Incidence and mortality trends of primary cutaneous melanoma: A 50-year Rochester Epidemiologic Project study

Background National cancer reporting-based registry data, although robust, lacks granularity for incidence trends. Expert opinion remains conflicted regarding the possibility of melanoma overdiagnosis in the context of rising incidence without a corresponding rise in mortality. Objective To characterize 10- and 50-year trends in melanoma incidence and mortality. Methods Multicenter, population-based epidemiologic study utilizing the Rochester Epidemiology Project for Olmsted County, Minnesota residents diagnosed with melanoma from 01/01/1970 to 12/21/2020. Age- and sex-adjusted incidence and disease-specific mortality are calculated. Results Two thousand three hundred ten primary cutaneous melanomas were identified. Current age- and sex-adjusted incidence rates increased 11.1-fold since 1970s (P < .001). Over the last decade, there is an overall 1.21-fold (P < .002) increase, with a 1.36-fold increase (P < .002) among females and no significant increase among males (1.09-fold increase, P < .329). Melanoma-specific mortality decreased from 26.7% in 1970s to 1.5% in 2010s, with a hazard ratio (HR) reduction of 0.73 (P < .001) per 5-year period. Increased mortality was associated with Breslow thickness (HR 1.35, P < .001), age at diagnosis (HR 1.13, P = .001) left anatomic site (HR 1.98, P = .016), and nodular histogenic subtype (HR 3.08, P < .001). Limitations Retrospective nature and focused geographic investigation. Conclusion Melanoma incidence has continued to increase over the past decade, most significantly in females aged 40+. Trend variations among age and sex cohorts suggests external factors beyond overdiagnosis may be responsible. Disease-specific mortality of melanoma continues to decrease over the last 50 years.


INTRODUCTION
][8] Melanoma poses an estimated 5-year relative mortality rate of 6.3%. 1 Although mortality trends had previously remained largely stable, some reports tracking SEER data between 2009 and 2018 indicate decreasing mortality between 17.9% and 30%. 2,9espite the crucial nature of national registries such as the SEER database, delays and underreporting are inherently prevalent. 10urthermore, SEER registries cover only 35% of the US population. 11Accordingly, high-quality populationbased studies are critical to strengthen the validity of national database findings.
3][14] These data reinforced the continued rise in melanoma incidence across all age groups, but highest among women aged 40 to 60 years old (24-fold increase), 13 and men aged 611 (11-fold increase). 14][14] A 2021 article in the New England Journal of Medicine utilizing national database data argued that relative stability in disease-specific mortality with rising incidence signifies overdiagnosis. 15[14]

Patient selection
Patients aged 181 with an initial lifetime diagnosis of primary cutaneous melanoma between January 1, 1970 and December 21, 2020 were identified using the Rochester Epidemiology Project (REP).
The REP is an aggregated multicenter medical record database for all residents of Olmsted County, Minnesota who obtain care from Mayo Clinic, Olmsted Medical Center, and other affiliated private practitioners.The comprehensive county-wide tracking features of the platform provide validity for epidemiologic studies. 16This patient population is largely non-Hispanic Caucasian with a comparable socioeconomic status to the entire United States and a population of 162,847 residents in 2020. 17EP patients were identified through ICD codes for malignant melanoma, followed by detailed manual chart review for verification.
Patients with a prior history of cutaneous melanoma, those with ocular or mucosal melanoma, or those who denied research authorization were excluded.Data collection identified melanoma tumor characteristics (Breslow thickness, histogenic subtype, anatomic site, and American Joint Committee on Cancer [AJCC] staging eighth edition), patient demographics, and mortality data.Individual death certificates were viewed to verify melanoma as a cause of death.[14]

Statistical analysis
Data analysis was completed from March 2022 to October 2023.Age-and sex-specific incidence rates (per 100,000 person-years) in Olmsted County were calculated based off REP data and adjusted to the total population of the United States.The 95% CI for incidence rates were calculated assuming a Poisson error distribution.Differences in incidence rates by sex, age group, and calendar period were assessed by fitting Poisson regression models to the incidence counts using the natural logarithm of the total person-years as an offset.
Cause-specific (ie, death due to melanoma) mortality rates were estimated using the Kaplan-Meier method.Univariable and multivariable associations with the risk of death due to melanoma were evaluated using Cox proportional hazards regression models and summarized with hazard ratios (HRs) and 95% CIs.Variable selection for these models was determined based on clinical expertise, previous literature, and the univariable analysis.Statistical significance was defined at a P \ .05.All statistical analyses were performed using RStudio Version 4.1.3(RStudio).

CAPSULE SUMMARY d
Rising incidence rates of primary cutaneous melanoma are welldocumented over the past 50 years, with conflicting expert opinion about the possibility of overdiagnosis.
When separated by aged cohorts (Fig the highest overall incidence of melanoma.The greatest rise in incidence over the most recent 10-year period (2005-2009 vs 2025-2020) is observed in females aged 40 to 60 and 611 (1.5-fold increase, P = .002;and 2.7-fold increase, P \ .001respectively).[20] Mortality and survival Melanoma-specific death was observed in 100 patients, accounting for an overall disease specificmortality rate of 4.3%.Of those, 93 deaths occurred within 10 years following melanoma diagnosis.Disease-specific death was not observed for melanoma in-situ.Table II displays disease-specific mortality related to numerous variables, with univariable and multivariable HR calculations.According to multivariable analysis, an increase in disease-specific mortality was significantly associated with the following: higher age at diagnosis (HR 1.13 per 5-year increase in age, 95% CI, 1.05-1.22,P = .011),increased Breslow thickness (HR 1.34, 95% CI, 1.25-1.43,P \ .001),left-sided anatomic site (HR 1.98, 95% CI, 1.14-3.34,P = .016),and nodular histogenic subtype (HR 3.08, 95% CI, 1.75-5.42,P \.001).Characteristics associated with an increase in disease-specific mortality, but without statistical significance upon multivariable analysis include male sex (HR 1.53, 95% CI, 0.91 to 2.56, P = .107)and anatomic site of head or neck (HR 1.48, 95% CI, 0.83-2.66,P = .184).Increasing pathologic stage (AJCC eighth edition) was associated with statistically significant increase in disease-specific mortality upon univariable analysis but was not utilized for multivariable analysis.Caucasian race was not significantly associated with mortality (HR 1.39, P = .741).A variable importance plot depicting the relative importance of each variable according to diseasespecific mortality is shown in Fig 5.

Incidence
3][14] Our results demonstrate a continued increase in age-and sex-adjusted melanoma incidence since 2009, which is largely consistent with recently reported incidence data showing continued increase in melanoma incidence among adults aged 40 years or older. 8ur findings demonstrate that males aged 611 continue to have the highest overall incidence of melanoma, but that females aged 401 are experiencing the greatest growth of incidence over the past decade.Female incidence remains higher than males among adults aged\ 60.Furthermore, the slight rise observed in male incidence over the past 10 years was not statistically significant and suggests a potential stabilization of melanoma incidence among adult males.
It is important to address recent concerns for melanoma overdiagnosis, suggested from rising incidence rates without a corresponding rise in mortality. 15,21,22Multiple groups have attempted to analyze and quantify the magnitude of this proposed overdiagnosis. 23,24A 2024 ecological study examining the SEER database calculated an overestimation of melanoma diagnosis in 2018 of 49.7% in white men and 64.6% in white women, with an estimated 85.4% to 89.4% overdiagnosis for melanoma in situ. 24 2022 cohort study used the trends in melanoma mortality among Black patients as a marker for improvements in medical care (a potential theory for declining mortality) and found an estimated overdiagnosis of melanoma in 59% to 60% of white patients. 23Despite using adept epidemiologic methods, the SEER database used in these studies presents inherent limitations with known delays, underreporting, and limited scope covering approximately 35% of the US population. 10,11ur data show converging rates among males and females overall, with stabilization in incidence of males of any age range and females\40 years of age.If overdiagnosis were the primary cause of rising incidence, a more uniform increase in melanoma rates across demographics might be expected.Instead, our data suggest that external demographic-specific risk factors and exposures may influence a rise in incidence among certain subsets of the population.
These findings are unable to confirm or deny the presence of overdiagnosis but support the multifactorial nature of melanoma incidence.Overdiagnosis alone would not fully account for the findings of our study which demonstrates the need to stratify population-based data into sex and age groups to fully elucidate incidence trends.
6][27] Exogenous female hormones have also been suggested as a melanoma risk factor, but with conflicting evidence. 28ltimately, additional high-quality studies addressing risk factors for development of melanoma are warranted.
The present data demonstrating 10.3% of melanomas arising within a pre-existing nevus on histology is on the low end of previous reports ranging from 10.8% 29 to 57.6%. 30Although this may suggest a higher rate of de novo melanomas, it should be noted that 50.4% of our pathology reports described preexisting nevus as unknown.Importantly, these data cannot exclude loss of identifiable histopathologic features of a pre-existing nevus in late transformation of melanomas arising from pre-existing nevus.
The literature suggests a higher prevalence of left-sided melanomas, 31,32 with 1 study of 6 population-based cancer registries showing a left:right (L/R) ratio of 1.10. 33Our study supports this prior data, with left-sided melanomas accounting for 46.5% of tumors, compared with 43.2% on the right (L/R ratio, 1.07).Increased left-sided ultraviolet exposure during vehicle driving has been suggested, but this does not explain the increased left-sided laterality of skin cancer in countries with vehicle driving on the right (Australia, England, Scotland) as seen in Brewster et al 33 or for increased incidence of melanomas on the leg which may receive less UV exposure when driving. 34An additional theory includes the potential asymmetry of melanocytic embryonic development. 33,35

Mortality
Our findings show a decrease in melanomaspecific mortality over the past decade, and a continued decline over the past 50 years.7][38] Improved surveillance efforts have also been suggested as influential in reducing mortality rates of melanoma through the increased detection of thin melanomas. 39ur results show that factors associated with disease-specific mortality include increasing Breslow thickness, higher age at diagnosis, left anatomic site, and increasing AJCC V8 pathologic stage and are consistent with previous reports in the literature. 40,41odular melanoma was the only histogenic subtype found to have a statistically significant increased risk of death due to melanoma.This is consistent with a 2015 study in Europe that demonstrated the lowest survivorship associated with nodular histogenic subtype (highest with lentigo maligna subtype). 41Very thick melanomas ([4.00 mm) have been shown to be disproportionately represented by the nodular subtype 42 which supports the idea that different histogenic subtypes have variations in biologic behavior and expected natural histories.It is possible that Breslow thickness and nodular subtype are confounding variables with respect to melanoma-specific mortality, but our findings further suggest that nodular melanomas present a higher risk of mortality.
Our data also demonstrate a 1.98-fold increase in melanoma-specific mortality associated with leftsided melanomas.Previously discussed theories to explain the higher incidence of left-sided tumors can also be considered to explain the observed association with mortality (increased UV exposure from driving, 31,32 asymmetric embryonic development of melanocytes); however, these postulations explain elevated incidence, rather than mortality. 33,35This relatively high HR warrants further exploration into left-sided melanoma tumorigenesis.
Disease-specific mortality associated with male sex was found to be elevated among a univariable calculation (HR 2.00, P = .001)but when calculated using a multivariable analysis, this risk was reduced and not statistically significant (HR 1.53, P = .107).This suggests that confounding variables may influence the observed increased mortality among male sex such as histogenic subtype, Breslow thickness, and anatomic location.
Anatomic site was not associated with a statistically significant difference in mortality, consistent with previous reports. 41Race was also found to have no significant association with melanoma-specific mortality.The authors suggest a potential limitation in the dataset's statistical power with only 2.2% of melanomas identified in non-White patients.

Limitations
Our study is limited by the focused nature of the REP which tracks residents of Olmsted County, Minnesota.While this population tends to have a similar socioeconomic status to the greater United States, the population is largely non-Hispanic Caucasian and not racially representative of other demographics.The retrospective nature of chart review is inherently limiting.This study did not assess for UV exposure history, family history, or genetic influences as factors for melanoma incidence and mortality.Melanoma incidence and mortality among patients \18 was not assessed.

CONCLUSION
The incidence of primary cutaneous melanoma in Olmsted County, Minnesota has increased steadily over the last 50 years, with a continued overall increase of 1.21-fold over the past 10 years.The highest increase over the past decade is among females aged 40 and older, with a relative stabilization among males.This divergence in sex incidence may indicate other external factors are contributing beyond overdiagnosis.Death due to melanoma has declined steadily over the past 50 years, with patient characteristics of age at diagnosis, nodular histogenic subtype, Pathologic stage, left anatomic site, and Breslow thickness most strongly associated with mortality.Although this study observes a confined subset of the US population, the granular, highquality nature of the data with reliable long-term follow-up is important in determining trends in incidence and melanoma.

d
High-quality population-based epidemiologic data with long-term follow-up validates rising melanoma incidence and decreasing mortality.Trend divergence among sex and age cohorts indicates external influence beyond overdiagnosis.

Fig 2 ,
A depicts disease-specific mortality by decade of diagnosis.Death due to melanoma decreased with each progressive decade and a more recent diagnosis was associated with a lower disease-specific death due to melanoma (HR 0.73 per

Fig 2 .
Fig 2. Disease-specific mortality of patients with primary cutaneous melanoma in Olmsted County, Minnesota.A, Overall mortality.B, Mortality within 10 years of diagnosis.

Fig 3 .
Fig 3. Hazard ratio for risk of death due to melanoma according to calendar year of diagnosis.

Fig 4 .
Fig 4. Projected disease-specific mortality following diagnosis of primary cutaneous melanoma.A, Overall mortality.B, Mortality according to Breslow thickness.C, Mortality according to sex.

Fig 5 .
Fig 5. Variable importance plot depicting the relative importance of multiple variables according to disease-specific mortality of melanoma.

Table I .
Patient demographics and clinicopathologic characteristics of 2310 primary cutaneous melanomas in Olmsted County, Minnesota 1, B-D), adults aged 40 to 60 and 611 continue to account for

Table II .
Univariable and multivariable models for disease-specific mortality of melanoma AJCC, American Joint Committee on Cancer; HR, hazard ratio; LM, lentigo maligna; MM, malignant melanoma; SD, standard deviation.