. The most important prognostic factor of these patients was the RFST previous to lymph node metastasis. Age, gender, and the clinicopathologic characteristics of primary tumor were not significant in the multivariate analysis. The number of the involved lymph nodes, which is a well-recognized prognostic marker in patients with regional lymph node metastasis, was not available for analysis in our study.1 Of particular interest is the finding that patients diagnosed with recurrent disease in the regional lymph nodes during 1990–2001 as compared with 1976–1989 had more favorable survival. This finding persisted in the multivariate analysis adjusted for clinicopathologic characteristics of the primary tumor.
Patients with recurrence to the regional lymph nodes are treated with radical lymph node dissection and are candidates for adjuvant systemic therapy. The only substance that has been reproducibly shown to affect disease behavior is systemic interferon-alpha (IFN-α).23–25 The available weight of evidence does not indicate that therapy with IFN-α produces a benefit in terms of long-term OS.26 However, both recurrence-free survival and short-term survival are improved significantly.27 The possibility that the use of interferon has contributed to the improvement of the overall survival in patients with regional lymph node recurrence cannot be totally excluded, but the use of IFN-α as standard adjuvant therapy in melanoma still remains highly controversial in clinical oncology. In Europe there is little enthusiasm for adopting IFN-α treatment as the standard of care because of its significant toxicity, its high cost, and a lack of clear benefit on long-term survival.28
Approximately 4% of patients diagnosed with primary melanoma experienced the first tumor progression to distant anatomical sites. The outcome of these patients is generally very poor and depends mainly on the location of metastases.1 Even after adjusting for site of metastasis, patients who developed primarily distant disease during 1990–2001 as compared with 1976–1989 had a significantly more favorable outcome. The only treatment that is known to significantly prolong the survival of patients with stage IV melanoma is complete surgical resection of metastases.29–32 Patients with resectable distant disease represent a small proportion of patients with stage IV melanoma and are more likely to have early-detected, isolated, or small in size and number of metastases.30–32
Since the early 1990s the screening for metastasis in our institution was performed according to the guidelines of the German Society of Dermatology, which suggested intense follow-up examinations including imaging techniques with the routine use of lymph node ultrasound.33 This follow-up strategy was recently evaluated in our institution and yielded a high proportion of detection of metastasis by the physicians carrying out the follow-up examinations.11 Additionally, half of recurrences were classified as early detection, indicating the possibility of surgical resection with curative intent. The same study demonstrated a survival benefit of patients with early-detected recurrences, which, however, remains to be established in comparative randomized trials.11 Until then, we cannot exclude the possibility that the improved survival of CM patients with progressive disease after 1990 reported here is, at least in part, due to earlier detection of recurrences in the frame of a more intense scheme for metastasis screening.
The hypothesis of beneficial effects of management factors in the survival of patients with melanoma reported here depends mainly on the lower stage-adjusted hazard of patients diagnosed with primary tumor or melanoma recurrence during the second period. This study neither aimed at nor had the strength to identify which management factors contributed to the survival improvement after 1990. This has to be investigated with targeted prospective studies. Additionally, the results of our study have to be interpreted carefully because of the presence of some sources of bias. First of all, the duration of follow-up of patients diagnosed with primary melanoma during 1990–2001 was systematically shorter as compared with 1976–1989. This might have contributed to a lower proportion of events observed during the second time period. In order to decrease this type of bias we included only cases diagnosed until 2001 in our analysis in order to have at least 5 years of follow-up (evaluation 2006).
Our study is also limited by the presence of lead and length time bias, which affects most studies analyzing the trends of the survival of CM patients. Regardless of whether treatment is available and effective, earlier detection will artificially increase survival by bringing forward the time of diagnosis (lead-time bias), whereas slow-progressing lesions with less aggressive potential will preferably be detected (length-time bias, overdiagnosis). We tried to reduce this bias as much as possible. First, all histopathologic slides from external institutions were requested to be sent to our institution and were reevaluated by the board of histopathologists in our institution. There has been a consistent reporting of melanoma diagnosis and its prognostic factors throughout the entire time period. This was controlled, for instance, by comparisons of tumor thickness and invasion levels throughout this time period. Second, we used multivariate analysis to adjust for detailed stage of the disease. Third, we analyzed survival time from the time of first diagnosis of the primary melanoma following the example of Balch et al.34 who used similar methods to adjust for lead-time bias at stage III disease in the latest AJCC staging system of melanoma. Finally, we had no way to take into account significant lifestyle changes that occurred in our population during the last 30 years. For example, increased sun exposure was recently suggested by Berwick et al.35 as a positive prognostic factor in patients with CM and sun exposure probably has increased in the German population over this period.
In conclusion, the OS of CM patients diagnosed between 1990 and 2001 as compared with patients diagnosed between 1976 and 1989 was clearly more favorable. Well-established prognostic factors were examined in this study and all showed strong trends toward early detection and diagnosis of tumors with more favorable prognosis. However, this study provides evidence for the first time that the improvement in OS may not be exclusively attributable to earlier diagnosis of CM. Multivariate analysis revealed the more recent time period of diagnosis to be an independent prognostic factor in addition to the effects of early diagnosis. Better outcome of patients who developed primarily regional lymph node or distant metastasis during the second period likewise contributed to the observed differences. Therefore, differences in management of melanoma patients between the 2 time periods may have contributed to this improvement of prognosis of melanoma patients in all tumor stages. Possible elements of melanoma management changes such as structured follow-up examinations and early detection of recurrences, the introduction of the sentinel node biopsy, adjuvant IFN-α treatment, and a more frequent surgical approach to distant metastasis require further investigation. Prospective randomized trials are required to test these hypotheses.