Melanoma Thickness and Survival Trends in the United States, 1989–2009

Abstract

Background:

With melanoma incidence rising and mortality stable, some question whether the melanoma epidemic is real. Melanoma thickness and survival trends may provide insights, but previous studies have been limited because of missing data on thickness.

Methods:

With a validated imputation method for missing thickness data, we characterized melanoma thickness and survival trends among men and women in the Surveillance, Epidemiology, and End Results (SEER)–9 registries between 1989 and 2009. A total of 98,498 cases of invasive melanoma were identified. All statistical tests were two-sided.

Results:

Incidence per 100 000 person-years increased (13.94, 95% confidence interval [CI] = 13.65 to 14.23, to 21.87, 95% CI = 21.56 to 22.19, P < .001) between 1989 to 1991 and 2007 to 2009, fatal incidence remained stable (2.32, 95% CI = 2.2 to 2.4, to 2.08, 95% CI = 2.0 to 2.2, P = .20) between 1989 to 1991 and 1998 to 2000, and five-year survival increased (88.29%, 95% CI = 87.60% to 88.95%, to 91.68%, 95% CI = 91.22% to 92.12%, P < .001) between 1989 to 1991 and 2001 to 2003. Increase in incidence occurred across all thickness groups. Median thickness decreased (0.73 to 0.58mm). Geometric mean thickness decreased (0.77 to 0.65mm) 4.6% (95% CI = 4.2% to 5.0%) every three years in multivariable analysis. Thickness decreased among T1/T2 tumors (0.01–1.00 and 1.01–2.00mm) and among all age and sex groups, whites, non-Hispanics, and all body sites. However, thickness increased among T3/T4 tumors (2.01–4.00 and > 4.00mm) and nodular melanomas; acral lentiginous melanomas approached statistical significance. Thickness remained unchanged among some racial minorities. Melanoma-specific survival improved (hazard ratio [HR] = 0.89, 95% CI = 0.88 to 0.91) every three years in multivariable analysis. Improvements in survival occurred across all subgroups except nonblack minorities, and nodular and acral lentiginous subtypes.

Conclusions:

Increasing incidence across all thickness groups coupled with T3/T4 lesions becoming thicker suggests that the melanoma epidemic is real and not simply an artifact of increased detection pressure of earlier-stage T1/T2 lesions. Survival is generally improving independent of thickness, but improvements in survival have not been experienced by certain minorities, and nodular and acral lentiginous subtypes.

In the United States, melanoma is the fifth most common cancer among men and the seventh most common cancer among women (1). Over the past two decades, melanoma incidence has dramatically risen but melanoma mortality has only recently stabilized (2). This discrepancy has led to the controversy of whether the melanoma epidemic is real or an artifact of increased detection pressure. Thickness is the most important factor in melanoma prognosis; therefore, examining the most recent trends in thickness and survival may provide key insights. Furthermore, subgroup analysis of melanoma thickness and survival may elucidate novel means to decrease the stubbornly stable mortality rate. Previous studies analyzing tumor thickness have encountered missing data on thickness within the Surveillance, Epidemiology, and End Results (SEER) Program database (3–6), thus undermining findings because of a major selection bias. However, multiple imputation methods using the SEER database have been shown to overcome this limitation (7). Thus, our aim was to thoroughly characterize invasive melanoma thickness and survival trends in the United States employing a previously validated multiple-imputation technique for thickness.

Methods

Study Population

Data for this retrospective cohort study was obtained from the National Cancer Institute’s SEER Program. We restricted our analysis to the nine original SEER registries (Atlanta, GA; CT; Detroit, MI; HI; IA; NM; San Francisco-Oakland, CA; Seattle-Puget Sound, WA; and UT) as the longevity of these registries allowed for time trend analysis compared with other registries that start after 1992. The SEER-9 registries comprise approximately 9.4% of the US population.

Cases

We limited our analysis to nonoccult, microscopically confirmed, invasive melanoma cases with active follow-up from 1989 to 2009. The International Classification of Disease for Oncology, Third Edition (ICD-O-3) histology and site codes, 8720–8790 and C44.01-C44.9, respectively, were used to define cases of invasive melanoma.

Outcome Measures

The main outcome measures were overall incidence, ultimately fatal incidence, melanoma-specific five-year survival rates, Breslow thickness measured on a 0.01 to 9.90mm scale, death because of melanoma based on death certificate–reported cause of death (alive/death not because of melanoma, or death because of melanoma), and survival time (measured in months from time of diagnosis). SEER initiated collection of Breslow thickness starting in 1988. From 1989 through 2003, Breslow thickness was measured on a 0.01 to 9.90mm scale, whereas from 2004 to 2009 it was measured on a 0.01 to 9.89mm scale (with any measurement greater than 9.89mm being coded as 9.89mm). Thickness was classified based on the American Joint Committee on Cancer thickness categories (ie, T1 [0.01–1.00 mm], T2 [1.01–2.00 mm], T3 [2.01–4.00 mm], and T4 [> 4.00 mm]). A survival time of 0 months was recoded as 0.5 months to include cases who died within one month of diagnosis but did not reach the one-month threshold. Ultimately fatal incidence is defined as the incidence of a case patient for whom the cause of death listed on the death certificate is melanoma. The use of ultimately fatal incidence as opposed to mortality rate facilitates fair year-to-year comparison to overall incidence because it uses year of diagnosis instead of year of death.

Covariates

Covariates of interest in the analysis included year of diagnosis and sociodemographic, tumor, staging, and therapeutic characteristics. Sociodemographic covariates are sex, age at diagnosis, race (white, black, American Indian/Alaskan Native, Asian/Pacific Islander, other, or unknown), Hispanic ethnicity (Hispanic or non-Hispanic), marital status (single, married, separated, divorced, widowed, or unknown), SEER registry location, county-level socioeconomic status indicators (median household income, % with less than a high school education, % unemployed, % white-collar jobs, % living below the poverty level, and % living in urban areas). Tumor covariates are tumor location (face [C44.0–44.3], scalp/neck [C44.4], trunk [C44.5], upper limb [C44.6], lower limb [44.7], overlapping [C44.8], and location not-otherwise-specified [C44.9]) and histologic subtypes (superficial spreading melanoma [code 8743], nodular melanoma [code 8721], lentigo maligna melanoma [code 8742], acral lentiginous melanoma [code 8744], amelanotic melanoma [code 8730], other uncommon melanomas [codes 8722, 8723, 8728, 8740, 8741, 8745, 8746, 8761, 8770–8774, and 8780], and melanoma not-otherwise-specified [code 8720]). Staging covariates include thickness, Clark level (II, III, IV, V, or unknown), ulceration positivity (yes, no, or unknown), lymph node positivity (yes [defined as any case with positive regional lymph nodes, in-transit metastasis, microscopic lymph nodes, or macroscopic lymph nodes], no, or unknown), metastasis positivity (yes [defined as any case with distant lymph nodes, metastasis to skin/subcutaneous tissue beyond regional lymph nodes, lung metastasis, visceral metastasis, or distant metastasis not-otherwise-specified], no, or unknown). Finally, a therapeutic characteristic is cancer-directed surgery (any [defined as any case patient treated with local tissue destruction, excision, amputation, or surgery not-otherwise-specified], none, or unknown).

Statistical Analysis

Because thickness was missing for 13.35% of cases, we employed a previously described and validated method known as multivariable imputation with chained equations (MICE), a sequential multivariable regression technique, to impute thickness values for cases with missing data on thickness (7). We used to the following factors to imputed thickness: year of diagnosis, age, sex, race, Hispanic ethnicity, marital status, SEER registry, county-level socioeconomic status indicators (median household income, % with less than a high school education, % unemployed, % white-collar jobs, % living below the poverty level, and % living in urban areas), tumor location, histologic subtype, Clark level, ulceration/lymph node/metastasis positivity, cancer-directed surgery, vital status, and survival time. Given there was no statistically significant or substantive difference in values of geometric mean thickness and its variance when running one compared with five imputations, we used one imputation to simplify analysis and use a consistent dataset.

Trends in incidence rates over time were determined by using jointpoint regression models (8) via SEER’s Jointpoint Regression Program (version 3.4.3). Because overall relative survival for melanoma stabilizes in approximately nine years (data not shown), calculation of ultimately fatal incidence rates was limited to melanoma cases diagnosed with at least nine years of active follow-up (ie, cases diagnosed on or before 2000). Calculation of five-year survival rates was limited to cases with at least five years of active follow-up (ie, cases diagnosed on or before 2004).

Because of the lognormal distribution of thickness, we characterized the central tendency using the median and geometric mean (defined as the product of x numbers taken to the x^th root). Thickness trends were assessed by construction of linear regression models using the natural logarithm of thickness as the dependent variable and the three-year time period as the independent variable. The following formula was used to calculate the percentage change (PC) in geometric mean thickness per three-year time period: $PC=(exp[\beta coefficentofthreeyeartimeperiod]-1)\times 100$ . Modeling was performed using univariate linear regression with subgroup analysis and multivariable linear regression. Trends in melanoma survival were determined using univariate and multivariable Cox proportional hazards regression, with proportionality verified via Schoenfeld plots. Univariate hazard ratios with subgroup analysis and multivariate hazard ratios were generated, which represented risk of death because of melanoma per three-year time period. Missing data for covariates other than thickness were coded as defined unknown groups (dummy variables), thus allowing for complete case analysis in all models. All analyses were conducted on the complete dataset (cases with known or imputed thickness) and repeated on the original dataset (cases with only known thickness). Results for the original dataset were similar and are shown in Supplementary Tables 1–4 (available online). All regression analyses were conducted using Stata IC (version 11.0). All P values are two-sided and considered statistically significant at an alpha level of .05.

Results

Overall Incidence, Ultimately Fatal Incidence, and Relative Survival Rate Trends

A total of 98 498 invasive melanomas were identified within the original nine SEER registries from 1989 through 2009, of which 13.35% (13 153 cases) had missing data on thickness. From 1989 through 1991 to 2007 through 2009, the overall incidence rate per 100 000 person-years increased from 13.94 (95% confidence interval [CI] = 13.65 to 14.23) to 21.87 (95% CI = 21.56 to 22.19), P_trend < .001 (Table 1). T1 cases rose from 8.73 to 15.61 (P_trend < .001), T2 cases rose from 2.65 to 3.23 (P_trend = .004), T3 cases rose from 1.61 to 1.79 (P_trend < .001), and T4 cases rose 0.95 to 1.24 (P_trend = .01)

Table 1.

Melanoma age-adjusted overall incidence per 100 000 person-years, age-adjusted ultimately fatal incidence per 100 000 person-years, and survival trends stratified by AJCC thickness category, SEER 9, United States, 1989–2009 (n = 98 498)

Outcome	1989–1991	1992–1994	1995–1997	1998–2000	2001–2003	2004–2006	2007–2009	Ptrend*
Incidence
Overall	13.94	14.92	16.94	18.19	19.17	21.07	21.87	<.001
T1	8.73	9.56	11.29	12.52	13.44	14.97	15.61	<.001
T2	2.65	2.79	3.07	3.03	3.01	3.21	3.23	.004
T3	1.61	1.66	1.69	1.69	1.74	1.77	1.79	<.001
T4	0.95	0.91	0.90	0.96	0.98	1.12	1.24	.01
Ultimately fatal incidence
Overall	2.32	2.38	2.30	2.08	--	--	--	.20
T1	0.58	0.68	0.63	0.62	--	--	--	.80
T2	0.63	0.63	0.63	0.50	--	--	--	.20
T3	0.59	0.60	0.60	0.55	--	--	--	.30
T4	0.52	0.47	0.44	0.41	--	--	--	.01
Melanoma-specific 5-year survival, %
Overall	88.29	88.36	89.67	90.75	91.68	--	--	<.001
T1	96.54	95.82	96.44	95.50	97.08	--	--	<.001
T2	84.68	83.70	85.18	86.87	87.87	--	--	<.001
T3	69.67	70.74	69.92	71.18	74.24	--	--	.02
T4	49.96	51.25	52.40	58.07	56.40	--	--	.001

* P_trend value calculated using jointpoint regression and Cox proportional hazards regression; all P values are two-sided. AJCC = American Joint Committee on Cancer; SEER = Surveillance, Epidemiology, and End Results.

From 1989 through 1991 to 1998 through 2000, ultimately fatal incidence rate per 100 000 person-years remained stable from 2.32 (95% CI = 2.2 to 2.4) to 2.08 (95% CI = 2.0 to 2.2), P_trend = .20 (Table 1). The ultimately fatal incidence rates remained unchanged among T1 (0.58 to 0.62, P_trend = .80), T2 (0.63 to 0.50, P_trend = .20), and T3 (0.59 to 0.55, P_trend = .30) cases, whereas it statistically significantly declined among T4 (0.52 to 0.41, P_trend = .01) cases. From 1989 through 1991 to 2001 through 2003, five-year survival increased from 88.29% (95% CI = 87.60% to 88.95%) to 91.68% (95% CI = 91.22% to 92.12%), P_trend < .001. It increased among T1 (96.54% to 97.08%, P_trend < .001), T2 (84.68% to 87.87%, P_trend < .001), T3 (69.67% to 74.24%, P_trend = .02), and T4 (49.96% to 56.40%, P_trend= .001) cases (Table 1).

Melanoma Thickness Trends

From 1989 through 1991 to 2007 through 2009, overall median thickness decreased from 0.73 to 0.58mm (Table 2). Overall geometric mean thickness decreased from 0.77 to 0.65mm (Table 2) with an average 3.05% (95% CI = 2.72% to 3.37%) and 4.6% (95% CI = 4.2% to 5.0%) decrease per three-year time period in univariate and multivariable analyses, respectively (Table 3). Geometric mean thickness statistically significantly decreased among T1 cases from 0.41 to 0.39mm (PC = -1.22% per three-year time period, 95% CI = -1.49% to -0.95%) and among T2 cases from 1.41 to 1.39mm (PC = -0.28%, 95% CI = -0.43% to -0.12%). Geometric mean thickness statistically significantly increased among T3 cases from 2.78 to 2.84mm (PC = 0.38%, 95% CI = 0.17% to 0.59%) and among T4 cases from 6.03 to 6.25mm (PC = 0.76%, 95% CI = 0.38% to 1.14%). Overall geometric mean thickness decreased after controlling for sociodemographic, tumor, staging, and therapeutic and survival characteristics (Table 4, Models 1–3).

Table 2.

Melanoma thickness trends by three-year period, SEER 9, United States, 1989–2009 (n = 98 498)*

Time period	Median thickness, mm	Geometric thickness mean, mm
	Overall	Overall	T1	T2	T3	T4
1989–1991	0.73	0.77	0.41	1.41	2.78	6.03
1992–1994	0.70	0.75	0.41	1.41	2.78	6.03
1995–1997	0.66	0.71	0.41	1.42	2.76	6.05
1998–2000	0.62	0.67	0.40	1.39	2.81	5.93
2001–2003	0.60	0.66	0.40	1.41	2.79	6.02
2004–2006	0.57	0.64	0.38	1.39	2.82	6.31
2007–2009	0.58	0.65	0.39	1.39	2.84	6.25

* AJCC = American Joint Committee on Cancer; SEER = Surveillance, Epidemiology, and End Results.

Table 3.

Melanoma trends: crude percentage change in geometric mean thickness and crude hazard ratios predicting risk of death from melanoma per three-year period, SEER 9, United States, 1989–2009 (n = 98 498)

Characteristic	No. (%)	PC (95% CI)	HR (95% CI)
Overall		-3.05* (-3.37 to -2.72)	0.92* (0.91 to 0.93)
AJCC thickness category
T1	67 792 (68.83)	-1.22* (-1.49 to -0.95)	0.94* (0.92 to 0.96)
T2	16 185 (16.43)	-0.28* (-0.43 to -0.12)	0.93* (0.91 to 0.95)
T3	9121 (9.26)	0.38* (0.17 to 0.59)	0.96* (0.94 to 0.98)
T4	5400 (5.48)	0.76* (0.38 to 1.14)	0.94* (0.92 to 0.96)
Sex and age, y
Men	54 920	-3.19* (-3.63 to -2.75)	0.92* (0.90 to 0.93)
<50	14 933 (15.16)	-2.88* (-3.66 to -2.10)	0.89* (0.86 to 0.92)
50–64	16 866 (17.12)	-3.92* (-4.69 to -3.14)	0.87* (0.85 to 0.89)
≥65	23 121 (23.47)	-3.42* (-4.14 to -2.69)	0.93* (0.91 to 0.95)
Women	43 578	-2.99* (-3.46 to -2.52)	0.91* (0.89 to 0.92)
<50	18 975 (19.26)	-2.70* (-3.34 to -2.06)	0.86* (0.82 to 0.89)
50–64	11 203 (11.37)	-4.39* (-5.29 to -3.49)	0.87* (0.83 to 0.91)
≥65	13 400 (13.60)	-3.03* (-3.98 to -2.06)	0.93* (0.90 to 0.95)
Race
White	95 091 (96.54)	-3.05* (-3.38 to -2.72)	0.92* (0.90 to 0.93)
Black	445 (0.45)	-0.48 (-5.86 to 5.20)	0.89 (0.80 to 1.00)
American Indian/Alaskan Native	189 (0.19)	-9.52* (-16.70 to -1.72)	0.94 (0.75 to 1.19)
Asian/Pacific Islander	705 (0.72)	2.12 (-2.56 to 7.02)	0.91 (0.82 to 1.02)
Other	119 (0.12)	4.82 (-6.71 to 17.76)	2.77 (0.66 to 11.58)
Unknown	1949 (1.98)	-3.46* (-5.27 to -1.62)	1.00 (0.63 to 1.59)
Hispanic Ethnicity
Non-Hispanic	96 715 (98.19)	-3.05* (-3.38 to -2.73)	0.92* (0.90 to 0.93)
Hispanic	1783 (1.81)	-2.76 (-5.45 to 0.01)	0.92* (0.85 to 0.99)
Marital status
Single	11 005 (11.17)	-2.66* (-3.66 to -1.65)	0.91* (0.88 to 0.94)
Married	49 892 (50.65)	-2.20* (-2.66 to -1.73)	0.91* (0.90 to 0.93)
Separated	294 (0.30)	3.62 (-3.85 to 11.66)	0.89 (0.75 to 1.05)
Divorced	4361 (4.43)	-2.01* (-3.72 to -0.27)	0.90* (0.86 to 0.94)
Widowed	6463 (6.56)	-1.97* (-3.41 to -0.52)	0.95* (0.92 to 0.98)
Unknown	26 483 (26.89)	-4.59* (-5.09 to -4.08)	0.94* (0.91 to 0.98)
Tumor location
Face	13 013 (13.21)	-2.76* (-3.68 to -1.82)	0.99 (0.96 to 1.03)
Scalp/neck	6897 (7.00)	-2.71* (-4.07 to -1.34)	0.95* (0.92 to 0.98)
Trunk	32 854 (33.35)	-2.81* (-3.35 to -2.28)	0.90* (0.88 to 0.92)
Upper limb	23 820 (24.18)	-2.44* (-3.09 to -1.79)	0.91* (0.89 to 0.94)
Lower limb	19 022 (19.31)	-2.89* (-3.62 to -2.16)	0.91* (0.88 to 0.94)
Overlapping	70 (0.07)	-2.77 (-15.68 to 12.12)	0.88 (0.61 to 1.28)
Location not-otherwise-specified	2822 (2.87)	-7.76* (-9.74 to -5.74)	1.05 (1.01 to 1.08)
Histologic subtype
Superficial spreading melanoma	36 717 (37.28)	-0.97* (-1.41 to -0.52)	0.91* (0.89 to 0.94)
Nodular melanoma	7104 (7.21)	1.70* (0.50 to 2.91)	1.00 (0.97 to 1.03)
Lentigo maligna melanoma	7525 (7.64)	-3.27* (-4.26 to -2.27)	0.99 (0.92 to 1.07)
Acral lentiginous melanoma	1063 (1.08)	3.09 (-0.36 to 6.65)	0.95 (0.88 to 1.03)
Amelanotic melanoma	389 (0.39)	-9.41* (-14.31 to -4.23)	0.82* (0.74 to 0.91)
Other uncommon melanomas	3517 (3.57)	-0.75 (-2.83 to 1.38)	0.93* (0.88 to 0.99)
Melanoma not-otherwise-specified	42 183 (42.83)	-6.02* (-6.51 to -5.53)	0.87* (0.85 to 0.88)
Clark level
2	35 295 (35.83)	-3.14* (-3.50 to -2.78)	0.91* (0.86 to 0.96)
3	20 887 (21.21)	-5.55* (-6.01 to -5.08)	0.85* (0.82 to 0.88)
4	32 479 (32.97)	-0.82* (-1.36 to -0.27)	0.97* (0.95 to 0.98)
5	3093 (3.14)	2.37* (0.32 to 4.47)	0.99 (0.96 to 1.02)
Unknown	6744 (6.85)	-5.19* (-6.58 to -3.78)	1.02 (1.00 to 1.04)
Ulceration positivity
Yes	6930 (7.04)	-0.05 (-1.50 to 1.42)	0.99 (0.96 to 1.01)
No	83 382 (84.65)	-4.22* (-4.54 to -3.90)	0.89* (0.87 to 0.90)
Unknown	8186 (8.31)	-8.97* (-10.12 to -7.82)	0.89* (0.88 to 0.91)
Lymph node positivity
Yes	4902 (4.98)	1.43 (-0.19 to 3.08)	0.87* (0.85 to 0.89)
No	62 881 (63.84)	-5.07* (-5.50 to -4.63)	0.87* (0.85 to 0.88)
Unknown	30 715 (31.18)	-2.98* (-3.64 to -2.33)	1.04* (1.02 to 1.07)
Metastasis positivity
Yes	2624 (2.66)	-5.03* (-7.18 to -2.82)	0.96* (0.94 to 0.99)
No	91 026 (92.41)	-2.49* (-2.83 to -2.16)	0.93* (0.92 to 0.94)
Unknown	4848 (4.92)	-6.55* (-7.91 to -5.14)	0.88* (0.85 to 0.92)
Cancer-directed surgery
Yes	94 300 (95.74)	-2.88* (-3.21 to -2.55)	0.90* (0.89 to 0.92)
No	4106 (4.17)	-4.97* (-6.63 to -3.29)	1.09* (1.06 to 1.12)
Unknown	92 (0.09)	-7.02 (-15.68 to 2.53)	1.00 (0.71 to 1.41)

* P < .05; P value calculated using linear regression and Cox proportional hazards regression. All P values are two-sided. AJCC = American Joint Committee on Cancer; CI = confidence interval; HR = hazard ratio; PC = percentage change; SEER = Surveillance, Epidemiology, and End Results.

Table 4.

Multivariable analysis of trends in melanoma thickness and melanoma survival per three-year period, SEER 9, United States, 1989–2009 (n = 98 498)

Melanoma thickness trends: PC in geometric mean thickness per three-year period
Multivariable linear regression model	PC (95% CI)
Model 1: Crude model (three-year period only)	-3.05* (-3.37 to -2.72)
Model 2: Model 1 plus adjusted for age, sex, race, Hispanic ethnicity, marital status, SEER registry and county-level SES (median household income, % high school education, % unemployed, % white collar jobs, % below poverty, and % urban)	-3.16* (-3.59 to -2.73)
Model 3: Model 2 plus adjusted for tumor location and histologic subtype	-3.20* (-3.60 to -2.79)
Model 3a: Model 3 restricted to T1	-1.51* (-1.87 to -1.14)
Model 3b: Model 3 restricted to T2	-0.47* (-0.68 to -0.25)
Model 3c: Model 3 restricted to T3	0.33* (0.04 to 0.63)
Model 3d: Model 3 restricted to T4	1.21* (0.68 to 1.75)
Melanoma survival trends: HRs predicting risk of death from melanoma per three-year period
Multivariable Cox regression model	HR (95% CI)
Model A: Crude model (three-year period only)	0.92* (0.91 to 0.93)
Model B: Model A plus adjusted for age, sex, race, Hispanic ethnicity, marital status, SEER registry and county-level SES (median household income, % high school education, % unemployed, % white collar jobs, % below poverty, and % urban)	0.91* (0.89 to 0.93)
Model C: Model B plus adjusted for tumor location and histologic subtype	0.92* (0.91 to 0.94)
Model D: Model C plus adjusted for tumor thickness	0.93* (0.92 to 0.95)
Model E: Model D plus adjusted for Clark level and ulceration positivity	0.92* (0.90 to 0.94)
Model F: Model E plus adjusted for lymph node positivity and metastasis positivity	0.90* (0.89 to 0.92)
Model G: Model F plus adjusted for cancer-directed surgery	0.89* (0.87 to 0.91)

* P < .05; P value calculated using linear regression and Cox proportional hazards regression. All P values are two-sided. CI = confidence interval; HR = hazard ratio; PC = percentage change; SEER = Surveillance, Epidemiology, and End Results.

In subgroup analysis (Table 3), thickness statistically significantly decreased among T1/T2 tumors, both sexes, all age groups, whites, non-Hispanics, nonseparated individuals, all body sites, and most histologic subtypes; thickness among Hispanics approached but did not meet statistical significance (P = .05). However, thickness statistically significantly increased among T3/T4 tumors and nodular melanoma; thickness among acral lectiginous melanoma approached but did not meet statistical significance (P = .08). No statistically significant change was observed among most racial minorities, separated individuals, and uncommon melanoma histologic subtypes.

Melanoma Survival Trends

Melanoma-specific survival statistically significantly improved (hazard ratio [HR] = 0.92, 95% CI = 0.91 to 0.93) per three-year time period (Table 3); this improvement was observed even after controlling for sociodemographic, tumor, staging, and therapeutic characteristics (HR = 0.89, 95% CI = 0.88 to 0.91) (Table 4, Models A-G). In subgroup analysis (Table 3), statistically significant improvement in survival occurred across T categories, both sexes, all age groups, whites, Hispanics and non-Hispanics, nonseparated individuals, nonfacial tumors, superficial spreading, and amelanotic and other uncommon melanoma histologic subtypes; improvement in survival among blacks approached but did not meet statistical significance (P = .05). No statistically significant change in survival was seen among nonblack racial minorities, separated individuals, facial tumors, and nodular, lentigo maligna, and acral lentiginous histologic subtypes.

Discussion

Absolute increases in overall incidence occurred across T categories with the greatest rise among T1 tumors. However, there is a growing divergence: T1/T2 tumors are becoming thinner, while T3/T4 tumors are becoming thicker. Given the small relative change in geometric mean thickness among T1/T2 lesions compared with T3/T4 lesions (Table 2), the likely explanation for the overall decrease in thickness is the high prevalence of T1/T2 lesions coupled with the rapid rise in incidence of T1/T2 lesions compared with T3/T4 lesions.

Several studies have demonstrated the overall rise in incidence across all thickness groups (5,9,10). This has been attributed to either: 1) a true rise in melanoma incidence because of increased risk (ie, UV radiation, and indoor tanning) or 2) an artifact of detection pressure resulting in the increased diagnosis of indolent disease. Linos et al. shed light on this debate by showing increases in melanoma incidence across all thickness categories, even among low socioeconomic status groups with presumably limited access to early detection efforts (3). Our findings of an increasing incidence across all thickness groups coupled with growing divergence in thickness between T1/T2 and T3/T4 lesions, even after controlling for sociodemographic factors (Model 3, C-D), suggest that the overall burden of disease is truly increasing, albeit primarily stemming from increases in T1/T2 disease potentially because of increased detection. Although our study is unable to determine the reasons for these phenomena, possible explanations may include disparities in education and more vigilant screening and prevention efforts amongst various subgroups.

Melanoma-specific survival improved independent of changes in sociodemographic, tumor, staging (including thickness), and surgical therapeutic characteristics. This may be explained by the preferential detection of indolent vs clinically relevant melanomas or therapeutic advances in melanoma management. Less likely, but possible, is the idea that increased detection efforts are in fact improving survival by early detection of thin but ultimately fatal melanomas. In other words, these melanomas are capable of early metastatic spread independent of tumor depth. Support for the view is based on 30% of melanoma deaths being attributable to thin lesions (5), as well as mitotic rate and ulceration status modifying the prognosis of T1 melanomas (11). Theories underlying the rationale of early detection may need to include other disease progression markers besides depth of invasion, such as tumor genetic profiles and biomarkers.

Teasing out the reasons for improving melanoma-specific survival is challenging. Trends in incidence of ultimately fatal melanoma stratified by thickness group may provide some insight. As with previous studies, ultimately fatal incidence has remained stable; however, unlike previous studies there was a statistically and clinically significant decline in ultimately fatal incidence among the T4 lesions. This may indicate a modest but clinically significant improvement in management of stage 2B/2C (T4) and possibly stage 3 and 4 diseases. Improvements in management may include more effective surgical therapies (ie, standardization of wide local excisions and introduction of sentinel lymph node dissections leading to improved staging and tailored therapy) and the arrival of immunotherapy (interferon-alpha and interleukin-2) (12–14).

In addition, viewing trends in overall incidence may provide insight into gains in melanoma-specific survival over time as well. In general, while ultimately fatal incidence is stable, overall incidence is rapidly increasing, with the greatest rise among T1 tumors. Even among high-risk elderly males, for whom ultimately fatal incidence is increasing, the overall incidence is increasing at a more rapid rate (data not shown). The result is an increasing pool of people with melanoma who are less likely to die of melanoma. This phenomenon will increase the melanoma-specific survival rate possibly because of a lead-time, length, and/or overdiagnosis bias. Indeed, Welch et al. in a study analyzing general cancer trends found a high correlation between changes in tumor incidence and five-year survival and little or no correlation between changes in tumor mortality and five-year survival (15). Thus, it appears that improvements in melanoma-specific survival are likely a combination of improved management of T4 disease, a shift toward detection of thinner melanomas within T1/T2 categories, and increased detection of T1/T2 disease.

Health disparities are widespread in the United States, including for melanoma. Numerous studies have documented disparities in melanoma thickness, stage, and survival in regards to race and ethnicity (16–18), marital status (19, 20), and histologic subtypes (21–24). Results from our study show that positive trends in melanoma thickness and melanoma-specific survival have not been experienced by certain subgroups, notably some minorities, separated individuals, and nodular and acral lentiginous histologic subtypes. The impact is a widening of the already present disparity in thicker disease and poorer survival within these subgroups. However, given the relatively small sample sizes in some minorities and separated individuals it is possible our study was underpowered to detect a statistically significant difference. Targeting these subgroups with primary and secondary prevention efforts may not reduce mortality or be cost-effective given their relative infrequency. However, they may benefit from more improved management such as c-KIT or tyrosine kinase inhibitors for acral lentiginous melanomas (25). In addition, focus should be sharpened on a deeper understanding of the unique biology of nodular and acral lentiginous melanomas given their disproportionate but clinically significant impact on melanoma mortality relative to their incidence (24).

In general, the magnitude of the trends in thickness and survival were attenuated using the original dataset with only the known thickness cases (Supplementary Tables 1–4, available online) compared with the complete dataset with known and imputed thickness cases. Our previous study showed that unknown thickness cases tended to have later-stage disease (ie, higher Clark level and increased likelihood of positive lymph node involvement and metastasis) with lower survival, increased mortality, and lower rates of surgical therapy compared with known thickness cases (7). By incorporating these unknown thickness cases into our analysis, through the employment of multiple imputation our analysis provides a more complete and robust evaluation of underlying trends in melanoma thickness and survival.

A major limitation of this study was the inability to adequately assess the impact of therapeutic details on melanoma thickness and survival trends given the lack of detailed information on surgery (ie, margin size, use of sentinel lymph node biopsies), immunotherapy, chemotherapy, and radiation therapy. Although socioeconomic factors were controlled in our analysis, they were assessed at the county level, not the individual level, thus making our analysis vulnerable to ecologic fallacy. Some inherent limitations of the SEER database could not be overcome, such as missing data on other variables, thickness being measured on a fixed 0.01 to 9.90mm scale, and lack of independent diagnostic verification of cases. Also, we could not analyze trends across all SEER-18 registries, given the addition of the several of the registries after 1989.

Despite these limitations, our study has several noteworthy strengths. We used a national, population-based, quality-controlled database. We overcame SEER’s lack of thickness information on 13% of cases by employing a previously validated multiple imputation for thickness (7). For future studies that utilize the SEER database, multiple imputation as a means to overcome limitations of missing data on thickness should be considered. Finally, subgroup and multivariate analysis was performed and utilized sociodemographic, tumor, staging, and therapeutic and survival characteristics to help minimize confounding.

This study has several novel and noteworthy implications. The melanoma epidemic is not simply a product of increased detection of T1 disease, because T2/T3/T4 disease is increasing as well. However, the majority of the increase in incidence comes from T1 disease. There is a growing divergence in melanoma disease stages as T1/T2 lesions are becoming thinner and T3/T4 lesions are becoming thicker, presumably because of disparities in diagnosis and management among various subgroups. Melanoma survival is improving independent of changes in thickness, and treatment in melanoma mortality is being made in regards to T4 melanomas, though the exact reasons for both these phenomena are difficult to determine. Finally, disparities in melanoma thickness and survival are growing in certain subgroups and histologic subtypes that may justify more improved management in certain cases.

Future studies should analyze trends in AJCC stages, identify unique characteristics of clinically relevant vs indolent disease, and determine the impact of specific prevention and management activities on epidemiologic trends. Reducing melanoma mortality must be driven by continued emphasis on prevention, early detection, and improved therapeutics.