WHAT IS ALREADY KNOWN ON THIS TOPIC

The incidence of cutaneous melanoma has risen significantly in the USA over the past 40 years without an equivalent increase in mortality, epidemiological features concerning for overdiagnosis. How much overdiagnosis is attributable to melanoma in situ versus invasive melanoma is unknown.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICYIntroduction

The incidence of cutaneous melanoma has increased dramatically in the USA rising from 8.3 per 100 000 in 1975 to 49.1 per 100 000 in 2016.1 Despite this increase in incidence, mortality has remained remarkably flat over this period. This epidemiological pattern is concerning for melanoma overdiagnosis—the diagnosis of cancers that meet the pathological definition for cancer but are not destined to cause harm—even if left untreated.2 Overdiagnosis and consequent overtreatment is an important harm from cancer screening and other early detection activities, as described with breast, thyroid, prostate and other cancers.3 4

There is concern that overdiagnosis is a growing problem in melanoma, a cancer in which population-based screening has been promoted,1 5 and opportunistic screening is common in countries like Australia and the USA.6 7 Quantifying the frequency of overdiagnosis is important for policy makers, healthcare providers and patients to make judgements about the potential harms and benefits of screening. This is an issue that has been reviewed by the US Preventive Services Task Force who concluded that there was insufficient evidence for recommending melanoma screening, citing in their evidence review sparse information on harms in both the 2016 and 2023 recommendations.8–10 A Cochrane review reached similar conclusions.11

Overdiagnosis of an individual is only rarely determined, usually in cases where a cancer diagnosis is made but the individual remains untreated long term.12 More commonly it is indirectly measured at a population level, through comparison of epidemiological markers in groups exposed and unexposed to screening and diagnostic activities. The ideal method to make estimates of cancer overdiagnosis is using data from randomised controlled trials (RCTs) of screening with long-term follow-up.13 14 For melanoma there are no such RCTs.8 11 Other methods for overdiagnosis quantification include: modelling the natural history of disease, pathological studies (eg, epidemiological necropsy) and observational studies (ecological and cohort).13 Each of these methods have strengths and weaknesses.

A previous cohort study in a US population estimated melanoma overdiagnosis in white Americans by using black Americans (who develop melanoma but are not routinely screened) as an external standard to make judgements about melanoma trends. It was estimated that 60% of melanoma in white American men and 59% of melanoma in white American women were overdiagnosed in 2014.15 However, these findings did not distinguish between invasive and in situ (more likely overdiagnosed) melanoma. Furthermore, the study was limited by assumptions made using black Americans as an external standard. Moreover, it is only one method to estimate overdiagnosis where multiple approaches are necessary to triangulate the approximate true extent of the problem.2 13 A novel method of analysing observational data for evidence of overdiagnosis uses current and past lifetime risk of melanoma diagnosis. This method has been used to estimate cancer overdiagnosis in Australia for breast, melanoma, thyroid, renal and prostate cancers.16 17 In this study we use this method to estimate the proportion of melanoma (both invasive and in situ) likely overdiagnosed in the USA.

MethodsOverview

Overdiagnosis can be conceived as the amount of cancer detected in a population minus the amount of true clinically relevant cancer in that population. While the amount of cancer that is detected in a population is a known quantity captured by cancer registries, the amount of clinically relevant cancer is not knowable. Despite this challenge, estimates of overdiagnosis can be inferred using population statistics.18 Here, we aimed to quantify melanoma overdiagnosis in the USA by comparing current and past lifetime risk of being diagnosed.17 An underlying assumption in our analysis is that there was no overdiagnosis in 1975. This is a conservative assumption that would likely underestimate overdiagnosis in future years. We restricted our analysis to white Americans because they have the highest rate of developing and dying from melanoma.19 This population is also more likely to undergo cutaneous melanoma screening by a clinician, making them susceptible to overdiagnosis.20

We followed the Strengthening the Reporting of Observational Studies in Epidemiology reporting guideline.

Data

In this ecological study, we collected incidence data for cutaneous melanoma (in situ and invasive) from the National Cancer Institute’s Surveillance, Epidemiology and End Results (SEER) 9 registry database between 1975 and 2018.21 Data related to mortality was from the National Vital Statistics System, an established database providing cause of death information on all deaths in the USA.22 Specifically, we collected both melanoma specific and all-cause mortality for each year between 1975 and 2018. We extracted crude annual incidence, mortality and population counts stratified by age group of 5-year intervals (eg, 0, 5, 10 … ≥85). Our cohort was restricted to the first matching record for each patient to avoid multiple entries for patients with multiple primary melanomas. For each case of melanoma, we gathered year of diagnosis, age at diagnosis, sex and tumour staging (this was a binary variable: in situ or invasive). Because we aimed to estimate excess lifetime risk for each year during the study period, we did not use the default setting for Devcan (which includes data grouped in 3-year increments), instead we manually entered the crude annual data collected from SEER into Devcan.

To make assessments and estimates about the diagnosis of melanoma in the entire US population in 2018 we collected data from the US Cancer Statistics Public Use Databases (https://www.cdc.gov/cancer/uscs/public-use/index.htm).23

Analysis

Estimates of lifetime risks were made using the Devcan V.6.7.7 software from the National Cancer Institute (https://surveillance.cancer.gov/devcan). This software programme uses cross-sectional incident and mortality counts and converts them into probabilities of developing or dying from cancer within a hypothetical population. Using the software, we generated lifetables for each year between 1975 and 2018. We compared the past lifetime risk and current lifetime risk of a melanoma diagnosis for each year. Because risk of death (all-cause) has changed over time, we adjusted for competing risk of death by applying the mortality rates for each 5-year age group from 2018 to all other calendar years (moving lifetime risk estimates upwards for a melanoma diagnosis in years earlier than 2018 upwards). We also adjusted for changes in risk factors for clinically relevant (‘true’) melanoma over time by applying the proportional increase in melanoma mortality between 1975 and 2014 (new treatments led to decreases in melanoma mortality after 2015), using the following formulas:

For men: adjustment=1+number years since 1975×0.061402*

For women: adjustment=1+number years since 1975×0.021748*

*These are the yearly proportional increases in melanoma mortality for white men and women.

Hence, for example, to make the 1975 lifetime risk for men comparable to the 2018 lifetime risk we applied the mortality rates for each 5-year age group from 2018 to adjust for longevity and multiplied the ensuing estimate by 3.64 (1+43×0.061402) to adjust for changes in risk factors. The difference in lifetime risk of developing melanoma between 1975 and 2018, after adjusting for changes in mortality trends and risk factors, was interpreted as probable melanoma overdiagnosis. We calculated the proportion of melanomas (all and invasive) that were overdiagnosed using the following formula:

To obtain the total number of people in the USA likely overdiagnosed with melanoma in 2018, we multiplied the numbers of melanomas (all and invasive) reported in the US Cancer Statistics Public Use Data for 2018 by the corresponding estimated overdiagnosis proportion.

The very low probability of a melanoma in situ diagnosis in 1975 means that using the excess lifetime risk to calculate proportion overdiagnosed in 2018 results in uncertain estimates. Instead, we calculated the total number of overdiagnosed melanoma in situ in 2018 by subtracting the total number of overdiagnosed invasive melanomas from the total number of overdiagnosed all melanomas. We calculated the proportion of overdiagnosed melanoma in situ in 2018 as:

Patient and public involvement

Patients were not involved in the development or conduct of this study. Informed consent was not required for this study as it was a secondary analysis of data.

Results

Between 1975 and 2018 the lifetime risk of being diagnosed with melanoma (invasive and in situ, adjusted for changes in longevity and risk factors) increased from 3.2% (1 in 31) to 6.4% (1 in 16) among white men, and from 1.6% (1 in 63) to 4.5% (1 in 22) among white women. We estimate that ((6.4–3.2)/6.4)=49.7% of all melanomas diagnosed in 2018 represent overdiagnosis in white men and ((4.5–1.6/4.5)=64.6% in white women. This represents 43 709 men and 38 943 women overdiagnosed with melanoma in the USA in 2018, of which 7848 men and 11 050 women were overdiagnosed with invasive melanoma (table 1).

Table 1

Estimated number of cutaneous melanomas overdiagnosed in US males and females, 2018

Over the same period, melanoma in situ increased from 0.17% (1 in 588) to 2.7% (1 in 37) in white men and 0.08% (1 in 1250) to 2.0% (1 in 50) in white women. We estimate that 89.4% of melanoma in situ diagnosed in white men and 85.4% in white women represent overdiagnosis in 2018. This represents 35 961 men and 27 893 women overdiagnosed with melanoma in situ in 2018 (table 1).

The adjusted lifetime risk of a melanoma diagnosis and the estimated adjusted excess lifetime risk of a melanoma diagnosis are shown in figure 1 for the years 1976–2018. The estimated excess lifetime risk of a melanoma diagnosis in 2018 was 3.2% among white men (figure 1A) and 2.9% among white women (figure 1B) (corresponding to the numerators in the calculations of proportions overdiagnosed that are presented above). For other calendar years’ lifetime risk estimates (adjusted and unadjusted) see online supplemental table 1. For other calendar years’ estimated excess lifetime risk of melanoma see online supplemental table 2 (all melanoma), online supplemental table 3 (invasive melanoma) and online supplemental table 4 (melanoma in situ). Much of the excess risk across all calendar years is from melanoma in situ which had excess lifetime risk of 2.6% among white men and 1.9% of white women in 2018. This can be appreciated in the curves in online supplemental figures 1 and 2. For melanoma in situ (online supplemental figure 2) the lifetime risk and excess risk curves are close to one another signifying much of the increase is likely attributable to overdiagnosis whereas the curves are further apart for invasive melanoma (online supplemental figure 1) signifying less of a contribution of overdiagnosis.

Figure 1

(A) Lifetime risk and estimated excess lifetime risk of melanoma diagnosis in white males in the USA from 1976 to 2018, adjusted for changes in longevity and risk factors. (B) Lifetime risk and estimated excess lifetime risk of melanoma diagnosis in white females in the USA from 1976 to 2018, adjusted for changes in longevity and risk factors.

Discussion

We used a life table method approach to estimate the proportion of melanoma (both in situ and invasive) likely representing overdiagnosis in the USA. We found that 49.7% of melanomas in white men and 64.6% of melanomas in white women may have been overdiagnosed in 2018. These overdiagnosis estimates represent approximately 44 000 melanomas in white men and 39 000 in white women. Melanoma in situ represents most of the overdiagnosed cancers as we estimate that 89% of melanoma in situ are overdiagnosed in white men and 85% in white women. These estimates further contribute to the growing body of evidence showing the significant and growing number of patients that are overdiagnosed with melanomas.1 24–26 A recently published scoping review on overdiagnosis in melanoma including 35 studies found that melanoma incidence has been increasing in Europe, the USA and Australia without commensurate increases in mortality, an epidemiological signature concerning for overdiagnosis.27 The study found that much of the increase was among screen detected thin and in situ melanomas, which are the most likely to be overdiagnosed.

Our estimates of the proportion of overdiagnosed melanoma are similar to estimates from a previous Australian study which showed that 58% of melanoma in men and 54% in women were overdiagnosed in 2012.17 As with the current study, authors of an Australian study concluded that most of the overdiagnosed melanomas in Australia were melanoma in situ—a cancer whose incidence is rapidly growing and now diagnosed two times as often as invasive melanoma in Australia.28 In a US-based study using Black Americans as an external standard, the proportion of melanomas overdiagnosed was estimated at 59% in white women and 60% in white men, although no distinction was made between contributions of in situ versus invasive melanoma.15 Moreover, in a study that also uses SEER to assess trends in overdiagnosis during a similar time period (1975–2017), Kurtansky et al similarly concluded that melanoma overdiagnosis appears to be greatest in white women and for melanoma in situ.25 While this study did not quantify the extent of melanoma overdiagnosis it points to a population in which the phenomenon may be greater. Taken together, our estimates using the life table method are consistent with previous reports, emphasising the substantial scope of the problem of melanoma overdiagnosis in the USA.

One difference between the estimates in Australia and USA, is the overdiagnosis rate for invasive melanoma in women. The primary analysis for the Australian data adjusted for changes in risk factors by applying the proportional increases in rates of thick melanomas, but used proportional increases in melanoma mortality (as was done in the current US-based study) in a sensitivity analysis. While this resulted in an estimated 22% overdiagnosis of invasive melanoma in Australian women (=0.010 excess/0.0460), the figure in the current study was 40% for US white women (=0.010 excess/0.0250). The reason for this difference between Australia and the USA seems to be that while there is the same excess risk in absolute terms (0.010 or 1% lifetime risk), the underlying risk of clinically important invasive melanoma is much higher in Australia (0.0460 or 4.6% lifetime risk) than USA (0.025 or 2.5% lifetime risk). The higher melanoma overdiagnosis risk for women than for men observed in the USA data has been observed by others, and gender differences have also been observed for thyroid cancer.25 29 The reasons for this gender difference are likely complex, but may include differences in healthcare encounter frequency (more opportunities for diagnosis) as well as underlying risk of a lethal cancer (fewer diagnoses of clinically important cancer).

One of the novel contributions of our study was our estimates of the proportion of melanoma in situ overdiagnosed in the USA. Our finding that most melanoma in situ is overdiagnosed may seem surprising to some but given the epidemiology of this cancer this is not unexpected. Melanoma in situ is considered the earliest form of melanoma, stage 0, or at least a precursor to invasive melanoma. However, while the incidence of melanoma in situ has risen dramatically in the USA over the past 40 years, removal of these lesions appears to have had minimal appreciable effect on the rate of invasive melanoma (or death from melanoma) over the same period. This suggests that much of melanoma in situ may not represent true ‘cancers’, and that melanoma in situ may not be an obligate precursor to invasive melanoma.1 28 30 31 Furthermore, patients diagnosed with melanoma in situ have a higher relative survival to age/sex/race matched individuals without melanoma in situ.32 These epidemiological realities present a possible opportunity to study de-escalation of care for patients with melanoma in situ.33 For example, RCTs could establish the safety of clinical monitoring of borderline lesions rather than immediate surgical removal, and using narrower margins for the surgical excisions of melanoma in situ. National organisations such as the National Comprehensive Cancer Network and American Academy of Dermatology might also revisit their guidelines that suggest all patients with melanoma in situ be seen screened frequently after diagnosis, a practice that is not promoted in other parts of the world.34–36

The reasons for melanoma overdiagnosis are multifactorial. Rates of opportunistic melanoma screening skin exams have continued to increase in the USA, particularly among people that identify as white Americans.1 Concomitant with the rise in skin exams has been an increase in biopsy rate which has been linked to higher detection of early melanoma.37 The thresholds for pathologists to label histopathological slides of pigmented lesions as melanoma has also fallen over time, further contributing to the increasing rate of melanoma diagnosis.38 39 The combination of these factors point to the conclusion that most of the increased detection of melanoma in the USA is likely of overdiagnosed indolent disease. Patients who are overdiagnosed can only be harmed—they cannot benefit, as there is nothing that needs to be fixed. These harms are physical, psychological and economic.40 Physical harms can include overtreatment, repeat skin biopsies, scarring, pain, infection and/or functional impairment. Psychological harms can include labelling as a patient with cancer, anxiety about being outdoors or about cancer recurrence, or guilt for past UV exposure causing melanoma.41 In the USA, economic harms can include treatment cost for the immediate diagnosis as well as costs for future surveillance and possible denial of life insurance as a cancer survivor.

Limitations

Our study has several important limitations. Both cohort and ecological studies estimating overdiagnosis have a high risk of confounding and selection bias, particularly when using a reference population (eg, a population from the 1970s); however, our adjustments accounting for changes in underlying risk factors is an important strength decreasing the probability of such biases in our estimates.13 Our study also assumes no overdiagnosis in 1975 which may not be the case, underestimating overdiagnosis. However, at the beginning of the study period screening for skin cancer was much less a focus of clinicians than today. Finally, the ideal study to estimate overdiagnosis would be a follow-up to a large, randomised control trial of melanoma screening, however, such data are not available. Until such a trial exists then we must rely on multiple observational studies to triangulate the extent of melanoma overdiagnosis.

Conclusion

Melanoma overdiagnosis among white Americans is significant and increasing over time with over 80 000 melanomas estimated as overdiagnosed in 2018. A large proportion of overdiagnosed melanomas are in situ cancers, pointing to a potential area to focus for an intervention de-escalation of the intensity of treatment and survivorship care.

Data availability statement

Data are available in a public, open access repository. All data used is publicly available and de-identified.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

This research project was considered by The University of Texas at Austin Institutional Review Board as non-human participant research, and no ethics approval was required.