Competing risks models with two time scales.

Data from the 11 Surveillance, Epidemiology, and End Results tumor registries cover 14% of the U.S. population and are often used to provide national estimates of cancer incidence and survival. Cancer mortality data for the U.S. and for SEER are compared to assess the representation of SEER to the U.S. Comparisons between US and SEER cancer mortality were made for 16 of the leading causes of cancer death. Cancer site-specific comparisons were also made by race and sex. In addition, tobacco-related cancers were considered. Analyses were performed for the years when all 11 SEER registries were included in the SEER program. Poisson regression was used to estimate site-specific cancer mortality rate ratios between the U.S. and SEER. Cancer site-specific mortality rates derived from SEER data tend to under-represent the U.S. cancer mortality experience for white males and females and black males. Under-representation is observed across the majority of SEER registries, with the highest amount of under representation in Utah and New Mexico. Under-representation of SEER data compared with US data is noticeably greater among tobacco-related cancers, particularly in Utah and New Mexico. For certain cancer sites, particularly tobacco-related cancers, the SEER coverage population is not representative of the U.S. population.

Machine-learning models are increasingly used in orthopaedic oncology to predict survival outcomes for patients with osteosarcoma. Typically, these models are trained on a single data set, such as the Surveillance, Epidemiology, and End Results (SEER) or the National Cancer Database (NCDB). However, because any single database, even if it is large, may emphasize different data points and may include errors, models trained on single data sets may learn database-specific patterns rather than generalizable clinical relationships, limiting their clinical utility when applied to different patient populations. We developed separate machine-learning models using SEER and NCDB databases and (1) compared the accuracy of SEER- and NCDB-trained models in estimating 2- and 5-year overall survival when validated on their respective databases, (2) assessed which database produced a more generalizable machine-learning model (defined as one that maintains high performance when applied to unseen external data) by using the model trained on one database to externally validate the other, and (3) identified key factors contributing to prediction accuracy. From 2000 to 2018 (SEER) and 2004 to 2018 (NCDB), we identified 15,241 SEER patients and 11,643 NCDB patients with osteosarcoma. After excluding patients with tumors outside the extremities/pelvis, including unconfirmed osteosarcoma histology results (52% [7989] SEER, 22% [2537] NCDB) and those with missing metastasis, treatment, or prognosis data (20% [2974] SEER, 43% [5057] NCDB), we included 4049 patients from NCDB and 4278 patients from SEER, all with confirmed osteosarcoma. SEER provides population-based coverage with detailed staging but limited treatment information, while NCDB offers hospital-based data with comprehensive treatment details. We developed separate models for each data set, randomly splitting each into training (80%) and validation (20%) sets. This separation was crucial because it allowed us to test how well our models performed on completely new, unseen data-to test whether a model will work in real-world clinical practice. Primary outcomes included accuracy (proportion of correct predictions), area under the receiver operating characteristic curve (AUC) (discriminative ability between survival outcomes, with values > 0.8 indicating good performance), Brier score (probabilistic prediction accuracy, with values < 0.25 indicating useful models), precision (proportion of positive predictions that were correct), recall (sensitivity for identifying actual outcomes), and F1 score (harmonic mean of precision and recall). The median patient age was 22 years in the NCDB versus 17 years in SEER (p = 0.005), with similar sex distributions (56% male in NCDB, 56% male in SEER) but different racial compositions and overall survival rates (72% and 52% at 2 and 5 years, respectively, for NCDB versus 65% and 43% for SEER). Internal validation showed excellent performance: NCDB-trained models achieved an AUC of 0.93 (95% confidence interval [CI] 0.92 to 0.94) at 2 years and 0.91 (95% CI 0.90 to 0.92) at 5 years, while SEER-trained models achieved 0.90 (95% CI 0.89 to 0.91) and 0.92 (95% CI 0.91 to 0.92), respectively. These AUC values > 0.90 indicate excellent discriminative ability; the models can reliably distinguish between patients who will survive and those who will not. The small differences between NCDB and SEER models (95% CI 0.90 to 0.93) are not clinically meaningful given overlapping confidence intervals. However, external validation revealed poor transferability: NCDB models tested on SEER data achieved an AUC of 0.67 (95% CI 0.65 to 0.68) and 0.60 (95% CI 0.58 to 0.62), while SEER models tested on NCDB data achieved 0.61 (95% CI 0.59 to 0.62) and 0.56 (95% CI 0.55 to 0.58). These external validation AUC values < 0.70 indicate poor predictive performance (barely better than chance and unsuitable for clinical decision-making). This dramatic performance drop demonstrates that models cannot be reliably transferred between different healthcare databases. NCDB models prioritized treatment variables, while SEER models emphasized demographic factors, reflecting the different clinical information captured by each database and explaining why cross-database application fails. Models should be validated within the same database environment where they will be applied. These results highlight differences between the NCDB and SEER data sets, showing that models learn database-specific patterns rather than generalizable disease patterns. Cross-database application of models leads to poor predictive performance and should be avoided without revalidation. Level III, prognostic study.

SummaryBackgroundEpidemiological evidence on the influence of ultra-processed foods (UPFs) on breast cancer prognosis is scarce. No study has examined the UPF-mortality association among Black breast cancer survivors.MethodsWe examined the UPF-mortality relationships among Black women diagnosed with primary breast cancer in New Jersey between 2005 and 2019 (n = 1733), who were participants of the Women's Circle of Health Study and Women's Circle of Health Follow-Up Study. Foods and drinks consumed one year before breast cancer diagnosis were assessed during home interviews by validated food-frequency questionnaires. UPFs were classified according to their degree of processing using the standard NOVA classification system. Death outcomes were ascertained through linkage with New Jersey State Cancer Registry files. Multivariable-adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for UPF-mortality associations were estimated using Cox and competing risks models, as appropriate.FindingsAfter a median of 9.3 years of follow-up since diagnosis, 394 total deaths (206 breast cancer-related) were identified. In the multivariable-adjusted model, compared to those in the lowest tertile (median = 2.6 servings/day), women in the highest tertile (median = 8.1 servings/day) of UPF intake had statistically significantly higher breast cancer-specific mortality (HR = 1.40, 95% CI = 1.00–1.96, Ptrend = 0.02) and all-cause mortality (HR = 1.36, 95% CI = 1.06–1.74, Ptrend <0.01). Dose-response analyses revealed a J-shaped association of UPF intake with mortality outcome, with lower risk at under 4 servings/day and higher risk at greater consumption (breast cancer mortality Pfor nonlinearity = 0.01). These associations were attenuated after additional adjustment for total energy intake.InterpretationThis large investigation of UPFs and breast cancer prognosis among Black breast cancer survivors suggests that higher consumption of UPFs, which is associated with greater total energy intake, may adversely influence breast cancer prognosis among Black women, a vulnerable population facing the highest risk of breast cancer mortality in the US.FundingThis work was supported by grants from R01CA185623 (Drs. Bandera, Demissie, and Hong), P30CA072720-5929 (Dr. Bandera), R01CA100598 (Dr. Ambrosone), P01CA151135 (Drs Ambrosone, Palmer, and Olshan), R00CA267557 (Dr. Wang) from the National Cancer Institute, and funding from the American Cancer Society RSG-23-1143513-01-CTPS (Dr. Qin) and the Breast Cancer Research Foundation (Dr. Ambrosone). The New Jersey State Cancer Registry, Cancer Epidemiology Services, New Jersey Department of Health, is funded by the Surveillance, Epidemiology and End Results (SEER) Program of the National Cancer Institute under contract 75N91021D00009, the National Program of Cancer Registries (NPCR), Centers for Disease Control and Prevention under grant NU58DP007117 as well as the State of New Jersey and the Rutgers Cancer Institute.

To explore the reliability of Medicare Part A claims data for clinical and health services research related to the care of patients with cancer, the authors compared estimates of the incidence of and resection rates for cancer of the breast, colon, and lung derived from analysis of Medicare Part A data versus data from the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program. Incidence rates of breast, colon, and lung cancer estimated from Medicare Part A data were within 6% of estimates derived from SEER data. Resection rates estimated from Medicare Part A data, in contrast, were 12% to 27% lower than resection rates based on SEER data. This discrepancy is not explained by variations in practice between regions participating in versus those not participating in the SEER registries but may be due to undercoding of surgical procedures in Medicare Part A data. This analysis suggests that Medicare data can provide useful insights into the care of patients with cancer, but research regarding inpatient procedures employed in management of cancer should be based on analysis of Medicare Parts A and B data combined.

Background: Access to cancer screening, diagnosis and treatment in the United States (US) is affected by insurance status; whereas, access within a publicly-funded health care system is similar across the whole population. The aim of this study was to compare overall survival (OS) of de novo stage IV (metastatic) breast cancer (BC) pts in a Canadian province and in the US according to insurance status. Methods: All female pts 18-64 yrs of age diagnosed with de novo stage IV BC from Jan 1, 2010 through Dec 31, 2014 with available biomarker information were included. Pts diagnosed by death certificate or autopsy and pts ≥ 65 yrs were excluded due to unreliable insurance status classification in the US SEER database. The Alberta cohort (AB) was obtained from the Alberta Health Services CancerControl Breast Data Mart (BDM), a repository of information on all pts diagnosed with their first BC diagnosis from Jan 1, 2004 onwards in the province of Alberta, Canada. The U.S. cohort was obtained from the US Surveillance, Epidemiology, and End Results (SEER) program cancer database. A total of 9,604 pts from the SEER database and 294 pts from the BDM were analyzed. OS was evaluated over a 2 yr period and median and 2 yr OS were estimated. Unadjusted associations were compared using the log-rank test, and hazard ratios (HR) were estimated using the Cox proportional hazards model with US insured set as reference group. Results: Comparison of AB and US cohorts showed no differences based on age group (18-49 vs 50-64), yr of diagnosis or receipt of primary surgery. The AB cohort had a higher incidence of hormone receptor positive (HR+), similar frequency of HER2+, and a lower incidence of triple negative (TN) BC relative to the US cohort: HR+ 60.5% vs 56.4%; HER2+ 30.6% vs 28.8%; and TN 8.8% vs 14.8%, respectively [p=0.017]. The distribution of HR+, HER2+ and TN BC was consistent between the SEER insured, Medicaid and uninsured groups. AB cohort estimated 2 yr OS was 70.1%, similar to the insured group of 66.0% and significantly better than the Medicaid or uninsured pts [53.2% and 50.9%; p&amp;lt;0.0001]. Subgroup analysis based on biomarker status, surgery and age group showed similar results. Adjusting for these variables, AB OS remained similar to the insured group [HR=0.92 (0.74-1.15) p=0.474] with worse OS noted in the Medicaid and uninsured populations [HR=1.44 (1.32-1.56) and HR=1.53 (1.33-1.77) p&amp;lt;0.001, respectively]. HR+HER2+TNNo SurgerySurgery18-4950-64InsuredReference Group: HR=1.00AB0.92 (0.68-1.24) p=0.5841.15 (0.76-1.72) p=0.5090.84 (0.49-1.42) p=0.5070.91 (0.71-1.17) p=0.4660.71 (0.45-1.11) p=0.1330.84 (0.56-1.27) p=0.4080.87 (0.67-1.12) p=0.284Medicaid1.43 (1.27-1.61) p&amp;lt;0.0011.57 (1.32-1.87) p&amp;lt;0.0011.37 (1.18-1.60) p&amp;lt;0.0011.46 (1.33-1.59) p&amp;lt;0.0011.55 (1.33-1.81) p&amp;lt;0.0011.75 (1.52-2.00) p&amp;lt;0.0011.47 (1.34-1.61) p&amp;lt;0.001Uninsured1.76 (1.44-2.15) p&amp;lt;0.0011.64 (1.20-2.22) p&amp;lt;0.0011.56 (1.20-2.05) p=0.0011.56 (1.34-1.80) p&amp;lt;0.0011.59 (1.16-2.18) p=0.0041.76 (1.35-2.29) p&amp;lt;0.0011.67 (1.43-1.94) p&amp;lt;0.001 Conclusion: OS in women with de novo stage IV BC in AB was similar to US insured. AB and US insured experienced superior OS compared with US Medicaid and uninsured. Citation Format: Kornaga EN, Matutino AR, Pereira AA, Verma S, Lupichuk S. Survival in women with de novo metastatic breast cancer: Comparison of real-world evidence from publicly-funded Canadian province and the United States by insurance status [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P2-08-13.

Colon and rectal cancers are the most common kinds of cancer globally. Colon cancer is more prevalent in men than in women. Early detection increases the likelihood of survival, and treatment significantly increases the likelihood of eradicating the disease. The Surveillance, Epidemiology, and End Results (SEER) programme is an excellent source of domestic cancer statistics. SEER includes nearly 30% of the United States population, covering various races and geographic locations. The data are made public via the SEER website when a SEER limited-use data agreement form is submitted and approved. We investigate data from the SEER programme, specifically colon cancer statistics, in this study. Our objective is to create reliable colon cancer survival and conditional survival prediction algorithms. In this study, we have presented an overview of cancer diagnosis methods and the treatments used to cure cancer. This paper presents an analysis of prediction performance of multiple deep learning approaches. The performance of multiple deep learning models is thoroughly examined to discover which algorithm surpasses the others, followed by an investigation of the network's prediction accuracy. The simulation outcomes indicate that automated prediction models can predict colon cancer patient survival. Deep autoencoders displayed the best performance outcomes attaining 97% accuracy and 95% area under curve-receiver operating characteristic (AUC-ROC).

Sociodemographic disparities in survival of cutaneous angiosarcoma

To the Editor: “Twenty-Year Follow-up of the Breast Cancers Diagnosed During the Breast Cancer Detection Demonstration Project” in the May/June 1997 issue1 was an excellent paper. In Table 5 the authors present the adjusted and observed survival rates of patients with breast cancer. As a clinician, I would find it helpful to have an explanation of the terms “adjusted survival” and “observed survival.” Authors' Reply: Dr. Lawrence's question about the definitions of “adjusted survival” and “observed survival” is important. As mentioned in the “Methods” section of the article, we considered various measures of survival in our analysis of the Breast Cancer Detection Demonstration Project (BCDDP). Adjusted survival appeared to be the most appropriate measurement of outcome for this study. In this analysis, information on the cause of death was taken from the death certificate, and “adjusted survival” counted only deaths from breast cancer as “events.” Breast cancer was considered the cause of death whether it was a primary or a contributing cause. Thus, in the survival analysis, women who had breast cancer but died of other causes contributed to follow-up until the time of death, but their deaths were not counted as events. Evaluating only breast cancer death as the outcome provided a way for us to “adjust” for other causes of death in this analysis. In contrast, the observed survival presented in Table 5 counts any death regardless of cause as an event. Because one goal of our analysis was to make comparisons among women diagnosed with breast cancer at different ages, comparing observed survival by age is problematic because death rates from all causes increase as women get older. However, since some argue that the cause of death is less important than whether the patient is alive or dead, we presented the observed survival for comparison. We also wanted the analysis to consider the question “to what extent did the diagnosis of breast cancer affect the survival of women participating in the BCDDP?” To answer this question we would need to know the expected survival rates for these women if they had not been diagnosed with breast cancer. As a substitute for this information, studies such as ours often use death rates from a national population. These rates would be the basis for estimating relative survival, or the ratio of the observed survival rate to the expected survival rate for the cohort. The women who volunteered for the BCDDP were not a representative sample of the US population, however, and other analyses have shown that their age-specific death rate for all causes is much lower than national rates.1 As described in the “Methods” section, the adjusted breast cancer survival and the relative survival from the population-based Surveillance, Epidemiology and End Results (SEER) program are quite similar (57% and 53%, respectively). Because we did not have appropriate external death rates to apply to the BCDDP population to use in the calculation of relative survival, adjusted breast cancer survival rates provided the closest measure to address this question in the BCDDP.

Male Breast Cancer: An Updated Surveillance, Epidemiology, and End Results Data Analysis

6529 Background: Racial, gender, and socioeconomic disparities in cancer clinical trials participation have been well documented. The intramural program of the NCI in the National Institutes of Health (NIH) Bethesda campus conducts clinical research where participants are referred by their physician or self-referred, treated without charge and receive aid in travel expenses. We sought to describe the demographics of participants and identify potential disparities. Methods: We used the clinical research data repository of the NIH, Biomedical Translational Research Information System (BTRIS), to extract self-reported demographic data of enrollees in NCI protocols from 2005-2020. Comparisons were made with the U.S. population of patients with cancer as captured in the Surveillance, Epidemiology, and End Results (SEER) Program (2005-2018). The average household income per zip code was derived from the American Community Survey at five-year increments from 2011, 2015, and 2019. Results: We obtained data on 40,007 participants; 38,531 (95%) came from the U.S., of which 18,606 (48%) from DC, Maryland, or Virginia (DMV). Median age at consent was 54 years (range 2 – 89). Proportions of basic demographic attributes and corresponding values from SEER data were presented. Notably, Hispanic male, non-Hispanic female, and white female participants were under-represented in NCI clinical research compared to SEER data. African Americans were under-represented in the non-DMV population (5%) vs. the DMV (19%). Also, for patients from the DMV area, median income was consistently less for African American participants compared to White and Asian participants. Conclusions: The identified disparities may be partially due to the specific research activities available at the NCI, but strategies aiming at increasing the participation and representation of under-represented groups are required. [Table: see text]

The Surveillance, Epidemiology, and End Results (SEER) program of the National Cancer Institute collects data on cancer diagnoses, treatment, and survival for approximately 30% of the United States (US) population. To reflect advances in research and oncology practice, approaches to cancer control are evolving from simply enumerating the development of cancers by organ site in populations to including monitoring of cancer occurrence by histopathologic and molecular subtype, as defined by driver mutations and other alterations. SEER is an important population-based resource for understanding the implications of pathology diagnoses across demographic groups, geographic regions, and time and provides unique insights into the practice of oncology in the US that are not attainable from other sources. It provides incidence, survival, and mortality data for histopathologic cancer subtypes, and data by molecular subtyping are expanding. The program is developing systems to capture additional biomarker data, results from special populations, and expand biospecimen banking to enable cutting-edge cancer research and oncology practice. Pathology has always been central and critical to the effectiveness of SEER, and strengthening this relationship in this modern era of cancer diagnosis could be mutually beneficial. Achieving this goal requires close interactions between pathologists and the SEER program. This review provides a brief overview of SEER, focuses on facets relevant to pathology practice and research, and highlights the opportunities and challenges for pathologists to benefit from and enhance the value of SEER data.

Cancer Centers offer a comprehensive approach to cancer prevention, diagnosis, and treatment through tertiary services and by providing coordinated, personalized care. We investigated whether this approach results in significantly better patient outcomes at an NCI-designated Comprehensive Cancer Center (NCI-CCC), City of Hope (COH), versus the U.S. Surveillance, Epidemiology, and End Results (SEER) national database. Patient-level data were abstracted from the COH Cancer Registry and SEER*Stat software, respectively. The cohort included patients diagnosed with incident cancer from 2004 to 2020. Overall survival was analyzed using multivariable Cox proportional hazards models. To reduce confounding from baseline differences, propensity score matching (PSM) methods were employed in a 1:5 ratio between COH and SEER, respectively, using age group at diagnosis category, gender, race, ethnicity, stage (for solid tumors), and diagnosis year. Patients treated at COH had significantly superior overall survival (OS) compared with those in the SEER cohort across all examined cancer types after multivariable adjustment for key baseline characteristics and in the PSM analyses. Median OS (when reached) was uniformly longer in the COH cohort compared with the SEER cohort in the full population adjusted model and in the PSM analyses. For the PSM analyses: non-small cell lung cancer (NSCLC) hazard ratio (HR) 0.73 (95% confidence interval [CI]: 0.70-0.76), breast cancer HR 0.83 (95% CI: 0.78-0.86), prostate cancer HR 0.62 (95% CI: 0.58-0.65), colorectal cancer HR 0.74 (95% CI: 0.69-0.79), pancreatic cancer HR 0.75 (95% CI: 0.70-0.81), acute myeloid leukemia HR 0.79 (95% CI: 0.75-0.84), acute lymphoid leukemia HR 0.7 (95% CI: 0.69-0.87), and multiple myeloma HR 0.70 (95% CI: 0.66-0.76) (all p <0.001). In addition, notable findings included substantially lower mortality risk for patients with any of the studied advanced solid cancers, including stage IV NSCLC (HR 0.71, 95% CI: 0.67-0.75, p <0.001) and stage IV breast cancer (HR 0.78, 95% CI: 0.69-0.87, p <0.001). Our results revealed patient care at an individual NCI-CCC employing state-of-the-art therapies and modern care is associated with survival benefits for cancer patients. Patients had significantly superior OS in the COH cohort compared with the SEER cohort for most of the analyzed solid tumors assessed for each stage (I-IV) and for 3 major hematologic malignancies assessed by age group.Until randomized comparisons studies are possible and the SEER data becomes more comprehensive, the specific causes of the observed benefit and identifying patients that benefit most may be subject to considerable debate. Additional research can help determine with greater granularity the drivers of the survival benefit and highlight an opportunity to make adjusted survival outcomes more accessible to patients, complementing the data currently available because of price transparency mandates. Such efforts may help ensure that all patients benefit from precision medicine approaches.

Human breast sarcoma and human breast implantation: A time trend analysis based on SEER data (1973–1990)

Older black women are less likely to undergo mammography and are more often given a diagnosis of advanced-stage breast cancer than older white women. To investigate the extent to which previous mammography explains observed differences in cancer stage at diagnosis between older black and white women with breast cancer. Retrospective cohort study using the Linked Medicare-Tumor Registry Database. Population-based data from three geographic areas of the United States included in the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) program (Connecticut; metropolitan Atlanta, Georgia; and Seattle-Puget Sound, Washington). Black and white women 67 years of age and older in whom breast cancer was diagnosed between 1987 and 1989. Medicare claims were used to classify women according to mammography use in the 2 years before diagnosis as nonusers (no previous mammography), regular users (> or =2 mammographies done at least 10 months apart), or peri-diagnosis users (mammography done only within 3 months before diagnosis). Information on mammography use was linked with SEER data to determine cancer stage at diagnosis. Stage was classified as early (in situ or local) or late (regional or distant). Black women were more likely to not undergo mammography (odds ratio [OR], 3.00 [95% CI, 2.41 to 3.75]) and to be given a diagnosis of late-stage disease (OR, 2.49 [CI, 1.59 to 3.92]) than white women. When women were stratified by previous mammography use, the black-white difference in cancer stage occurred only among nonusers (adjusted OR, 2.54 [CI, 1.37 to 4.71]). Among regular users, cancer was diagnosed in black and white women at similar stages (adjusted OR, 1.34 [CI, 0.40 to 4.51]). In logistic modeling, previous mammography alone explained about 30% of the excess late-stage disease in black women. In a separate model, previous mammography explained 12% of the excess late-stage disease among black women after adjustment for sociodemographic and comorbidity information. Differences in breast cancer stage at diagnosis between older black and white women are related to previous mammography use. Increased regular use of mammography may result in a shift toward earlier-stage disease at diagnosis and narrow the observed differences in stage at diagnosis between older black and white women.

Anorectal melanoma (AM) is a rare type of melanoma that accounts for 0.4% to 1.6% of total malignant melanomas. The incidence of AM increases over time, and it remains highly lethal, with a 5-year survival rate of 6% to 22%. Considering the rare nature of this disease, most studies on AM comprise isolated case reports and single-center trials, which could not provide comprehensive assessment of the disease. Therefore, we conducted a population-based study by using the Surveillance, Epidemiology, and End Results (SEER) program to provide the latest and best available evidence of AM.We extracted all cases of AM registered in the SEER database from 1973 to 2011 (April 2014 release) and calculated age-adjusted incidence. Only cases with active follow-up were included to predict factors associated with prognosis. Survival outcomes were also compared among different types of surgery.We identified 640 AM cases, which consisted of 265 rectal melanoma and 375 anal melanoma. The estimated annual incidence rates of AM per 1 million population were 0.259 in males and 0.407 in females, and it increased with advanced age and over time. Tumor stage and surgical treatment were independent predictors of survival. Results implied that surgery improved the prognosis of patients with local- and regional-stage AM but could not prolong the survival of patients with distant-stage AM. Moreover, the outcome of less extensive excision was not statistically different from that of more extensive excision.This study provides an up-to-date estimation of the incidence and prognosis of AM by using SEER data. The incidence of AM continuously increases over time, despite its rarity. This disease also exhibits poor prognosis. Thus, AM must be further investigated in future studies. We also recommend surgery as the optimal treatment for local- and regional-stage AM patients but not for those with distant metastasis.