This quality improvement study compares thickness-specific incidence of melanoma in screened vs unscreened patients following the initiation of a primary care–based skin cancer screening initiative.
Key Points
Question
Is primary care–based melanoma screening associated with melanoma incidence?
Findings
In this quality improvement study among 595 799 patients, primary care–based melanoma screening was associated with increased detection of thin melanoma (in situ and thickness ≤1 mm). While decreased incidence of thick melanoma (>2 and >4 mm) was observed in screened vs unscreened patients, this difference was not statistically significant.
Meaning
Population-based screening for melanoma was associated with increased detection of thin melanoma, raising concerns that screening carries the potential harm of overdiagnosis of indolent lesions; longer follow-up is needed to determine if this influences outcomes such as incidence of thick melanoma, distant metastases, or mortality.
Abstract
Importance
Population-based skin cancer screening is currently not recommended owing to lack of data to quantify the balance of benefits and harms.
Objective
To compare thickness-specific incidence of melanoma in screened vs unscreened patients following the initiation of a primary care–based skin cancer screening initiative.
Design, Setting, and Participants
This observational study of a quality improvement initiative was conducted from January 1, 2014, through December 31, 2018, among patients 35 years and older presenting for a primary care visit at primary care practices within an academic and community-based health care system during the study period. Data analysis was performed January 2020 to January 2022.
Interventions
Primary care clinicians were offered training in melanoma identification through skin examination and encouraged to offer annual screening to patients 35 years and older.
Main Outcomes and Measures
Thickness of melanomas diagnosed in screened and unscreened patients.
Results
Among 595 799 analyzed screen-eligible patients, 144 851 (24.3%) were screened at least once. Screened patients were older (median [IQR] age, 59 [49-67] vs 55 [45-66] years) and more likely to be female (82 244 [56.8%] vs 250 806 [55.6%]; P < .001) and non-Hispanic White (124 747 [86.1%] vs 375 890 [83.4%]; P < .001) than unscreened patients. After adjusting for age, sex, and race, screened patients were more likely than unscreened patients to be diagnosed with in situ (incidence, 30.4 vs 14.4; hazard ratio [HR], 2.6; 95% CI, 2.1-3.1; P < .001) or thin invasive (≤1 mm) melanoma (incidence, 24.5 vs 16.1; HR, 1.8; 95% CI, 1.5-2.2; P < .001). Screened patients were also more likely than unscreened patients to be diagnosed with in situ (incidence, 26.7 vs 12.9; HR, 2.1; 95% CI, 1.7-2.6; P < .001) or thin invasive (≤1 mm) interval melanomas (melanoma diagnosed at least 60 days after initial screening examination) (incidence, 18.5 vs 14.4; HR, 1.3; 95% CI, 1.0-1.7; P = .03). Incidence of melanoma thicker than 4 mm in unscreened and screened patients, respectively, was 3.3 and 2.7 (HR, 0.8; 95% CI, 0.4-1.4; P = .38) for all melanomas and 2.7 and 1.5 (HR, 0.6; 95% CI, 0.2-1.2; P = .15) for interval melanomas.
Conclusions and Relevance
In this quality improvement study, primary care–based melanoma screening was associated with increased detection of thin melanoma, raising concern about overdiagnosis. Further studies with longer follow-up are needed to determine the influence of screening on the incidence of thick melanoma and outcomes associated with high costs and poor outcomes, such as metastasis.
Introduction
Cutaneous melanoma was the fifth most common cancer in the US in 2021.1 Melanoma can be detected with a simple naked-eye examination, and stage at diagnosis is the best predictor of prognosis, so systematic skin cancer screening may reduce melanoma mortality. However, to our knowledge, no randomized clinical trials of melanoma screening have been performed. The US Preventive Services Task Force currently gives visual skin cancer screening of asymptomatic patients an “I” rating owing to insufficient evidence to assess the balance of benefits and harms of skin cancer screening.2 There is concern that melanoma screening may lead to overdiagnosis, or an increased detection of thin tumors that would not have been detected otherwise and that are unlikely to progress to a harmful cancer.3
In 2014, the University of Pittsburgh Medical Center (UPMC), a large health care system with both community-based and academic practices, began a primary care skin cancer screening quality initiative that has been previously described.4 Briefly, primary care clinicians employed by UPMC were invited to participate in a voluntary program in which patients 35 years and older who presented for routine office visits were offered full skin examination annually (guidance based on the screening guidelines in place in Germany at the time).5 Optional training in skin cancer detection was available using a validated training program via an online module.6 During the first year of the initiative, we found an increase in melanoma detection, particularly for in situ melanoma, and a decrease in the median thickness of invasive melanoma, but no difference in the incidence of melanoma thicker than 1 mm, in patients who were screened compared with those who were not screened.4 We now report findings of melanoma incidence and thickness from the first 5 years of the UPMC primary care screening program.
Methods
Approvals
The screening initiative, including clinician outreach and training and system changes to the electronic health record (EHR), was approved as a quality improvement project. Collection of outcome data about screening dispositions, patient demographics, and melanoma diagnoses was approved by the University of Pittsburgh Institutional Review Board, which waived the need for patient informed consent owing to determination that study presented minimal risk to participants and could not be practically carried out without a waiver of consent.
Program Overview
Primary care clinicians employed by UPMC were invited and encouraged, but not required, to complete a web-based, validated, skin cancer training (a previously validated system called INFORMED [INternet curriculum FOR Melanoma Early Detection])6 adapted for use in this context. This program trains primary care clinicians on how to perform a full body skin examination and to recognize melanoma, keratinocyte carcinomas, and commonly encountered benign lesions, such as seborrheic keratosis and cherry angiomas. Training rates were only tracked during the first year of the initiative and were previously reported.7 The EHR (EPIC) was modified to provide an alert when a patient 35 years or older did not have a documented skin cancer screening in the prior 12 months and to allow the clinician to note with a single click if screening was performed or was already completed. Clinicians were instructed to note screening to be completed if they performed the screening or if the patient reported they had been screened in the previous year by another clinician, but no formal documentation of the examination was required. Physicians were encouraged by physician leadership to participate and were educated about melanoma and this pilot program through a series of town hall meetings at the beginning of the initiative. Physicians were not compensated for screening above their standard billing for the office visit.4 Some offices posted information about the program in their office, and some screens could have been patient initiated. Screening could be performed by any physician, nurse practitioner, or physician assistant in the UPMC system. After detection of a lesion that was concerning for skin cancer or determination that a patient was at high risk for cancer, primary care clinicians could perform a biopsy or refer the patient to a dermatologist of their choosing for further evaluation through routine standard-of-care practice. Only patients in the screen-eligible population (defined below) were included in this analysis.
Screen-Eligible Population and Screening Status Case Definitions
Screen-eligible patients were defined as those 35 years and older who presented to a UPMC-employed primary care clinician for a primary care visit (defined as a visit to a primary care office in which a Current Procedural Terminology code for an office visit or annual physical examination was coded) between January 1, 2014, and December 31, 2018. Patients who were diagnosed with melanoma prior to their first primary care clinician visit in the study period or had a screening date that occurred during the study period but prior to their first primary care visit in the study period were excluded from analysis (Figure 1). Primary care clinicians were requested to ask about and offer skin cancer screening to eligible patients during office visits. Screening was documented by clicking a single button in the EHR or by using the International Classification of Diseases, Ninth Revision code V76.43 or International Statistical Classification of Diseases and Related Health Problems, Tenth Revision code Z12.83 (encounter for screening for malignant neoplasms of the skin). This first date on which this was noted was captured as a screening date.
Figure 1. Study Flow Diagram.
aOne patient was excluded because biopsy showed fragment of melanoma, and no information on thickness was available.
UPMC indicates University of Pittsburgh Medical Center.
Each patient in the screen-eligible cohort was assigned a screening status (screened or unscreened) and a primary visit date. For patients who were not screened during the study period, or who were diagnosed with melanoma prior to their first screening, the date of their first primary care visit during the study period was used as their entry date into the unscreened cohort. For patients with at least 1 screening (provided screening occurred prior to or on the day of their first melanoma diagnosis in the study period if applicable), the date their first screening was noted in the EHR was considered their primary screening visit date and was used as their date of entry into the screened cohort. For each screen-eligible patient, in addition to screening status and date, the following patient-level data were recorded: age, sex, and self-identified race and ethnicity.
Melanoma Diagnoses
Text-based searching of pathology reports for all screen-eligible patients available in the UPMC system was used to identify primary melanoma diagnoses and date of diagnosis. Reports were reviewed to exclude re-excisions and to confirm that only primary lesions were included. Breslow thickness was recorded for each case, using information from the initial pathology report and, when applicable, from pathology reports from re-excisions. To better understand the impact of screening on incident melanomas, we attempted to exclude prevalent melanomas (ie, those present at the time the patient entered the screen-eligible cohort). To do this, we categorized melanomas diagnosed at least 60 days after the first screening visit in screened patients, and at least 60 days after the first screen-eligible primary care visit in unscreened patients, as interval melanomas. For each eligible patient, only the first melanoma diagnosis during the study period was considered in the analysis, although the number of patients with multiple primary melanomas was collected.
Statistical Analysis
This study was reported using Standards for Quality Improvement Reporting Excellence (SQUIRE) reporting guidelines, which provide a framework for reporting new knowledge about system-level work to improve the quality, safety, and value of health care (eAppendix in the Supplement).8 The primary outcome of interests was the cumulative incidence of melanoma by T category in screened vs unscreened patients. Secondary analyses were performed for all melanomas and for interval melanomas and for all patients and for patients 65 years or older (chosen owing to older age as a risk factor for melanoma incidence and mortality and owing to the potential implications for guidelines relevant to a US Medicare population). Descriptive statistics were calculated for patient demographic and screening characteristics, and χ2 tests were used to compare rates of melanoma diagnosis by thickness in screened and unscreened patients. P values were 1-tailed with a significance level of P < .05. Age- and sex-adjusted cumulative incidence of melanoma diagnosis by thickness was calculated for screened and unscreened patients. Cox proportional hazards and frailty models were used to estimate hazard ratios (HRs) for screening on risk of melanoma diagnosis by thickness, adjusting for age, sex, and White race in the primary analysis as well as clinician practice in sensitivity analysis. Data analysis was performed January 2020 to January 2022 using R, version 4.0.5 (R Foundation for Statistical Computing) and SAS, version 9.4 (SAS Institute) software.
Results
Screened and Screen-Eligible Population
A total of 620 371 screen-eligible patients were identified. Patients were excluded from analysis if they were diagnosed with melanoma prior to their first primary care visit in the study period (n = 248) or if they had a screening date prior to their first primary care visit in the study period (n = 24 324), leaving 595 799 patients eligible for analysis. In this cohort, 144 851 (24.3%) were screened at least once, and 450 948 (75.7%) were unscreened (Figure 1). Compared with unscreened patients, those who were screened were slightly older (median [IQR] age, 59 [49-67] vs 55 [45-66] years), more likely to be female (82 244 [56.8%] vs 250 806 [55.6%]; P < .001), and more likely to have a self-identified race and ethnicity of non-Hispanic White (124 747 [86.1%] vs 375 890 [83.4%]; P < .001). Patients 65 years and older comprised 32.9% of the screened and 28.4% of the unscreened cohort (47 603 and 127 777 patients, respectively). Most patients entered the study cohort in 2014 (Table 1). A minority (18.2%) of screens were documented by dermatologists in the UPMC system.
Table 1. Demographic Characteristics of Screened and Unscreened Populations.
| Characteristic | No. (%) | P value | |
|---|---|---|---|
| Unscreened (n = 450 948 [75.7]) | Screened (n = 144 851 [24.3]) | ||
| Age (at first visit for unscreened or first screen for screened), y | |||
| Median (IQR) | 55 (45-66) | 59 (49-67) | NA |
| Mean (SD) | 56.6 (14.7) | 58.5 (12.9) | <.001a |
| ≥65 y | 127 777 (28.3) | 47 603 (32.9) | <.001b |
| Sex | |||
| Female | 250 806 (55.6) | 82 244 (56.8) | <.001b |
| Male | 200 142 (44.4) | 62 607 (43.2) | |
| Race and ethnicity | |||
| American Indian/Alaska Native | 592 (0.1) | 150 (0.1) | <.001b |
| Asian | 6686 (1.5) | 2124 (1.5) | |
| Black | 37 637 (8.3) | 9638 (6.7) | |
| Hispanic White | 2397 (0.5) | 688 (0.5) | |
| Non-Hispanic White | 375 890 (83.4) | 124 747 (86.1) | |
| No race reported | 9818 (2.2) | 1901 (1.3) | |
| Pacific Islander | 205 (0.05) | 36 (0.03) | |
| White, ethnicity unknown | 17 723 (3.9) | 5567 (3.8) | |
| Year first eligible (primary care clinician visit) | |||
| 2014 | 232 426 (51.5) | 103 869 (71.7) | <.001b |
| 2015 | 68 723 (15.2) | 17 401 (12.0) | |
| 2016 | 53 136 (11.8) | 10 890 (7.5) | |
| 2017 | 51 129 (11.3) | 8115 (5.6) | |
| 2018 | 45 534 (10.1) | 4576 (3.2) | |
| Calendar years with a screening visit | |||
| Median (IQR) | 0 (0-0) | 1 (1-2) | <.001c |
| First (within study period) melanoma thickness | |||
| In situ | 221 (34.6) | 172 (48.3) | <.001b |
| Thin invasive, ≤1 mm | 238 (37.3) | 132 (37.1) | |
| 1.01-2 mm | 78 (12.2) | 24 (6.7) | |
| 2.01-4 mm | 50 (7.8) | 16 (4.5) | |
| > 4 mm | 51 (8.0) | 12 (3.4) | |
| No. melanomas per patient | |||
| No melanoma | 450 310 | 144 494 | .26b |
| Single primary | 614 (96.2) | 348 (97.8) | |
| Multiple primary melanomas | 24 (3.8) | 8 (2.2) | |
| Mean person-years of follow up | 3.30 | 2.91 | <.001c |
Abbreviation: NA, not applicable.
Wilcoxon rank sum.
χ2.
t Test.
Melanoma Diagnoses in Screened and Unscreened Patients
In total, 994 patients were diagnosed with melanoma for which thickness could be determined during the study period (356 screened patients, 638 unscreened patients; 986 of 994 [99.2%] melanomas were diagnosed in patients who self-identified their race and ethnicity as non-Hispanic White). One patient in the screened population had a biopsy that showed fragments of melanoma with no further thickness information and was excluded from analysis. In the screened population, 110 melanomas were diagnosed within the 60 days following the first screen date. In the unscreened population, 73 melanomas were diagnosed on or within the 60 days following the first eligible primary care visit. The remaining melanomas (246 in the screened population, 565 in the unscreened population) were subsequently diagnosed during the observation period and were considered interval melanomas (Figure 1).
The cumulative incidences of melanoma in situ and invasive melanoma by thickness (≤1 mm, >1 mm, >2 mm, and >4 mm thick) are shown for all patients (Figure 2) and for those 65 years and older (eFigure in the Supplement). Melanoma detected at a first screen-eligible visit would have been present at the time the patient entered the study period (ie, would be considered melanoma prevalence, not incidence). To compare the association of a negative screen with the thickness of melanomas that later develop (referred to as interval melanomas), we also compared melanoma incidence by thickness in screened and unscreened patients, excluding those patients who had a melanoma diagnosed within 60 days of their first screen-eligible visit. The 60-day window was chosen as a reasonable period during which a patient would have been expected to be referred to a specialist for definitive diagnosis (eTable in the Supplement). After adjusting for age, sex, and White race, melanomas in screened patients were significantly more likely than those in unscreened patients to be in situ (incidence, 30.4 vs 14.4; HR, 2.6; 95% CI, 2.1-3.1; P < .001), thin invasive melanoma (≤1 mm) (incidence, 24.5 vs 16.1; HR, 1.8; 95% CI, 1.5-2.2; P < .001) (Table 2). We also found that screened patients were more likely than unscreened patients to be diagnosed with interval melanomas that were in situ (incidence, 26.7 vs 12.9; HR, 2.1; 95% CI, 1.7-2.6; P < .001) or thin invasive lesions (<1 mm) (incidence, 18.5 vs 14.4; HR, 1.3; 95% CI, 1.0-1.7; P = .03) (Table 3). While thin melanomas were more frequently diagnosed in screened than unscreened patients, we observed a trend toward a reduction in thicker melanomas (all and interval melanomas) among screened patients with HRs that were less than 1 for melanomas thicker than 2 and 4 mm, particularly in those 65 years and older (Figure 2 and Tables 2 and 3). However, these findings did not reach statistical significance.
Figure 2. Cumulative Incidence of Melanoma by Breslow Thickness and Cohort.
Cumulative incidence by Breslow thickness of melanoma in entire screen-eligible population by screening status for (A) all melanomas, (B) in situ melanoma, (C) invasive melanoma 1 mm or thinner, (D) melanoma greater than 1 mm, (E) melanoma greater than 2 mm, and (F) melanoma greater than 4 mm. Shaded areas indicate 95% CI.
Table 2. Total Melanoma Cases in Screened and Unscreened Patients for Entire Screen-Eligible Cohort and for Patients 65 Years and Older.
| Group | Melanomas diagnosed (% of population diagnosed with melanoma) | P value (χ2) | Age-sex adjusted incidence [per 100 000 person-years] (95% CI) | HR (95% CI) and P value adjusting for age, sex, and White race, Cox proportional hazard model | |||
|---|---|---|---|---|---|---|---|
| Unscreened | Screened | Unscreened | Screened | HR (95% CI) | P value | ||
| All melanomas | |||||||
| All | 638 (0.14) | 356 (0.25) | <.001 | 42.2 (39.0-45.6) | 65.3 (57.9-73.7) | 1.8 (1.6-2.1) | <.001 |
| Age ≥65 y | 270 (0.21) | 155 (0.33) | <.001 | 66.5 (58.7-75.3) | 98.7 (83.2-117.2) | 1.6 (1.3-2.0) | <.001 |
| In situ | |||||||
| All | 221 (0.05) | 172 (0.1) | <.001 | 14.4 (12.6-16.5) | 30.4 (25.5-36.2) | 2.6 (2.1-3.1) | <.001 |
| Age ≥65 y | 104 (0.08) | 71 (0.1) | <.001 | 25.7 (21.0-31.3) | 43.4 (33.5-56.4) | 1.9 (1.4-2.6) | <.001 |
| ≤1 mm | |||||||
| All | 238 (0.05) | 132 (0.09) | <.001 | 16.1 (14.2-18.3) | 24.5 (20.1-29.9) | 1.8 (1.5-2.2) | <.001 |
| Age ≥65 y | 83 (0.07) | 56 (0.1) | <.001 | 22.3 (17.8-27.9) | 36.1 (27.2-47.9) | 1.9 (1.3-2.6) | <.001 |
| >1 mm | |||||||
| All | 179 (0.04) | 52 (0.04) | .57 | 11.6 (10.0-13.5) | 10.4 (7.8-14.0) | 1.0 (0.7-1.3) | .75 |
| Age ≥65 y | 83 (0.07) | 28 (0.06) | .73 | 18.5 (14.8-23.2) | 19.3 (13.1-28.3) | 1.0 (0.7-1.6) | .90 |
| >2 mm | |||||||
| All | 101 (0.02) | 28 (0.02) | .56 | 6.5 (5.3-7.9) | 6.0 (4.1-8.8) | 0.9 (0.6-1.4) | .61 |
| Age ≥65 y | 60 (0.05) | 16 (0.03) | .29 | 14.1 (10.9-18.3) | 12.0 (7.3-19.9) | 0.8 (0.5-1.4) | .42 |
| >4 mm | |||||||
| All | 51 (0.01) | 12 (0.008) | .41 | 3.3 (2.5-4.3) | 2.7 (1.5-4.8) | 0.8 (0.4-1.4) | .38 |
| Age ≥65 y | 33 (0.03) | 5 (0.01) | .08 | 7.6 (5.3-10.8) | 3.4 (1.4-8.4) | 0.5 (0.2-1.2) | .11 |
Abbreviation: HR, hazard ratio.
Table 3. Interval Melanomas in Screened and Unscreened Patients for Entire Screen-Eligible Cohort and for Patients 65 Years and Older.
| Group | Melanomas diagnosed (% of population diagnosed with melanoma) | P value (χ2) | Age-sex adjusted incidence [per 100 000 person-years] (95% CI) | HR (95% CI) and P value adjusting for age, sex, and White race, Cox proportional hazard model | |||
|---|---|---|---|---|---|---|---|
| Unscreened | Screened | Unscreened | Screened | HR (95% CI) | P value | ||
| All melanomas | |||||||
| All | 565 (0.13) | 246 (0.17) | <.001 | 37.6 (34.6-40.8) | 53.5 (46.9-60.9) | 1.5 (1.2-1.7) | <.001 |
| Age ≥65 y | 243 (0.19) | 113 (0.24) | .06 | 60.8 (53.3-69.3) | 82.1 (68.0-99.1) | 1.3 (1.1-1.7) | .009 |
| In situ | |||||||
| All | 196 (0.04) | 122 (0.08) | <.001 | 12.9 (11.2-14.8) | 26.7 (22.2-32.2) | 2.1 (1.7-2.6) | <.001 |
| Age ≥65 y | 94 (0.07) | 51 (0.1) | .04 | 23.3 (18.9-28.7) | 37.2 (28.1-49.2) | 1.6 (1.1-2.2) | .01 |
| ≤1 mm | |||||||
| All | 213 (0.05) | 85 (0.06) | .10 | 14.4 (12.6-16.5) | 18.5 (14.8-23.1) | 1.3 (1.0-1.7) | .03 |
| Age ≥65 y | 75 (0.06) | 39 (0.08) | .11 | 20.3 (16.0-25.7) | 28.7 (20.8-39.5) | 1.4 (1.0-2.1) | .06 |
| >1 mm | |||||||
| All | 156 (0.04) | 39 (0.03) | .19 | 10.3 (8.8-12.0) | 8.3 (6.0-11.5) | 0.8 (0.6-1.2) | .31 |
| Age ≥65 y | 74 (0.06) | 23 (0.05) | .52 | 17.2 (13.6-21.8) | 16.2 (10.7-24.6) | 1.0 (0.6-1.5) | .86 |
| >2 mm | |||||||
| All | 89 (0.02) | 20 (0.01) | .18 | 5.7 (4.6-7.0) | 4.2 (2.6-6.5) | 0.7 (0.5-1.2) | .23 |
| Age ≥65 y | 57 (0.05) | 12 (0.03) | .09 | 13.5 (10.3-17.7) | 9.0 (5.0-16.0) | 0.6 (0.3-1.2) | .15 |
| >4 mm | |||||||
| All | 42 (0.009) | 7 (0.005) | .14 | 2.7 (2.0-3.7) | 1.5 (0.7-3.3) | 0.6 (0.2-1.2) | .15 |
| Age ≥65 y | 30 (0.02) | 3 (0.006) | .03 | 7.0 (4.8-10.1) | 2.2 (0.7-7.2) | 0.3 (0.1-1.0) | .05 |
Abbreviation: HR, hazard ratio.
To adjust for variations in factors such as those related to patients’ geography (sociodemographic factors, access to health care) and to the primary care physicians caring for these patients, frailty models were used to estimate HRs for screening on risk of melanoma diagnosis by thickness adjusting for age, sex, and White race as well as clinician practice (eTable in the Supplement). This did not significantly change our findings.
Discussion
In this study of a primary care–based melanoma screening initiative, we observed a higher cumulative incidence of thin and in situ melanoma among screened vs unscreened patients. While early detection of melanoma is a strategy to reduce melanoma morbidity and mortality, the value of a cancer screening program should most rigorously be measured not by the number of new, early cancers detected, but by its impact on the development of late-stage disease and its associated morbidity, cost, and mortality. The significant increase in melanoma incidence in the US, particularly thin melanoma, without a concomitant decrease in melanoma mortality, raises the concern that early detection efforts, such as visual skin screening, may result in overdiagnosis, meaning the detection of indolent lesions that would not have progressed to fatal melanoma prior to being detected by routine care.3,9 Similarly, in Europe, while there are significant disparities in melanoma thickness by health care expenditure, mortality rates have less variability, suggesting that wealthier countries with more regular and established melanoma screening and awareness programs have higher detection rates of indolent melanoma.10
We did also observe a lower cumulative incidence of thick melanomas in screened vs unscreened patients, with the most notable decrease observed for interval melanomas among patients 65 years and older (HR, 0.6; P = .15 for lesions >2 mm; and HR, 0.3; P = .05 for lesions >4 mm). However, this was a secondary end point and did not meet a threshold for statistical significance of P < .05. Explanations of why screening could reduce the incidence of thicker melanoma include that thinner lesions with the biologic potential to progress to thick melanoma may be diagnosed and removed in the screened population, preventing them from becoming thick melanomas. Alternatively, patients who are screened may be more knowledgeable about melanoma (either at baseline or through knowledge gained during the screening process) and thus more likely to identify and seek attention for melanoma before it reached a thickness of 4 mm.11 Few patients in our study were followed up for at least 5 years, and it is possible that, particularly for slower-growing superficial spreading melanomas, longer follow-up time is needed to fully appreciate any differences in the incidence of thick melanomas between the screened and unscreened populations. Additionally, a larger sample size would likely be needed to see a difference between these groups given the relatively low number of thick melanomas in population studied.
While to our knowledge there are no large, randomized studies of melanoma screening, other large observational studies have been reported. The SCREEN (Skin Cancer Research to provide Evidence for Effectiveness in Northern Germany) project conducted in Schleswig-Holstein Germany was a 1-year intervention in which all patients 20 years and older holding statutory health insurance were invited to receive a single skin cancer screening from a trained physician, most of whom were generalists. In the 5 years following this intervention, there was an initial apparent reduction in melanoma mortality of nearly 50%. However, this was not sustained on follow-up in Schleswig-Holstein, nor was the same outcome observed when skin cancer screening was offered on a biannual basis to all citizens 35 years and older in the rest of Germany.5,12,13 Other criticisms of the SCREEN project include that a 50% reduction in melanoma deaths is unlikely due to a screening intervention in 19% of the population and that the reduction in melanoma deaths in the screened region was accompanied by an increase in death from undefined malignant neoplasm, suggesting that misclassification rather than reduction in melanoma death could explain the initial findings.14 A retrospective analysis of the Schleswig-Holstein data comparing observed vs expected rates of interval melanomas (defined as those diagnosed within 2 years after an initial negative screen) found an increase in the incidence of melanoma in situ and a decrease in invasive melanoma. Compared with our study, this study involved a shorter window of time, included fewer melanomas, and used as a comparator reference the historic melanoma incidence prior to the SCREEN intervention.15 Melanoma screening and patient awareness can change over time, and a strength of the present study is our use of a contemporary unscreened population as a comparator group.
The greatest strength of the present study is our ability to collect real-world data over a 5-year period from nearly 600 000 patients. More than 144 000 patients were screened for skin cancer at least once, allowing us to compare the cumulative incidence of melanoma by thickness in contemporary cohorts of screened and unscreened patients. A prospective randomized study of this size would not be practical or cost-effective. Our initiative did not offer physician compensation specifically for screening and used existing health care infrastructure, resulting in low direct costs. Previous studies have shown that harms such as increased costs due to dermatology visits and procedures and increased patient anxiety secondary to screening or skin biopsy were not seen among patients in this screened population.16,17
Limitations
This study does have several limitations. It was a nonrandomized, retrospective analysis of a quality initiative. Although we adjusted our findings for patient age, sex, and race and ethnicity, there are likely differences not captured in patient demographics that may confer melanoma risk, such as skin type, family history of melanoma, and sun exposure history. Additionally, factors such as patients’ willingness to be screened and physicians’ decisions of which patients to screen were not taken into account, although we did not see a significant change in the findings when physician practice group was added to the model. It is also likely that some melanomas in this study were identified first by the patient and then brought to the attention of the screening physician, and thus all tumors in screened patients may not actually be screen-detected. While the screening education was standardized, it was optional, and training rates were reported for the first year of the intervention but were not tracked after this, limiting our ability to correlate melanoma thickness with primary care clinician participation in the optional training program.4 The actual screening process was neither standardized nor formally documented. It is unlikely that all screenings were equally thorough and that all screeners had the same sensitivity for melanoma. Additionally, because clinicians could mark the screening as completed if the patient reported a skin cancer screening from another health care clinician in the previous year, we cannot be certain that the screening date noted in the EHR is the exact date on which it was performed, nor by the person who performed the screening, and it is also likely that the same patient-reported screening was noted at 2 or more visits, making it impossible to quantify the number of times an individual patient was screened. We found that most patients were only screened once during the study period and thus cannot draw conclusions about what impact regular screenings every 1 or 2 years may have had. The skin cancer screening examination also likely has an educational and awareness component that may contribute to early melanoma detection and thus have benefits beyond direct lesion detection. Because UPMC is not a closed system, it is likely that some patients had melanomas diagnosed at non-UPMC facilities that were not captured in our data. Furthermore, UPMC is a dynamic organization, and thus the primary care clinicians and practices employed by UPMC fluctuated over the 5-year period of this study.
Conclusions
In this quality improvement study, our findings show that screening for skin cancer was associated with a higher incidence of thin melanoma. While we did observe a lower incidence of thick melanomas in screened patients, particularly interval melanomas and in patients 65 years and older, these results were not statistically significant. The number of thick (>4 mm) melanomas in our population was low, and median follow-up time was only about 3 years. Our findings and study design are important because we will be able to quantify the association of screening with melanoma overdiagnosis (a potential harm of screening), and with more follow-up time, with the incidence of thick melanomas (a benefit of screening that would be anticipated to influence survival, morbidity, and treatment costs). Modeling studies have suggested that screening older patients, particularly men, for melanoma may be cost-effective.18 Real-world data can add to this to help to determine which patients, if any, can benefit from screening and the optimal screening strategies that maximize benefit and minimize cost and harms. Longer follow-up is needed to fully determine the association of screening with outcomes such as the incidence of thick melanoma, treatment cost and morbidity, distant metastasis, and death.
eTable. Frailty model adjusting for age, sex, white race and provider practice with results by sex
eFigure. Cumulative incidence for interval melanomas and for patients age 65 years and older
eAppendix. SQUIRE 2.0 Checklist
References
- 1.National Cancer Institute Surveillance, Epidemiology, and End Results Program . Cancer stat facts: melanoma of the skin. Accessed March 6, 2022. https://seer.cancer.gov/statfacts/html/melan.html
- 2.Bibbins-Domingo K, Grossman DC, Curry SJ, et al. ; US Preventive Services Task Force . Screening for skin cancer: US Preventive Services Task Force recommendation statement. JAMA. 2016;316(4):429-435. doi: 10.1001/jama.2016.8465 [DOI] [PubMed] [Google Scholar]
- 3.Welch HG, Mazer BL, Adamson AS. The rapid rise in cutaneous melanoma diagnoses. N Engl J Med. 2021;384(1):72-79. doi: 10.1056/NEJMsb2019760 [DOI] [PubMed] [Google Scholar]
- 4.Ferris LK, Saul MI, Lin Y, et al. A large skin cancer screening quality initiative: description and first-year outcomes. JAMA Oncol. 2017;3(8):1112-1115. doi: 10.1001/jamaoncol.2016.6779 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Geller AC, Greinert R, Sinclair C, et al. A nationwide population-based skin cancer screening in Germany: proceedings of the first meeting of the International Task Force on Skin Cancer Screening and Prevention (September 24 and 25, 2009). Cancer Epidemiol. 2010;34(3):355-358. doi: 10.1016/j.canep.2010.03.006 [DOI] [PubMed] [Google Scholar]
- 6.Eide MJ, Asgari MM, Fletcher SW, et al. ; INFORMED (INternet course FOR Melanoma Early Detection) Group . Effects on skills and practice from a web-based skin cancer course for primary care providers. J Am Board Fam Med. 2013;26(6):648-657. doi: 10.3122/jabfm.2013.06.130108 [DOI] [PubMed] [Google Scholar]
- 7.Ferris LK, Saul MI, Lin Y, et al. A large skin cancer screening quality initiative: description and first-year outcomes. JAMA Oncol. 2017;3(8):1112-1115. doi: 10.1001/jamaoncol.2016.6779 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Ogrinc G, Davies L, Goodman D, Batalden P, Davidoff F, Stevens D. SQUIRE 2.0 (Standards for QUality Improvement Reporting Excellence): revised publication guidelines from a detailed consensus process. BMJ Qual Saf. 2016;25(12):986-992. doi: 10.1136/bmjqs-2015-004411 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Rubin R. Melanoma diagnoses rise while mortality stays fairly flat, raising concerns about overdiagnosis. JAMA. 2020;323(15):1429-1430. doi: 10.1001/jama.2020.2669 [DOI] [PubMed] [Google Scholar]
- 10.Forsea AM, Del Marmol V, de Vries E, Bailey EE, Geller AC. Melanoma incidence and mortality in Europe: new estimates, persistent disparities. Br J Dermatol. 2012;167(5):1124-1130. doi: 10.1111/j.1365-2133.2012.11125.x [DOI] [PubMed] [Google Scholar]
- 11.Zhang SH, Liu R, Siripong N, et al. Thick melanoma is associated with low melanoma knowledge and low perceived health competence, but not delays in care. J Am Acad Dermatol. 2020;83(2):587-590. doi: 10.1016/j.jaad.2019.05.068 [DOI] [PubMed] [Google Scholar]
- 12.Katalinic A, Waldmann A, Weinstock MA, et al. Does skin cancer screening save lives?: an observational study comparing trends in melanoma mortality in regions with and without screening. Cancer. 2012;118(21):5395-5402. doi: 10.1002/cncr.27566 [DOI] [PubMed] [Google Scholar]
- 13.Eisemann N, Waldmann A, Katalinic A. [Incidence of melanoma and changes in stage-specific incidence after implementation of skin cancer screening in Schleswig-Holstein]. Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz. 2014;57(1):77-83. doi: 10.1007/s00103-013-1876-1 [DOI] [PubMed] [Google Scholar]
- 14.Johansson M, Brodersen J, Gøtzsche PC, Jørgensen KJ. Screening for reducing morbidity and mortality in malignant melanoma. Cochrane Database Syst Rev. 2019;6(6):CD012352. doi: 10.1002/14651858.CD012352.pub2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Hübner J, Waldmann A, Geller AC, et al. Interval cancers after skin cancer screening: incidence, tumour characteristics and risk factors for cutaneous melanoma. Br J Cancer. 2017;116(2):253-259. doi: 10.1038/bjc.2016.390 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Weinstock MAMA, Ferris LKLK, Saul MIMI, et al. Downstream consequences of melanoma screening in a community practice setting: first results. Cancer. 2016;122(20):3152-3156. doi: 10.1002/cncr.30177 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Matthews NH, Risica PM, Ferris LK, et al. Psychosocial impact of skin biopsies in the setting of melanoma screening: a cross-sectional survey. Br J Dermatol. 2019;180(3):664-665. doi: 10.1111/bjd.17134 [DOI] [PubMed] [Google Scholar]
- 18.Adamson AS, Jarmul JA, Pignone MP. Screening for melanoma in men: a cost-effectiveness analysis. J Gen Intern Med. 2020;35(4):1175-1181. doi: 10.1007/s11606-019-05443-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eTable. Frailty model adjusting for age, sex, white race and provider practice with results by sex
eFigure. Cumulative incidence for interval melanomas and for patients age 65 years and older
eAppendix. SQUIRE 2.0 Checklist
