Abstract
The prognosis for advanced melanoma (AM) is extremely poor. Some patients are already in an advanced stage at the time of their first diagnosis and face a significant risk of early death. This study predicted all-cause early death and cancer-specific early death in patients with AM by identifying independent risk factors, building 2 separate nomogram models, and validating the efficiency of the models. A total of 2138 patients diagnosed with AM from 2010 to 2015 were registered in the Surveillance, Epidemiology and End Results (SEER) database and randomly assigned to a training cohort and a validation cohort. Logistic regression models were used to identify the associated independent risk factors. These factors have also been used to build nomograms for early deaths. Next, we validated the model’s predictive power by examining subject operating characteristic curves, then applied calibration curves to assess the accuracy of the models, and finally, tested the net benefit of interventions based on decision curve analysis. The results of the logistic regression model showed that marital status, primary site, histological type, N stage, surgery, chemotherapy, bone, liver, lung and brain metastases were significant independent risk factors for early death. These identified factors contributed to the creation of 2 nomograms, which predict the risk of all-cause early death and cancer-specific early death in patients with AM. In the all-cause early death model, the area under the curve was 0.751 and 0.759 for the training and validation groups, respectively, whereas in the cancer-specific early death model, the area under the curve was 0.740 and 0.757 for the training and validation groups. Calibration curves indicated a high degree of agreement between the predicted and observed probabilities, and the decision curve analysis demonstrated a high value for the model in terms of its applicability in clinical settings. These nomograms have practical applications in predicting the risk of early death in patients with AM, helping oncologists to intervene early and develop more personalized treatment strategies.
Keywords: advanced melanoma, early death, Epidemiology and End Results (SEER), nomogram, predictive model, surveillance
1. Introduction
Malignant melanoma is one of the most important and most severe types of skin tumors and is extremely aggressive.[1] Its incidence has increased rapidly over the past 50 years.[2,3] It tends to metastasize early in the progression of the disease. Many patients have metastasized at the time of diagnosis or experienced recurrence after initial curative treatment.[4,5] The most common sites of metastases from melanoma are the skin and subcutaneous tissue, followed by distant metastases in the lung, liver, bone and brain.[6,7] Despite tremendous advances in the treatment of advanced malignant melanoma with molecularly targeted therapies and immunotherapy,[8,9] the prognosis for advanced metastatic melanoma remains extremely poor, with a 5-year survival rate of less than 10% and a median overall survival (OS) of approximately 7.5 months.[5]
Clinicians currently assess the prognosis of melanoma primarily based on the American Joint Committee on Cancer (AJCC) 8th edition TNM staging system.[7] However, it has been well observed that despite the international staging system has been adopted, the variation in the conclusions reached by pathologists in the diagnosis of malignant melanoma is extremely high between different observers. As a result, the diagnosis of melanoma remains difficult to be accurate and consistent.[7,9] Therefore, the prognosis of melanoma is difficult to predict accurately by relying on the TNM staging alone.
Nomograms have been widely used in the medical field to predict the survival rate of tumor patients. Compared with the traditional TNM staging system, it can incorporate more possible prognostic factors and is more accurate in prognostic prediction.
Based on some previous studies,[10–12] we defined early death as death within 3 months of the first diagnosis. Therefore, in this study, we extracted and analyzed data from the Surveillance, Epidemiology, and End Results (SEER) database for patients diagnosed with advanced melanoma (AM) between 2010 and 2015. Risk factors associated with early death in AM patients were identified and developed nomograms to assess the prognosis of these patients more accurately. It helps oncologists to identify and intervene early and provide more individualized treatment plans.
2. Method
2.1. Patients selection
We conducted a retrospective cohort study using data from the SEER database, including demographic, clinicopathological, and survival data of cancer patients. This study used SEER*Stat software (www.seer.cancer.gov, software version 8.4.0) to extract data from the SEER database for patients diagnosed with AM from 2010 to 2015, applying the International Classification of Diseases for Oncology, and 3rd edition (ICD-O-3) criteria, to identify patients with melanoma based on the primary site (ICD-10/C43-C44), whose diagnosis was also confirmed histologically.
The exclusion criteria were as follows: patients without definitive histological examination; Unknown age, race, marital status, and stage; Liver, brain, lung, bone metastases unknown; Treatment unknown; and Missing follow-up data. The specific patient selection flow chart is shown in Figure 1. Based on the inclusion and exclusion criteria, a total of 2138 patients were ultimately included in the study and randomly assigned by computer to the training group cohort and the validation cohort in a 7:3 ratio.
Figure 1.
Flow chart of patient selection for included studies.
2.2. Statistical analysis
The endpoint of this study was the survival time, and patients who died within 3 months after the initial diagnosis were included in this study. All included variables except the survival time are described as numbers and percentages (N, %). As described above, these patients were randomly assigned to the training and validation cohorts. The training cohort was used to identify prognostic factors associated with early death in patients with AM and to develop nomograms. The nomogram is a common tool for prognostic evaluation of tumors, which transforms complex regression equations into visual graphs, making the results of predictive models more readable and facilitating patient assessment. With Nomogram, developers can provide an accurate digital probability of survival or risk for each patient, which can assist physicians in decision making and develop more personalized treatment strategies.
Variables that were significantly associated with early death in the multifactorial logistic analysis were included in the nomogram, and we developed nomograms for cancer-specific early death and all-cause early death, respectively. The validation cohort was used to verify the accuracy and efficiency of the model. Pearson chi-square tests were used to explore the differences between this training and validation cohorts.
Statistical analysis of the data was performed using R software (version 4.2.2) and SPSS statistics software (version 27, IBM Corp, Armonk, NY), and a two-sided P value < .05 was considered statistically significant.
After the model was built, to assess the reliability of the nomogram, we calculated the area under the curve (AUC) by ROC curves in both the training and validation cohorts to predict model efficacy, applied calibration curves to assess the model’s agreement with reality, and finally, tested the net benefit of interventions based on model results by decision curve analysis, which assesses the value of clinical application. The calibration curves represent the agreement between observed and predicted probabilities through 1000 resampling bootstraps.
3. Result
3.1. Demographic and clinical features of AM
Based on the inclusion and exclusion criteria, 2138 patients diagnosed with AM from 2010 to 2015 in the SEER database were included in this study (Fig. 1), and were randomized in a ratio of 7:3 into a training cohort and a validation cohort. Demographic and clinicopathological characteristics of patients in the training cohort (n = 1498) and validation cohort (n = 640) are presented in Table 1. Overall, there was a predominance of patients over 40 years old, and 70.2% were male patients (n = 1500). More than 90% of the patients (n = 2060) were white. Among the primary sites, 63.0% of the patients had melanoma of unknown origin (n = 1348), 32.9% had cutaneous melanoma (n = 703), and a very small percentage had primary sites in the external ear or choroid and other rare sites. The majority of patients had the pathologic type of malignant melanoma, which accounted for 83.3% of the total (n = 1781), nodular melanoma accounted for 8.9% (n = 191), and the rest of the pathologic types were less common. Most of these patients (73.2%) did not undergo surgery (n = 1564), while 42.6% of them received radiotherapy and 30.6% received chemotherapy. The percentage of patients with metastatic lesions of the liver, lungs, bone, and brain were: 29.7%, 50.7%, 25.1%, and 41.1%, respectively.
Table 1.
Baseline features for training and validation cohorts.
| Characteristics | Number of patients (%) | P value | ||
|---|---|---|---|---|
| Overall (n = 2138) | Training cohort (n = 1498) | Validation cohort (n = 640) | ||
| Age | ||||
| <40 | 135 (6.3) | 90 (6.0) | 45 (7.0) | .619 |
| 40–65 | 1024 (47.9) | 724 (48.3) | 300 (46.9) | |
| >65 | 979 (45.8) | 684 (45.7) | 295 (46.1) | |
| Sex | ||||
| Male | 1500 (70.2) | 1042 (69.6) | 458 (71.6) | .381 |
| Female | 638 (29.8) | 456 (30.4) | 182 (28.4) | |
| Race | ||||
| White | 2060 (96.4) | 1439 (96.1) | 621 (97.0) | .516 |
| Black | 30 (1.4) | 22 (1.5) | 8 (1.2) | |
| Other | 48 (2.2) | 37 (2.5) | 11 (1.7) | |
| Marital | ||||
| Married | 1157 (54.1) | 796 (53.1) | 361 (56.4) | .18 |
| Unmarried | 981 (45.9) | 702 (46.9) | 279 (43.6) | |
| Primary site | ||||
| Skin, NOS | 1348 (63.0) | 957 (63.9) | 391 (61.1) | .795 |
| Skin other/unspecific parts of face | 48 (2.2) | 32 (2.1) | 16 (2.5) | |
| Skin of scalp and neck | 121 (5.7) | 88 (5.9) | 33 (5.2) | |
| Skin of trunk | 272 (12.7) | 183 (12.2) | 89 (13.9) | |
| Skin of upper limb and shoulder | 113 (5.3) | 82 (5.5) | 31 (4.8) | |
| Skin of lower limb and hip | 149 (7.0) | 97 (6.5) | 52 (8.1) | |
| Choroid | 23 (1.1) | 16 (1.1) | 7 (1.1) | |
| External ear | 17 (0.8) | 11 (0.7) | 6 (0.9) | |
| Other rare sites | 47 (2.2) | 32 (2.1) | 15 (2.3) | |
| Melanoma type | ||||
| Malignant melanoma, NOS | 1781 (83.3) | 1256 (83.8) | 525 (82.0) | .192 |
| Nodular melanoma | 191 (8.9) | 126 (8.4) | 65 (10.2) | |
| Superficial spreading melanoma | 40 (1.9) | 29 (1.9) | 11 (1.7) | |
| Spindle cell melanoma, NOS | 36 (1.7) | 20 (1.3) | 16 (2.5) | |
| Amelanotic melanoma | 28 (1.3) | 23 (1.5) | 5 (0.8) | |
| Other rare types | 62 (2.9) | 44 (2.9) | 18 (2.8) | |
| T Stage | ||||
| T0 | 939 (43.9) | 661 (44.1) | 278 (43.4) | .451 |
| T1–T2 | 164 (7.7) | 123 (8.2) | 41 (6.4) | |
| T3–T4 | 389 (18.2) | 266 (17.8) | 123 (19.2) | |
| TX | 646 (30.2) | 448 (29.9) | 198 (30.9) | |
| N Stage | ||||
| N0 | 828 (38.7) | 572 (38.2) | 256 (40.0) | .143 |
| N1 | 416 (19.5) | 276 (18.4) | 140 (21.9) | |
| N2 | 86 (4.0) | 63 (4.2) | 23 (3.6) | |
| N3 | 145 (6.8) | 101 (6.7) | 44 (6.9) | |
| NX | 663 (31.0) | 486 (32.4) | 177 (27.7) | |
| Surgery | ||||
| No | 1564 (73.2) | 1110 (74.1) | 454 (70.9) | .145 |
| Yes | 574 (26.8) | 388 (25.9) | 186 (29.1) | |
| Radiation | ||||
| No | 1227 (57.4) | 878 (58.6) | 349 (54.5) | .089 |
| Yes | 911 (42.6) | 620 (41.4) | 291 (45.5) | |
| Chemotherapy | ||||
| No | 1484 (69.4) | 1050 (70.1) | 434 (67.8) | .319 |
| Yes | 654 (30.6) | 448 (29.9) | 206 (32.2) | |
| Bone metastasis | ||||
| No | 1602 (74.9) | 1133 (75.6) | 469 (73.3) | .273 |
| Yes | 536 (25.1) | 365 (24.4) | 171 (26.7) | |
| Brain metastasis | ||||
| No | 1260 (58.9) | 872 (58.2) | 388 (60.6) | .322 |
| Yes | 878 (41.1) | 626 (41.8) | 252 (39.4) | |
| Liver metastasis | ||||
| No | 1504 (70.3) | 1051 (70.2) | 453 (70.8) | .813 |
| Yes | 634 (29.7) | 447 (29.8) | 187 (29.2) | |
| Lung metastasis | ||||
| No | 1053 (49.3) | 724 (48.3) | 329 (51.4) | .209 |
| Yes | 1085 (50.7) | 774 (51.7) | 311 (48.6) | |
3.2. Mortality of early death
Among the 2138 patients with AM, 811 (37.9%) died early. Out of these, 767 patients died from cancer-specific causes, while 44 patients died from non-cancer-specific causes. Early mortality increased progressively with age and was much higher in men than in women. The vast majority of patients who died early did not receive surgery (n = 691, 85.2%) or chemotherapy (n = 668, 82.4%), and the study also showed that patients with lung and brain metastases were more likely to die early. Table 2 provides a detailed description of the demographic and clinicopathologic characteristics of patients with AM, distinguishing between those who experienced early death and those who did not.
Table 2.
Baseline characteristics of all-cause early death and cancer-specific early death in patients with advanced melanoma.
| Characteristics | Number of patients (%) | Cancer-specific early death (n = 767) | ||
|---|---|---|---|---|
| Overall (n = 2138) | No early death (n = 1327) | All-cause early death (n = 811) | ||
| Age | ||||
| <40 | 135 (6.3) | 94 (7.1) | 41 (5.1) | 40 (5.2) |
| 40–65 | 1024 (47.9) | 675 (50.9) | 349 (43.0) | 336 (43.8) |
| >65 | 979 (45.8) | 558 (42.0) | 421 (51.9) | 391 (51.0) |
| Sex | ||||
| Male | 1500 (70.2) | 923 (69.6) | 577 (71.1) | 546 (71.2) |
| Female | 638 (29.8) | 404 30.4) | 234 (28.9) | 221 (28.8) |
| Race | ||||
| White | 2060 (96.4) | 1274 (96.0) | 786 (96.9) | 743 (96.9) |
| Black | 30 (1.4) | 20 (1.5) | 10 (1.2) | 10 (1.3) |
| Other | 48 (2.2) | 33 (2.5) | 15 (1.8) | 14 (1.8) |
| Marital | ||||
| Married | 1157 (54.1) | 764 (57.6) | 393 (48.5) | 373 (48.6) |
| Unmarried | 981 (45.9) | 563 (42.4) | 418 (51.5) | 394 (51.4) |
| Primary site | ||||
| Skin, NOS | 1348 (63.0) | 744 (56.1) | 604 (74.5) | 572 (74.6) |
| Skin other/unspecific parts of face | 48 (2.2) | 35 (2.6) | 13 (1.6) | 12 (1.6) |
| Skin of scalp and neck | 121 (5.7) | 97 (7.3) | 24 (3.0) | 24 (3.1) |
| Skin of trunk | 272 (12.7) | 186 (14.0) | 86 (10.6) | 77 (10.0) |
| Skin of upper limb and shoulder | 113 (5.3) | 76 (5.7) | 37 (4.6) | 35 (4.6) |
| Skin of lower limb and hip | 149 (7.0) | 118 (8.9) | 31 (3.8) | 31 (4.0) |
| Choroid | 23 (1.1) | 20 (1.5) | 3 (0.4) | 3 (0.4) |
| External ear | 17 (0.8) | 14 (1.1) | 3 (0.4) | 3 (0.4) |
| Other rare sites | 47 (2.2) | 37 (2.8) | 10 (1.2) | 10 (1.3) |
| Melanoma type | ||||
| Malignant melanoma, NOS | 1781 (83.3) | 1055 (79.5) | 726 (89.5) | 684 (89.2) |
| Nodular melanoma | 191 (8.9) | 150 (11.3) | 41 (5.1) | 41 (5.3) |
| Superficial spreading melanoma | 40 (1.9) | 32 (2.4) | 8 (1.0) | 7 (0.9) |
| Spindle cell melanoma, NOS | 36 (1.7) | 25 (1.9) | 11 (1.4) | 11 (1.4) |
| Amelanotic melanoma | 28 (1.3) | 14 (1.1) | 14 (1.7) | 13 (1.7) |
| Other rare types | 62 (2.9) | 51 (3.8) | 11 (1.4) | 11 (1.4) |
| T Stage | ||||
| T0 | 939 (43.9) | 547 (41.2) | 392 (48.3) | 373 (48.6) |
| T1–T2 | 164 (7.7) | 127 (9.6) | 37 (4.6) | 36 (4.7) |
| T3–T4 | 389 (18.2) | 292 (22.0) | 97 (12.0) | 93 (12.1) |
| TX | 646 (30.2) | 361 (27.2) | 285 (35.1) | 265 (34.6) |
| N Stage | ||||
| N0 | 828 (38.7) | 538 (40.5) | 290 (35.8) | 273 (35.6) |
| N1 | 416 (19.5) | 282 (21.3) | 144 (17.8) | 140 (18.3) |
| N2 | 86 (4.0) | 72 (5.4) | 14 (1.7) | 14 (1.8) |
| N3 | 145 (6.8) | 108 (8.1) | 37 (4.6) | 32 (4.2) |
| NX | 663 (31.0) | 337 (25.4) | 326 (40.2) | 308 (40.2) |
| Surgery | ||||
| No | 1564 (73.2) | 873 (65.8) | 691 (85.2) | 651 (84.9) |
| Yes | 574 (26.8) | 454 (34.2) | 120 (14.8) | 116 (15.1) |
| Radiation | ||||
| No | 1227 (57.4) | 738 (55.6) | 489 (60.3) | 455 (59.3) |
| Yes | 911 (42.6) | 589 (44.4) | 322 (39.7) | 312 (40.7) |
| Chemotherapy | ||||
| No | 1484 (69.4) | 816 (61.5) | 668 (82.4) | 628 (81.9) |
| Yes | 654 (30.6) | 511 (38.5) | 143 (17.6) | 139 (18.1) |
| Bone metastasis | ||||
| No | 1602 (74.9) | 1032 (77.8) | 570 (70.3) | 542 (70.7) |
| Yes | 536 (25.1) | 295 (22.2) | 241 (29.7) | 225 (29.3) |
| Brain metastasis | ||||
| No | 1260 (58.9) | 862 (65.0) | 398 (49.1) | 373 (48.6) |
| Yes | 878 (41.1) | 465 (35.0) | 413 (50.9) | 394 (51.4) |
| Liver metastasis | ||||
| No | 1504 (70.3) | 1017 (76.6) | 487 (60.0) | 455 (59.3) |
| Yes | 634 (29.7) | 310 (23.4) | 324 (40.0) | 312 (40.7) |
| Lung metastasis | ||||
| No | 1053 (49.3) | 718 (54.1) | 335 (41.3) | 314 (40.9) |
| Yes | 1085 (50.7) | 609 (45.9) | 476 (58.7) | 453 (59.1) |
3.3. Identification of early death prognostic factor
Risk factors associated with early death in the training cohort were first analyzed by univariate logistic regression. The results showed that age, marital status, primary site, type of pathology, T stage, N stage, surgery, chemotherapy, bone metastasis, lung metastasis, liver metastasis, brain metastasis, and marital status were all significant risk factors for all-cause early death. However, among them, age was not a risk factor for cancer-specific early death. A more detailed presentation of the results of the univariate analysis can be seen in Table 3. Significant risk factors identified in the univariate logistic regression analysis were included in the multivariate logistic analysis. The multivariate analysis revealed that all of the above significant risk factors, except age and T-stage, were independent risk factors for predicting all-cause early death and cancer-specific early death in patients with AM. More detailed information can be seen in Table 4.
Table 3.
Univariate logistic regression analysis of the training cohort.
| Variable | All-cause early death | Cancer-specific early death | ||||
|---|---|---|---|---|---|---|
| OR | 95% CI | P value | OR | 95% CI | P value | |
| Age | ||||||
| <40 | 1.1 | 0.69–1.75 | .701 | 1.08 | 0.67–1.73 | .757 |
| 40–65 | 1.7 | 1.07–2.72 | .026 | 1.55 | 0.97–2.49 | .067 |
| >65 | 1 (ref) | 1 (ref) | ||||
| Sex | ||||||
| Male | 1 (ref) | 1 (ref) | ||||
| Female | 0.95 | 0.76–1.2 | .69 | 0.97 | 0.77–1.22 | .77 |
| Race | ||||||
| White | 1 (ref) | 1 (ref) | ||||
| Black | 1.08 | 0.46–2.54 | .86 | 1.2 | 0.51–2.82 | .679 |
| Other | 0.75 | 0.37–1.5 | .416 | 0.73 | 0.36–1.49 | .391 |
| Marital | ||||||
| Married | 1 (ref) | 1 (ref) | ||||
| Unmarried | 1.4 | 1.14–1.72 | .002 | 1.35 | 1.09–1.66 | .006 |
| Primary site | ||||||
| Skin, NOS | 1 (ref) | 1 (ref) | ||||
| Skin other/unspecific parts of face | 0.54 | 0.25–1.16 | .115 | 0.52 | 0.24–1.14 | .103 |
| Skin of scalp and neck | 0.27 | 0.15–0.46 | <.001 | 0.3 | 0.17–0.52 | <.001 |
| Skin of trunk | 0.64 | 0.46–0.89 | .009 | 0.6 | 0.43–0.85 | .003 |
| Skin of upper limb and shoulder | 0.59 | 0.36–0.95 | .029 | 0.59 | 0.36–0.95 | .031 |
| Skin of lower limb and hip | 0.31 | 0.19–0.52 | <.001 | 0.35 | 0.21–0.58 | <.001 |
| Choroid | 0.08 | 0.01–0.61 | .014 | 0.09 | 0.01–0.68 | .019 |
| External ear | 0.27 | 0.06–1.24 | .091 | 0.3 | 0.06–1.38 | .121 |
| Other rare sites | 0.33 | 0.14–0.78 | .011 | 0.37 | 0.16–0.87 | .023 |
| Melanoma type | ||||||
| Malignant melanoma, NOS | 1 (ref) | 1 (ref) | ||||
| Nodular melanoma | 0.46 | 0.3–0.7 | <.001 | 0.51 | 0.34–0.78 | .002 |
| Superficial spreading melanoma | 0.45 | 0.19–1.06 | .066 | 0.41 | 0.17–1.02 | .054 |
| Spindle cell melanoma, NOS | 0.47 | 0.17–1.3 | .145 | 0.52 | 0.19–1.45 | .214 |
| Amelanotic melanoma | 1.08 | 0.47–2.49 | .853 | 1.01 | 0.43–2.36 | .978 |
| Other rare types | 0.31 | 0.14–0.68 | .003 | 0.35 | 0.16–0.76 | .008 |
| T Stage | ||||||
| T0 | 1 (ref) | 1 (ref) | ||||
| T1–T2 | 0.34 | 0.22–0.55 | <.001 | 0.36 | 0.22–0.57 | <.001 |
| T3–T4 | 0.51 | 0.38–0.7 | <.001 | 0.52 | 0.38–0.71 | <.001 |
| TX | 1.13 | 0.89–1.44 | .319 | 1.06 | 0.83–1.35 | .654 |
| N Stage | ||||||
| N0 | 1 (ref) | 1 (ref) | ||||
| N1 | 1.06 | 0.78–1.43 | .718 | 1.09 | 0.8–1.47 | .58 |
| N2 | 0.27 | 0.13–0.58 | .001 | 0.3 | 0.14–0.64 | .002 |
| N3 | 0.68 | 0.42–1.09 | .108 | 0.6 | 0.37–0.99 | .045 |
| NX | 1.94 | 1.51–2.48 | <.001 | 1.87 | 1.45–2.39 | <.001 |
| Surgery | ||||||
| No | 1 (ref) | 1 (ref) | ||||
| Yes | 0.34 | 0.26–0.44 | <.001 | 0.36 | 0.28–0.48 | <.001 |
| Radiation | ||||||
| No | 1 (ref) | 1 (ref) | ||||
| Yes | 0.82 | 0.66–1.01 | .063 | 0.88 | 0.71–1.09 | .257 |
| Chemotherapy | ||||||
| No | 1 (ref) | 1 (ref) | ||||
| Yes | 0.32 | 0.25–0.41 | <.001 | 0.35 | 0.27–0.45 | <.001 |
| Bone metastasis | ||||||
| No | 1 (ref) | 1 (ref) | ||||
| Yes | 1.48 | 1.16–1.87 | .001 | 1.42 | 1.11–1.8 | .005 |
| Brain metastasis | ||||||
| No | 1 (ref) | 1 (ref) | ||||
| Yes | 1.89 | 1.53–2.34 | <.001 | 1.9 | 1.53–2.35 | <.001 |
| Liver metastasis | ||||||
| No | 1 (ref) | 1 (ref) | ||||
| Yes | 1.94 | 1.55–2.42 | <.001 | 1.98 | 1.58–2.48 | <.001 |
| Lung metastasis | ||||||
| No | 1 (ref) | 1 (ref) | ||||
| Yes | 1.76 | 1.43–2.17 | <.001 | 1.74 | 1.4–2.15 | <.001 |
Table 4.
Multivariate logistic regression analysis of the training cohort.
| Variable | All-cause early death | Cancer-specific early death | ||||
|---|---|---|---|---|---|---|
| OR | 95% CI | P value | OR | 95% CI | P value | |
| Age | ||||||
| <40 | 1 (ref) | NA | ||||
| 40–65 | 0.88 | 0.51–1.5 | .636 | |||
| >65 | 1.41 | 0.82–2.44 | .212 | |||
| Marital | ||||||
| Married | 1 (ref) | 1 (ref) | ||||
| Unmarried | 1.48 | 1.16–1.88 | .002 | 1.37 | 1.08–1.73 | .01 |
| Primary site | ||||||
| Skin, NOS | 1 (ref) | 1 (ref) | ||||
| Skin other/unspecific parts of face | 1.42 | 0.56–3.59 | .461 | 1.23 | 0.48–3.13 | .665 |
| Skin of scalp and neck | 0.47 | 0.23–0.97 | .04 | 0.53 | 0.26–1.08 | .078 |
| Skin of trunk | 1.16 | 0.68–1.98 | .579 | 0.98 | 0.58–1.66 | .938 |
| Skin of upper limb and shoulder | 1.24 | 0.64–2.38 | .522 | 1.15 | 0.6–2.21 | .667 |
| Skin of lower limb and hip | 0.6 | 0.31–1.18 | .14 | 0.66 | 0.34–1.28 | .219 |
| Choroid | 0.09 | 0.01–0.76 | .027 | 0.1 | 0.01–0.82 | .032 |
| External ear | 0.59 | 0.1–3.58 | .567 | 0.67 | 0.12–3.83 | .648 |
| Other rare sites | 0.42 | 0.15–1.13 | .086 | 0.5 | 0.19–1.34 | .168 |
| Melanoma type | ||||||
| Malignant melanoma, NOS | 1 (ref) | 1 (ref) | ||||
| Nodular melanoma | 0.91 | 0.51–1.62 | .757 | 0.97 | 0.55–1.71 | .911 |
| Superficial spreading melanoma | 0.76 | 0.28–2.09 | .594 | 0.67 | 0.24–1.88 | .448 |
| Spindle cell melanoma, NOS | 0.54 | 0.17–1.68 | .286 | 0.66 | 0.22–2.01 | .468 |
| Amelanotic melanoma | 1.09 | 0.44–2.66 | .857 | 1.02 | 0.41–2.54 | .962 |
| Other rare types | 0.54 | 0.23–1.3 | .171 | 0.64 | 0.27–1.51 | .31 |
| T stage | ||||||
| T0 | 1 (ref) | 1 (ref) | ||||
| T1–T2 | 1.08 | 0.53–2.22 | .824 | 1.09 | 0.53–2.21 | .82 |
| T3–T4 | 1.46 | 0.78–2.71 | .236 | 1.44 | 0.78–2.66 | .249 |
| TX | 1.22 | 0.91–1.63 | .19 | 1.16 | 0.87–1.55 | .317 |
| N Stage | ||||||
| N0 | 1 (ref) | 1 (ref) | ||||
| N1 | 1.29 | 0.91–1.82 | .147 | 1.27 | 0.9–1.79 | .168 |
| N2 | 0.39 | 0.17–0.9 | .027 | 0.4 | 0.18–0.91 | .03 |
| N3 | 1.16 | 0.66–2.04 | .613 | 0.95 | 0.54–1.7 | .874 |
| NX | 1.77 | 1.33–2.35 | 0 | 1.67 | 1.26–2.21 | 0 |
| Surgery | ||||||
| No | 1 (ref) | 1 (ref) | ||||
| Yes | 0.42 | 0.27–0.66 | 0 | 0.48 | 0.31–0.75 | .001 |
| Chemotherapy | ||||||
| No | 1 (ref) | 1 (ref) | ||||
| Yes | 0.27 | 0.2–0.36 | 0 | 0.29 | 0.22–0.38 | 0 |
| Bone metastasis | ||||||
| No | 1 (ref) | 1 (ref) | ||||
| Yes | 1.5 | 1.13–2 | .005 | 1.33 | 1–1.77 | .047 |
| Brain metastasis | ||||||
| No | 1 (ref) | 1 (ref) | ||||
| Yes | 2 | 1.56–2.55 | 0 | 1.91 | 1.5–2.43 | 0 |
| Liver metastasis | ||||||
| No | 1 (ref) | 1 (ref) | ||||
| Yes | 2.23 | 1.7–2.92 | 0 | 2.21 | 1.69–2.88 | 0 |
| Lung metastasis | ||||||
| No | 1 (ref) | 1 (ref) | ||||
| Yes | 1.45 | 1.14–1.84 | .003 | 1.46 | 1.15–1.86 | .002 |
3.4. Nomogram construction and validation
Based on the independent risk factors identified by multivariate logistic regression, we constructed 2 independent nomograms to predict the risk of all-cause early death (Fig. 2A) and cancer-specific early death (Fig. 2B) in patients with AM, respectively. Each influential factor in the model was assigned a score for each value level taken according to the degree of contribution of each factor to the outcome variable (magnitude of the regression coefficient), and then the individual scores were summed to obtain the total score. The probabilities of all-cause early death and cancer-specific early death ranged from 0.10 to 0.90, and the total score for most patients ranged from 150 to 350. From the nomogram, it was found that chemotherapy and N-stage primary site had a good prognostic value for predicting early death.
Figure 2.
Nomograms for predicting all-cause (A) and cancer-specific early death (B) in advanced melanoma patients.
The ROC curves of the nomogram used to assess overall and cancer-specific early death are shown in Figure 3. The AUC for overall early death in the training group was 0.751 (Fig. 3A; 95% CI: 0.726–0.776), while the AUC for cancer-specific early death was 0.794 (Fig. 3B; 95% CI: 0.714–0.765). The AUC for overall early death in the validation group was 0.759 (Fig. 3C; 95% CI: 0.722–0.797) and the AUC for cancer-specific early death was 0.757 (Fig. 3D; 95% CI: 0.718–0.780), respectively. The calibration plots of the model showed that the calibration curves fit the diagonal line relatively well, indicating that the predicted early death was highly consistent with the actual outcome (Fig. 4), while the decision curve analysis showed that the model has a high value for clinical application (Fig. 5). In the training group, the sensitivity, specificity, and odds ratio for all-cause early death were 0.732, 0.685, and 5.959, respectively. For cancer-specific early death, they were 0.729, 0.669, and 5.443. While in the validation set, the sensitivity, specificity, and odds ratio for all-cause early death were 0.716, 0.655, and 5.573, respectively. For cancer-specific early death, the data were 0.695, 0.705, and 5.451.
Figure 3.
ROC for discrimination of nomograms in predicting all-cause and cancer-specific early death in the training cohort (A, B) and the validation cohort (C, D). AUC indicates area under the curve, with higher values indicating higher prediction accuracy. AUC = area under the ROC curves, ROC = operating characteristic curves.
Figure 4.
Calibration curves for assessing the calibration of the nomogram in predicting all-cause early death and cancer-specific early death in the training cohort (A, B) and the validation cohort (C, D).
Figure 5.
DCA for the nomograms in predicting all-cause early death and cancer-specific early death in the training cohort (A, B) and the validation cohort (C, D). DCA = decision curve analysis.
4. Discussion
The prognosis for AM is extremely poor, with a 5-year survival rate of less than 10%.[5,13] With the successful development of targeted and immunotherapy for patients with metastatic melanoma, there has been a significant improvement in quality of life and overall survival,[14,15] but mortality rate for patients with AM remained high, with more than one-third of the patients in this study dying within 3 months of diagnosis. However, few studies have assessed early death in AM, so it is essential to explore the risk factors that lead to early death in order to increase patient survival as much as possible. To our knowledge, this is the first study to identify risk factors and construct a column chart for identifying early death in AM.
Melanoma is a highly heterogeneous disease, and despite clinical staging guidelines, the prognosis of metastatic melanoma remains variable and unpredictable. Several current studies suggest that nomograms can be used to predict the prognosis of patients with melanoma. Li et al[16] developed a predictive nomogram for cancer-specific death (CSD) in metastatic melanoma patients and a corresponding risk classification system was developed, with training cohort AUC values of 0.706, 0.700, and 0.706, respectively. A study by Du et al[17] combined a predictive model of the ki-67 clinical factor for predicting the prognosis of extremity melanoma.
The data in this study suggest that marital status, primary site, histological type, N stage, surgery, chemotherapy, bone, liver, lung and brain metastases are associated with the risk of all-cause early death and cancer-specific early death. The results of univariate and multivariate analyses demonstrate the significant prognostic value of the TNM staging for metastatic melanoma. A study conducted by Cozzolino et al[18] showed that tumor site, tissue type, distant metastases, were significantly associated with overall survival, which is in agreement with our findings. In this study, we incorporated both TNM staging, clinical and pathological features, and identified a total of 10 independent risk factors to construct nomograms for predicting early death in patients with AM. The area under the curve (AUC) for overall early death in the training group was 0.751, while the AUC for cancer-specific early death was 0.794. In comparison, a previous study on the prognostic accuracy of the AJCC staging system, version 8, with a total cohort of 1462 patients, reported an AUC of 0.74.[19]
In patients with advanced disease, aggressive treatment is an important means of reducing early mortality. Both surgery and chemotherapy can reduce the risk of early death. Appropriate surgical treatment is essential for the diagnosis, staging, and treatment of melanoma. First, biopsy surgery provides clarity in the diagnosis and staging of melanoma and also helps prevent surgical site errors.[20] Second, enlarged excision of the primary melanoma site can reduce the risk of local recurrence.[21,22] In patients with a single metastasis or a few metastases, surgical excision of all metastases can occasionally lead to lasting improvement.[23,24] However, in patients with mucosal melanoma, local recurrence is common even with aggressive surgical treatment, and most patients die from distant lesions rather than from uncontrolled local lesions.[25,26] Overall, deciding whether to operate on a patient with metastatic melanoma is a complex and often difficult process, regardless of whether the aim is curative or palliative treatment.
Metastases lead to a significant risk of early death, with over 95% of early deaths being metastatic melanoma and leading to cancer-specific death. Lung is the most common site of distant metastases from melanoma, accounting for as much as 40% of stage IV disease, and lung metastasectomy may improve survival,[27] In today’s era of targeted and immune checkpoint blockade therapy, the choice of lung metastasectomy should also take into account the patient’s outcome with systemic therapy.[28–30] Brain metastases are a common complication for patients with regionally advanced and metastatic melanoma, and a significant cause of complications and death. Various tools have been developed to determine the prognosis of patients with melanoma brain metastases.[31] Once patients at a high risk of premature death are identified, oncologists may be able to give more advanced treatment strategies, such as clinical trials, and closer follow-up.
The present study also has some unavoidable limitations, such as the model did not incorporate some known risk factors, such as Eastern Tumor Collaborative Group performance scores, gene mutation information, and molecular pathology indicators, which may significantly improve the efficacy of existing models. Secondly, our study was retrospective and potential selection bias may have adversely affected the conclusions. Finally, the existence of inconsistent registry staging in the SEER database[32] may have influenced our analysis of prognosis, and our nomograms were only validated internally and further external validation through larger prospective studies is needed.
5. Conclusion
In summary, we extracted clinical and pathological factors from the SEER database for patients with AM and found that more than one-third of patients with advanced melanoma died within 3 months. The study also analyzed independent risk factors associated with early death. Based on these factors, we developed and validated nomograms for all-cause early death and cancer-specific early death risk. The good performance of the nomograms suggests that these models can help clinicians identify patients with a high risk of early death and provide them with personalized treatment, thus improving their survival benefit, as well as provide an aid to clinical trial design. Certainly, further studies confirming its practical application in the management of patients with advanced melanoma are of great necessity.
Author contributions
Conceptualization: Bin Liu.
Data curation: Bin Liu, Cunli Yin, Xi Yang, Chunyu Wang.
Investigation: Cunli Yin, Xi Yang, Yingchun Lu, Chunyu Wang.
Methodology: Bin Liu, Siru Li.
Resources: Siru Li, Xi Yang, Yingchun Lu.
Software: Siru Li, Xi Yang, Yingchun Lu.
Supervision: Bin Liu, Cunli Yin, Yingchun Lu.
Validation: Siru Li, Chunyu Wang.
Writing – original draft: Siru Li.
Writing – review & editing: Siru Li.
Abbreviations:
- AM
- advanced melanoma
- AUC
- area under the ROC curves
- ICD
- International Classification of Diseases
- ROC
- operating characteristic curves
- SEER
- Surveillance, Epidemiology and End Results
Since the data published in the SEER database is a public use database and does not contain personal privacy, informed patient consent and ethical certification are not required for the data published in the SEER database.
The authors have no funding and conflicts of interest to disclose.
The datasets generated during and/or analyzed during the current study are publicly available.
How to cite this article: Li S, Yin C, Yang X, Lu Y, Wang C, Liu B. Risk factors and predictive models for early death in patients with advanced melanoma: A population-based study. Medicine 2023;102:40(e35380).
Contributor Information
Siru Li, Email: lsr770880@163.com.
Cunli Yin, Email: yclrumor@163.com.
Xi Yang, Email: yangximuse@stu.cdutcm.edu.cn.
Yingchun Lu, Email: luyingchun1115@163.com.
ChunYu Wang, Email: rebeccawan9@163.com.
References
- [1].Ahmed B, Muhammad IQ, Ghafoor S. Malignant melanoma: skin cancer-diagnosis, prevention, and treatment. Crit Rev Eukaryot Gene Expr. 2020;30:291–7. [DOI] [PubMed] [Google Scholar]
- [2].Siegel RL, Miller KD, Jemal A. Cancer statistics, 2019. CA Cancer J Clin. 2019;69:7–34. [DOI] [PubMed] [Google Scholar]
- [3].Kosary CL, Altekruse SF, Ruhl J, et al. Clinical and prognostic factors for melanoma of the skin using SEER registries: collaborative stage data collection system, version 1 and version 2. Cancer. 2014;120:3807–14. [DOI] [PubMed] [Google Scholar]
- [4].Bedrosian I, Faries MB, Guerry D, et al. Incidence of sentinel node metastasis in patients with thin primary melanoma (< or = 1 mm) with vertical growth phase. Ann Surg Oncol. 2000;7:262–7. [DOI] [PubMed] [Google Scholar]
- [5].Sundararajan S, Thida AM, Yadlapati S, et al. Metastatic Melanoma. Treasure Island, FL: StatPearls Publishing; 2022. [PubMed] [Google Scholar]
- [6].Damsky WE, Rosenbaum LE, Bosenberg M. Decoding melanoma metastasis. Cancers (Basel). 2010;3:126–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [7].Gershenwald JE, Scolyer RA, Hess KR, et al.; for members of the American Joint Committee on Cancer Melanoma Expert Panel and the International Melanoma Database and Discovery Platform. Melanoma staging: evidence-based changes in the American Joint Committee on Cancer eighth edition cancer staging manual. CA Cancer J Clin. 2017;67:472–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].Seedor RS, Orloff M. Treatment of metastatic melanoma in the elderly. Curr Oncol Rep. 2022;24:825–33. [DOI] [PubMed] [Google Scholar]
- [9].Davis LE, Shalin SC, Tackett AJ. Current state of melanoma diagnosis and treatment. Cancer Biol Ther. 2019;20:1366–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [10].Ye EZ, Ye EH, Ye RZ. DeepImageTranslator V2: Analysis of Multimodal Medical Images using Semantic Segmentation Maps Generated through Deep Learning. Cold Spring Harbor Laboratory; 2021. [Google Scholar]
- [11].Ying Z, Xiongfeng F, Lanqing W, et al. A predictive nomogram for early death of metastatic gastric cancer: a retrospective study in the SEER database and China. J Cancer. 2020;11:5527–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [12].Shen H, Deng G, Chen Q, et al. The incidence, risk factors and predictive nomograms for early death of lung cancer with synchronous brain metastasis: a retrospective study in the SEER database. BMC Cancer. 2021;21:825. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [13].Millet A, Martin AR, Ronco C, et al. Metastatic melanoma: insights into the evolution of the treatments and future challenges. Med Res Rev. 2017;37:98–148. [DOI] [PubMed] [Google Scholar]
- [14].Ancuceanu R, Neagu M. Immune based therapy for melanoma. Indian J Med Res. 2016;143:135–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Jamie G, Charlotte A. Update on immunotherapy in melanoma. Surg Oncol Clin N AM. 2015; 24:337–46. [DOI] [PubMed] [Google Scholar]
- [16].Li W, Xiao Y, Xu X, et al. A novel nomogram and risk classification system predicting the cancer-specific mortality of patients with initially diagnosed metastatic cutaneous melanoma. Ann Surg Oncol. 2021;28:3490–500. [DOI] [PubMed] [Google Scholar]
- [17].Du Y, Li C, Mao L, et al. A nomogram incorporating Ki67 to predict survival of acral melanoma. J Cancer Res Clin Oncol. 2023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [18].Cozzolino C, Buja A, Rugge M, et al. Machine learning to predict overall short-term mortality in cutaneous melanoma. Discov Oncol. 2023;14:13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Kanaki T, Stang A, Gutzmer R, et al. Impact of American Joint Committee on Cancer 8th edition classification on staging and survival of patients with melanoma. Eur J Cancer. 2019;119:18–29. [DOI] [PubMed] [Google Scholar]
- [20].Swetter SM, Tsao H, Bichakjian CK, et al. Guidelines of care for the management of primary cutaneous melanoma. J Am Acad Dermatol. 2019;80:208–50. [DOI] [PubMed] [Google Scholar]
- [21].Wheatley K, Wilson JS, Gaunt P, et al. Surgical excision margins in primary cutaneous melanoma: a meta-analysis and Bayesian probability evaluation. Cancer Treat Rev. 2016;42:73–81. [DOI] [PubMed] [Google Scholar]
- [22].Utjés D, Malmstedt J, Teras J, et al. 2-cm versus 4-cm surgical excision margins for primary cutaneous melanoma thicker than 2 mm: long-term follow-up of a multicentre, randomised trial. Lancet. 2019;394:471–7. [DOI] [PubMed] [Google Scholar]
- [23].Fife KM, Colman MH, Stevens GN, et al. Determinants of outcome in melanoma patients with cerebral metastases. J Clin Oncol. 2004;22:1293–300. [DOI] [PubMed] [Google Scholar]
- [24].Paek SH, Audu PB, Sperling MR, et al. Reevaluation of surgery for the treatment of brain metastases: review of 208 patients with single or multiple brain metastases treated at one institution with modern neurosurgical techniques. Neurosurgery. 2005;56:1021–34; discussion 1021. [PubMed] [Google Scholar]
- [25].Meleti M, René LC, de Bree R, et al. Head and neck mucosal melanoma: experience with 42 patients, with emphasis on the role of postoperative radiotherapy. Head Neck. 2008;30:1543–51. [DOI] [PubMed] [Google Scholar]
- [26].Krengli M, Masini L, Kaanders Johannes HAM, et al. Radiotherapy in the treatment of mucosal melanoma of the upper aerodigestive tract: analysis of 74 cases. A Rare Cancer Network study. Int J Radiat Oncol Biol Phys. 2006;65:751–9. [DOI] [PubMed] [Google Scholar]
- [27].Hanna TP, Chauvin C, Miao Q, et al. Clinical outcomes after pulmonary metastasectomy for melanoma: a population-based study. Ann Thorac Surg. 2008;106:1675–81. [DOI] [PubMed] [Google Scholar]
- [28].Bello DM, Panageas KS, Hollmann T, et al. Survival outcomes after metastasectomy in melanoma patients categorized by response to checkpoint blockade. Ann Surg Oncol. 2020;27:1180–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [29].Klemen ND, Shindorf ML, Sherry RM. Role of surgery in combination with immunotherapy. Surg Oncol Clin N Am. 2019;28:481–7. [DOI] [PubMed] [Google Scholar]
- [30].Ascierto PA, Long GV, Robert C, et al. Survival outcomes in patients with previously Untreated BRAF wild-type advanced melanoma treated with nivolumab therapy: three-year follow-up of a randomized phase 3 Trial. JAMA Oncol 2019; 5(2):187–194 [DOI] [PMC free article] [PubMed] [Google Scholar]
- [31].Sperduto PW, Kased N, Roberge D, et al. Summary report on the graded prognostic assessment: an accurate and facile diagnosis-specific tool to estimate survival for patients with brain metastases. J Clin Oncol. 2012;30:419–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [32].Cullison CR, Zheng DX, Levoska MA, et al. Inconsistencies in cutaneous melanoma staging Within SEER Registries. JAMA Dermatol 2018;157:727–9. [DOI] [PMC free article] [PubMed] [Google Scholar]