A novel staging system for adrenocortical carcinoma better predicts survival in patients with stage I/II disease
Elliot A. Asare, MD,a,b,c Tracy S. Wang, MD, MPH,” David P. Winchester, MD,b Katherine Mallin, PhD,b Electron Kebebew, MD,d and Cord Sturgeon, MD, MS,e Chicago, IL, Milwaukee, WI, and Bethesda, MD
Background. Current American Joint Committee on Cancer/International Union against Cancer (AJCC/UICC) and European Network for the Study of Adrenal Tumors staging for adrenocortical carcinoma (ACC) have not shown a survival difference between patients with stage I/II disease. This study evaluates current staging systems for survival prediction using a larger cohort and assesses whether incorporating age into ACC staging improves survival predictions.
Methods. Patients in the National Cancer Data Base (1985-2006) with a diagnosis of ACC were identified and staged using a novel TNM-A staging system: Stage I (T1/T2NOMO, age ≤55), stage II (T1/T2N0M0, age >55), stage III (T1/T2N1M0 or T3/T4N0-N1M0, any age), or stage IV (any T any NM1, any age). Differences in overall survival (OS) by stage were compared using a Cox proportional hazards model.
Results. Staging was derived for 1,579 of 3,262 patients. Median age was 54 years; mean tumor size was 11.6 cm. Using current staging, differences in 5-year OS was observed only between patients with stages II/III and III/IV ACC. With TNM-A staging, differences in 5-year OS between all stages was significant (I/II [P < . 003], II/III [P < . 0001], III/IV [P <. 0001]).
Conclusion. A staging system that incorporates patient age better predicts 5-year OS among patients with stages I/II ACC. Consideration should be given to including age in staging for ACC, because it may better inform providers about treatment and prognosis. (Surgery 2014;156:1378-86.)
From the American Joint Committee on Cancer (AJCC);” the Cancer Programs,b American College of Surgeons, Chicago, IL; the Department of Surgery,” Medical College of Wisconsin, Milwaukee, WI; the Endocrine Oncology Branch,ª Center for Cancer Research, National Cancer Institute, Bethesda, MD; and the Depart- ment of Surgery,e Feinberg School of Medicine, Northwestern University, Chicago, IL
ADRENOCORTICAL CARCINOMA (ACC) is a highly aggressive tumor with an estimated annual inci- dence of 0.5-2 cases per million per year.1-4 Owing to the rarity of ACC, studies evaluating the value and prognostication of staging systems have been limited by relatively small sample size.5-7 Numerous modifications to ACC staging have been made since McFarlane devised the first tu- mor, node, metastasis (TNM) staging system for ACC in 1958, and several competing staging
Elliot Asare is supported by the American College of Surgeons Clinical Scholars in Residence Program.
Accepted for publication August 11, 2014.
Reprint requests: Elliot A. Asare, MD, Cancer Programs, Amer- ican College of Surgeons, 633 N. St. Clair Street, 22nd Floor, Chicago, IL 60611. E-mail: easare@facs.org.
0039-6060/$ - see front matter
@ 2014 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.surg.2014.08.018
systems are currently in practice, although the American Joint Committee on Cancer (AJCC)/ International Union against Cancer (UICC) criteria for staging ACC is the most widely used (Table I).5,7-12
A validation study of the AJCC/UICC staging system did not find a statistically significant differ- ence in disease-specific survival (DSS) between stages I and II and stages II and III.5 A proposal to redefine and limit stage IV to patients with distant (M1) disease only yielded a significant difference in DSS between stages II and III but no significant difference in DSS between stages I and II.5,10,11 A validation of this proposed staging system data did not find a difference in DSS be- tween stages I and II and between stages I and III.6 To better evaluate the current competing stag- ing systems for ACC, an adequately powered study is needed, which would require a greater number of patients within each stage of disease.
| Stage | T* | N* | M* |
|---|---|---|---|
| I | T1 | N0 | M0 |
| II | T2 | N0 | M0 |
| III | T1 | N1 | M0 |
| T2 | N1 | M0 | |
| T3 | N0 | M0 | |
| IV | T3 | N1 | M0 |
| T4 | N0 | M0 | |
| T4 | N1 | M0 | |
| Any T | Any N | M1 |
*T1: ≤5 cm and confined to the adrenal gland. T2: >5 cm and confined to the adrenal gland. T3: Local invasion. T4: Invasion of adjacent struc- tures. N0: No regional lymph node metastases. N1: Positive regional lymph nodes. M0: No distant metastases. M1: Distant metastases.
This research was undertaken to evaluate the prognostic abilities of the current staging systems for ACC using the patients registered in the National Cancer Data Base (NCDB), a database that contains the largest cohort of patients with ACC in the United States. The hypothesis was that neither the current AJCC/UICC nor ENSAT stag- ing system would show differences in prognosis between stages I and II in the larger NCDB cohort and that, because age has been shown to be an independent predictor of survival in patients with ACC,13-16 incorporating age at the time of diag- nosis into ACC staging would improve survival prediction.
MATERIALS AND METHODS
Study design. We performed a retrospective review of a cohort of patients with the histologic diagnosis of ACC recorded in the NCDB from 1985 to 2006. The end date of 2006 was chosen to capture ≥5 years of follow-up for the entire cohort.
Data source and study population. The NCDB is a national cancer outcomes database jointly run by the Commission on Cancer of the American Col- lege of Surgeons and the American Cancer Society. It captures approximately 70% of all newly diag- nosed cancers in the United States from >1,500 hospitals. Variables of interest present in the NCDB are tumor size, Surveillance, Epidemiology, and End Results (SEER) summary stage, margin status, number of positive lymph nodes, grade, AJCC/UICC TNM stage (since 2010) and vital status. The 4 available grades were well differenti- ated (1), moderately differentiated (2), poorly differentiated (3), and undifferentiated (4). The SEER summary stages are localized (confined to the adrenal gland), regional (positive regional
lymph nodes or extension to adjacent tissues), and distant (metastatic) disease.3,4 Patients with In- ternational Classification of Diseases for Oncology (sec- ond and third editions) histology codes 8010, 8140, and 8370 and primary site codes C74.0- C74.9 were selected. Patients were excluded from analysis if they were <18 years at the time of diag- nosis or if data were missing for the histologic diag- nosis code, tumor size, SEER summary stage, or sex. Tumors <1 or >50 cm were considered out- liers and excluded from further analysis.
Using a combination of SEER summary stage, tumor size, and lymph node status, recoding to AJCC/UICC TNM stage and ENSAT stage was performed, using previously published recoding methods.1,4 Verification of the recoded results was performed by comparing the AJCC/UICC stage derived from the combination of SEER summary stage, size, and lymph node status with available AJCC/UICC TNM stage data since 2010. Overall survival (OS) was the calculated endpoint because the NCDB does not capture cause-specific mortality.
Modifying ACC staging to incorporate age. The variables in the database with prognostic signifi- cance for ACC outcome are shown in Table II. Dichotomized age of ≤55 or >55 years at the time of diagnosis was chosen because this threshold approximated the median age of the cohort (54 years; range, 18-90). The TNM staging was then modified to incorporate age (Table III).
Statistical analysis. The life-table method was used to generate 5-year OS curves based on AJCC stage, ENSAT stage, and the proposed new stage, using age (subsequently referred to as TNM-A). Data for some variables are reported as number (percent) or mean values ± standard deviation. Survival was calculated from the date of diagnosis to the date of death. Patients who were known to be alive at last contact were censored. Differences between stages were compared using the log-rank test. Univariable Cox proportional hazard regres- sion models were used to examine the association between AJCC/UICC stage, ENSAT stage, tumor size (≤5 or >5 cm and ≤12 or >12 cm), age (≤55 or >55 years), the TNM-A stage, margin status, and grade. Histologic grades 1 and 2, and grades 3 and 4 were combined owing to the small sample size in each grade category. A multivariable Cox propor- tional hazards model was created using each of the 3 staging systems and the variables margin status and grade. Graphic confirmation of the propor- tional hazard assumption was done. Hazard ratios (HR) and descriptive P values were determined. Logistic regression analysis was used to determine
| Predictor | Univariable analysis | Multivariable analysis | Multivariable analysis | Multivariable analysis | ||||
|---|---|---|---|---|---|---|---|---|
| HR (CI) | P value | AJCC/UICC stage, HR (CI) | P value | ENSAT stage, HR (CI) | P value | TNM-A stage, HR (CI) | P value | |
| AJCC/UICC stage | ||||||||
| I | Referent | Referent | ||||||
| II | 1.12 (0.84-1.51) | .4 | 1.10 (0.82-1.49) | .5 | ||||
| III | 1.82 (1.33-2.49) | .0002 | 1.57 (1.15-2.15) | .005 | ||||
| IV | 5.0 (3.75-6.69) | <. 0001 | 3.92 (2.92-5.27) | <. 0001 | ||||
| ENSAT stage | ||||||||
| I | Referent | Referent | ||||||
| II | 1.12 (0.84-1.51) | .4 | 1.10 (0.82-1.48) | .5 | ||||
| III | 2.10 (1.56-2.84) | <. 0001 | 1.81 (1.33-2.45) | .0001 | ||||
| IV | 5.60 (4.18-7.51) | <. 0001 | 4.29 (3.18-5.79) | <. 0001 | ||||
| Margin status | ||||||||
| R0 | Referent | Referent | Referent | Referent | ||||
| R+ | 2.22 (1.82-2.71) | <. 0001 | 1.46 (1.18-1.79) | .0004 | 1.50 (1.22-1.84) | .0002 | 1.48 (1.21-1.82) | <. 0001 |
| Grade | ||||||||
| 1/2 | Referent | Referent | Referent | Referent | ||||
| 3/4 | 2.94 (2.31-3.75) | <. 0001 | 1.91 (1.49-2.45) | <. 0001 | 1.83 (1.42-2.35) | <. 0001 | 1.82 (1.42-2.34) | <. 0001 |
| Age (y) (dichotomous) | ||||||||
| ≤55 | Referent | |||||||
| >55 | 1.51 (1.34-1.70) | <. 0001 | ||||||
| TNM-A stage | ||||||||
| I | Referent | Referent | ||||||
| II | 1.91 (1.55-2.36) | <. 0001 | 1.86 (1.50-2.29) | <. 0001 | ||||
| III | 2.62 (2.14-3.20) | <. 0001 | 2.26 (1.84-2.77) | <. 0001 | ||||
| IV | 7.0 (5.79-8.46) | <. 0001 | 5.42 (4.44-6.61) | <. 0001 | ||||
CI, Confidence interval; HR, hazard ratio; R0, microscopically negative resection margins; R+, microscopically or grossly positive resection margins.
| New stage | Criteria | Frequency (%) | Effective sample size at 5 years | 5-Year overall survival (%)* |
|---|---|---|---|---|
| I | Age ≤55; T1 or T2, NOMO | 371 (23.5) | 202 | 70 |
| II | Age >55; T1 or T2, N0M0 | 316 (20.0) | 131 | 53 |
| III | Any age; T1-T2N1M0 T3-T4N0-N1M0 | 376 (23.8) | 111 | 37 |
| IV | Any age; T1-T4N0-N1M1 | 516 (32.7) | 40 | 10 |
*Log-rank P values for comparisons of survival by stage: Stage I vs II (P= . 003); stage II vs III (P <. 0001); stage III vs IV (P <. 0001).
the likelihood of undergoing surgical resection or having a positive resection margin status in each age group. All statistical analysis was done using SAS version 9.4 (SAS Institute, Cary, NC).
RESULTS
Between 1985 and 2006, a total of 3,262 patients in the NCDB were diagnosed with ACC. Patients with sex other than male or female (n=1), <18 years of age (n = 188), or missing tumor size (n = 848) were excluded from further analysis, as were pa- tients with tumors <1 cm (n = 7) or >50 cm (n = 11) in size. The final cohort of 1,579 patients
with a known SEER summary stage and all relevant variables included 895 females (57%). Mean age of the cohort was 53.7 + 15.2 years (median, 54; range, 18-90). The mean tumor size was 11.6 ± 6.2 cm (median, 10.5; range, 1.0-50.0; Table IV).
Operative resection was performed in 1,188 patients (75.2%); 308 (19.5%) underwent no operative intervention, 79 (5.0%) underwent an unspecified operative procedure, 2 (0.13%) were recorded as having undergone destruction of the tumor, and 2 (0.13%) had an unknown status. Histologic grade was available in 470 patients
| Variable | Results |
|---|---|
| Gender, n (%) | |
| Male | 684 (43) |
| Female | 895 (57) |
| Age (y) | |
| Mean | 53.7 ± 15.2 |
| Median | 54 |
| Range | 18-90 |
| Size (cm) | |
| Mean | 11.6 ± 6.2 |
| Median | 10.5 |
| Mode | 10 |
| Range | 1-50 |
| Grade, n (%) | |
| 1 | 96 (6.1) |
| 2 | 75 (4.7) |
| 3 | 214 (13.6) |
| 4 | 85 (5.4) |
| Unknown | 1,109 (70.2) |
| Treatment, n (%) | |
| Resection | 1,188 (75.2) |
| Tumor destruction | 2 (0.13) |
| Surgery, not otherwise specified | 79 (5.0) |
| No surgery | 308 (19.5) |
| Unknown | 2 (0.13) |
| Surgical resection margin status, n (%) | |
| R0 | 794 (50.3) |
| R+ | 153 (9.7) |
| Unknown | 632 (40.0) |
(29.8%): Grade 1 included 96 patients (6.1%), grade 2, 75 (4.7%), grade 3, 214 (13.6%), and grade 4, 85 (5.4%). Of the 947 patients with known resection margins, 794 (83.8%) had microscopi- cally negative resection margins (R0) and 159 (16.2%) had microscopically positive (R1) or grossly positive (R2) resection margins.
The 5-year OS for the cohort based on AJCC/ UICC stage and ENSAT staging is shown in Table V and Fig 1. There was no difference in survival be- tween AJCC/UICC stages I and II (P = . 4). There was a significant difference between AJCC/UICC stages II and III (P < . 0001), and III and IV (P < . 0001). With the ENSAT staging system, sur- vival for stages I and II was similar to AJCC/ UICC stage (P = . 4). The difference in 5-year OS between ENSAT stages III and IV was significant (P < . 0001). In the TNM-A staging system, stages I, II, III, and IV had 5-year OS of 70%, 53%, 37%, and 10%, respectively (Table III; Fig 2). There was a significant difference in OS between patients with stages I and II (P = . 003), stages II and III (P < . 0001), and stages III and IV (P <. 0001) disease.
On univariable Cox modeling of stage using stage I as reference, an HR of 1.12 (95% CI, 0.84- 1.51; P = . 4), 1.12 (95% CI, 0.84-1.51; P= . 4), and 1.91 (95% CI, 1.55-2.36; P < . 0001) was obtained for AJCC/UICC stage II, ENSAT stage II, and TNM-A stage II, respectively (Table II). Multivari- able Cox proportional hazards model with stage, resection margin status, and grade as covariates produced HR of 1.10 (95% CI, 0.82-1.49; P = . 5), 1.10 (95% CI, 0.82-1.48; P = . 5), and 1.86 (95% CI, 1.50-2.29; P < . 0001) for AJCC/UICC stage II, ENSAT stage II, and TNM-A stage II, respec- tively (Table II). Positive resection margin status and grade (3 and 4) retained their statistical signif- icance on multivariable analysis with each staging system (Table II). A sensitivity analysis of the effect of tumor size was performed using thresholds of ≤5 or >5 cm, and ≤12 or >12 cm, which did not show significance (results not shown). The odds ratio for receipt of surgical resection in the cohort >55 years old was 0.59 (95% CI, 0.46- 0.77; P < . 0001) compared with 1.68 (95% CI, 1.31-2.17; P < . 0001) among the cohort ≤55 years old. Older patients who underwent operative resection did not have an increased odds of positive margin status (odds ratio, 0.95; 95% CI, 0.65-1.32; P = . 69).
DISCUSSION
There are several staging systems for ACC, including the AJCC/UICC staging system and the ENSAT staging system.5,7,12 Owing to the rarity of ACC, studies that have attempted to validate the staging systems have been underpowered.5,6 Therefore, the objective of this study was to use a large cohort of patients to evaluate current staging systems for their ability to predict survival in pa- tients with ACC and to determine if the incorpora- tion of age into staging system might better predict survival.
A valid staging system should have prognostic validity such that survival should be increasingly worse from stages I to IV. In addition, the variables used to derive disease stages should be widely accepted, have standard definitions, be reproduc- ible, and be readily available for collection by registrars across the wide spectrum of cancer treatment centers in the world. Although the variables T, N, and M meet these criteria, the current staging systems for ACC do not yield significant survival differences between stages I and II.5,12
Studies proposing use of the ENSAT staging system, and the subsequent validation study, used relatively small sample sizes of 416 and 573
| Staging system | TNM criteria | Frequency (%) | Effective sample size | 5-Year overall survival (%)* |
|---|---|---|---|---|
| AJCC/UICC | ||||
| I | T1N0M0 | 104 (6.6) | 56 | 68 |
| II | T2N0M0 | 583 (36.9) | 277 | 61 |
| III | T1-T2N1M0, T3N0M0, | 270 (17.1) | 94 | 43 |
| IV | T3N1M0, T4N0M0, T4N1M0, Any T, Any NM1 | 622 (39.4) | 57 | 12 |
| ENSAT | ||||
| I | T1N0M0 | 104 (6.6) | 56 | 68 |
| II | T2N0M0 | 583 (36.9) | 277 | 61 |
| III | T1-T2N1M0 T3-T4N0-N1M0 | 376 (23.8) | 111 | 37 |
| IV | T1-T4N0-N1M1 | 516 (32.7) | 40 | 10 |
*Log-rank Pvalues for comparisons of survival by stage; AJCC/UICC: stage I vs II (P= . 4); stage II vs stage III (P <. 0001); stage III vs stage IV (P <. 0001); ENSAT: stage I vs II (P= . 4); stage II vs III (P <. 0001); stage III vs IV (P <. 0001).
patients.5,6 The initial study by the ENSAT group, which examined 23, 176, 67, and 150 patients in AJCC/UICC stages I, II, III, and IV, respectively, did not find a difference in survival between AJCC/UICC stages I and II, and II and III.5 A sepa- rate study using recoded SEER data by Lughezzani et al6 examined 19, 182, 46, and 326 patients in AJCC/UICC stages I, II, III, and IV, respectively. Lughezzani et al® also did not find a statistically sig- nificant survival difference between AJCC/UICC stages I and II or between stages II and III.
In the current study, by recoding available NCDB data to TNM stage, a larger number of patients within each stage group was obtained, compared with previously published works. In contrast with previous studies,5,6 the results of this study demonstrated significant differences in 5-year OS between AJCC/UICC stages II and III. This discrepancy may be owing to the larger sam- ple size in our study, which makes a type II error less likely. However, similar to previous reports, no difference in survival was seen between AJCC/ UICC stage I and II in the current study.
The lack of difference in survival between AJCC/UICC and ENSAT stages I and II persisted despite the relatively large sample size in this analysis. These results suggest that current staging systems do not adequately separate lower risk groups based on probability of survival. Both the AJCC/UICC and ENSAT staging criteria for stage I and II differ only by the size criterion of ≤5 or >5 cm. Therefore, it seems that maximal tumor diameter is not a factor in predicting survival from ACC; this is further supported by the results of this and other studies, in which size dichotomized by
thresholds of 5 cm or even 12 cm was not a significant predictor of survival.15,16
The current study reaffirms the prognostic significance of resection margin status, tumor grade, and age that has been previously identi- fied.1-3,13,14 However, only age was incorporated in the proposed TNM-A staging system. First, given the large number of patients with undocumented histologic grades in currently available databases, incorporating grade into any potential staging sys- tem for ACC, as suggested by Miller et al,7 would be challenging and may lead to the inability to stage patients in retrospective databases. Moreover, tumor grading of ACC is subjective. Although tu- mor size lacks prognostic significance as shown in this study and other reported studies, 15-17 addition of age to current size criteria allows for significant differences to be observed. Incorporation of age into the ENSAT staging system will not be histori- cally disruptive, and will not lead to large numbers of unstageable patients in retrospective databases, but will allow for significant differences in OS be- tween all stages to be observed and, more impor- tant, lead to a better prediction of survival for each stage of disease, based on calculated HRs. An international expert panel on ACC has defined the low-risk group to be stage I-II with microscop- ically negative resection margins and a Ki67 of ≤10%.18 Challenges to using the international panel risk stratification model include the lack of standardization in immunohistochemical staining for Ki67 as well as interobserver variability.19 Also, the ENSAT staging system used in devising the risk stratification by the panel does not show a dif- ference between stages I and II.5 This study shows
1.0
a
0.8
Survival Probability
0.6
0.4
O
0.2
O
0.0
0
1
2
3
4
5
OS(Years)
UICCstage
1
2
3
4
1.0
0.8
Survival Probability
0.6
0.4
o
0.2
0
00
0
1
2
3
4
5
OS(Years)
ENSAT
1
2
3
4
that patients aged ≤55 or >55 with Tl or T2 tu- mors have significantly different prognoses. Accordingly, the TNM-A system may allow better risk stratification with less variability to guide adju- vant chemotherapy treatment. Given the prog- nostic significance of age on survival, potential differences in the tumor biology between younger patients and older patients may exist. Future research could examine existing molecular data,
including tumor profiling data, and query based on age to determine potential differential molecu- lar drivers in young patients versus old patients with ACC.
There are several limitations of this study. First, the exclusion of patients with missing variables may have introduced bias. However, missing vari- ables seemed to be missing at random and the final sample size utilized in this study represents
1.0
0.8
Survival Probability
0.6
0.4
0.2
Đ
0.0
0
1
2
3
4
5
OS(Years)
newstage
o
1
℮
2
o
3
℮
4
the largest cohort of patients in which staging of ACC has been evaluated. Second, recoding of SEER summary stage to AJCC/UICC TNM did not have a 100% agreement with the available prospectively collected AJCC/UICC TNM stage (since 2010); therefore, a small number of patients recoded as AJCC/UICC stage III may belong to AJCC/UICC stage IV. The impact of this is likely small, given that the overall agreement rate be- tween recoded SEER summary stage and prospec- tively collected AJCC/UICC TNM stage was 92%. Also, despite the high data quality standards of the NCDB, as evidenced by routine audits and use of trained registrars for data abstraction, some data entry errors may have occurred. Therefore, we excluded outlying tumor sizes. The lack of cause- specific mortality is an additional limitation of this study. Although it is recognized that most patients with ACC who die will do so from complications of recurrent or metastatic disease,14,18 competing causes of death may have impacted the relatively poor survival in the older age group. Another po- tential confounder to the observed difference in survival may be owing to the difference in likeli- hood to receive surgery between younger patients and older patients. However, both older and younger patients who underwent surgery had similar odds of having a positive margin status.
Despite these limitations, this study examined the largest available cohort of patients with ACC. The findings of this study corroborate those of
previous studies, which utilize smaller datasets, in that neither the AJCC/UICC nor ENSAT staging system is able to predict significant differences in survival between patients with stages I and II ACC. The development of a novel staging system, TNM- A, which incorporates age at the time of diagnosis, demonstrates significant differences in survival among patients with stage I, II, III, or IV ACC. The ability to improve predictions of OS is critical in patients with ACC, because better prognostica- tion may help to guide treatment decisions, such as the use of adjuvant chemotherapy in patients with stage I and II disease and appropriate selection of patients for clinical trials. Large single- and multi- institutional databases that capture specific cause of death are needed to validate the proposed TNM-A staging system. Future work will seek to build prognostic calculators that include other pertinent risk factors in addition to TNM/TNM-A staging. This is critically important in the manage- ment of patients with ACC, a disease in which OS is poor and options for postoperative adjuvant ther- apy remain limited.
REFERENCES
1. Kebebew E, Reiff E, Duh QY, Clark OH, McMillan A. Extent of disease at presentation and outcome for adrenocortical carcinoma: have we made progress? World J Surg 2006;30: 872-8.
2. Ng L, Libertino JM. Adrenocortical carcinoma: diagnosis, evaluation and treatment. J Urol 2003;169:5-11.
3. Schteingart DE, Doherty GM, Gauger PG, Giordano TJ, Hammer GD, Korobkin M, et al. Management of patients with adrenal cancer: recommendations of an international consensus conference. Endocr Relat Cancer 2005;12:667-80.
4. Paton BL, Novitsky YW, Zerey M, Harrell AG, Norton HJ, As- bun H, et al. Outcomes of adrenal cortical carcinoma in the United States. Surgery 2006;140:914-20.
5. Fassnacht M, Johanssen S, Quinkler M, Bucsky P, Willen- berg HS, Beuschlein F, et al. Limited prognostic value of the 2004 International Union Against Cancer staging classi- fication for adrenocortical carcinoma: proposal for a Revised TNM Classification. Cancer 2009;115:243-50.
6. Lughezzani G, Sun M, Perrotte P, Jeldres C, Alasker A, Is- barn H, et al. The European Network for the Study of Ad- renal Tumors staging system is prognostically superior to the international union against cancer-staging system: a North American validation. Eur J Cancer 2010;46:713-9.
7. Miller BS, Gauger PG, Hammer GD, Giordano TJ, Doherty GM. Proposal for modification of the ENSAT staging system for adrenocortical carcinoma using tumor grade. Langen- becks Arch Surg 2010;395:955-61.
8. Macfarlane DA. Cancer of the adrenal cortex; the natural history, prognosis and treatment in a study of fifty-five cases. Ann R Coll Surg Engl 1958;23:155-86.
9. Sullivan M, Boileau M, Hodges CV. Adrenal cortical carci- noma. J Urol 1978;120:660-5.
10. Icard P, Louvel A, Chapuis Y. Survival rates and prognostic factors in adrenocortical carcinoma. World J Surg 1992;16: 753-8.
11. Lee JE, Berger DH, el-Naggar AK, Hickey RC, Vassilopou- lou-Sellin R, Gagel RF, et al. Surgical management, DNA content, and patient survival in adrenal cortical carcinoma. Surgery 1995;118:1090-8.
12. AJCC Cancer Staging Manual. In: Edge SB, Byrd DR, Comp- ton CC, Fritz AG, Greene FL, Trotti A III, editors. AJCC can- cer staging manual. 7th ed. New York: Springer; 2011.
13. Abiven G, Coste J, Groussin L, Anract P, Tissier F, Legmann P, et al. Clinical and biological features in the prognosis of adrenocortical cancer: poor outcome of cortisol-secreting tumors in a series of 202 consecutive patients. J Clin Endo- crinol Metab 2006;91:2650-5.
14. Tritos NA, Cushing GW, Heatley G, Libertino JA. Clinical features and prognostic factors associated with adrenocor- tical carcinoma: Lahey Clinic Medical Center experience. Am Surg 2000;66:73-9.
15. Bilimoria KY, Shen WT, Elaraj D, Bentrem DJ, Winchester DJ, Kebebew E, et al. Adrenocortical carcinoma in the United States: treatment utilization and prognostic factors. Cancer 2008;113:3130-6.
16. Canter DJ, Mallin K, Uzzo RG, Egleston BL, Simhan J, Wal- ton J, et al. Association of tumor size with metastatic poten- tial and survival in patients with adrenocortical carcinoma: an analysis of the National Cancer Database. Can J Urol 2013;20:6915-21.
17. Kutikov A, Mallin K, Canter D, Wong YN, Uzzo RG. Effects of increased cross-sectional imaging on the diagnosis and prognosis of adrenocortical carcinoma: analysis of the Na- tional Cancer Database. J Urol 2011;186:805-10.
18. Berruti A, Fassnacht M, Baudin E, Hammer G, Haak H, Leb- oulleux S, et al. Adjuvant therapy in patients with adreno- cortical carcinoma: a position of an international panel. J Clin Oncol 2010;28:e401-2.
19. Terzolo M, Daffara F, Ardito A, Zaggia B, Basile V, Ferrari L, et al. Management of adrenal cancer: a 2013 update. J Endocrinol Invest 2014;37:207-17.
DISCUSSION
Dr Dave Schneider (Madison, WI): I have a question about your outcome. It seems like, if you are using overall survival, then you will bias yourself to finding that older patients do worse. Can you comment on that?
Dr Elliot A. Asare (Chicago, IL): Yes, there is po- tential for bias. Unfortunately, the National Cancer Data Base does not collect information on cause- specific mortality, so we were only limited to overall survival. However, the overall 5-year survival for the entire cohort was 38%, which is comparable with what has been reported for disease-specific survival in previous studies from single institutions or using SEER data. This is because most patients with adre- nocortical carcinoma die of complications related to the disease; hence, the potential for bias may be small.
Dr Dave Schneider: I would just add that SEER does have some disease-specific survival, so you might want to look in that subset alone.
Dr Michael J. Demeure (Scottsdale, AZ): Did you show that grade was independent of age? Is age an independent variable, or do patients who are older have higher grade tumors? Given that ACC is a rare disease and it has been shown that there is a wide disparity in pathologists, how can you be sure that the grade is accurate? Some peo- ple use the Weiss criteria and other things, so there is really no standardization.
Dr Elliot A. Asare: Thank you for raising very important questions. One of the rationales for trying to devise this staging system was to come up with something where the variables that are going to be included in the staging system will be standardized, readily available, and would also be easily collected in all cancer registries across the world. Such an approach would allow for cross- institutional collaborations to enhance future research for rare tumors by providing larger sam- ple size.
In our analysis, grade was found to be a predic- tor of survival on both univariable and multivari- able analysis. There was, however, a considerable number of patients who did not have recorded histologic grade. We agree with you that there may be wide variation in the assignment of grade even within institutions and across institutions. We did not look for interactions between histologic grade and age at diagnosis. Although there may be potential differences in the tumor biology between older patients and younger patients, we cannot conclude from our study whether there is a confounding relationship between age and grade.
Age proved to be a significant prognostic factor in our analysis, both as a continuous variable (data not shown) and as a dichotomized variable. Our results are consistent with some previously pub- lished studies.
Dr Steven Dejong (Maywood, IL): With the different philosophies on postoperative chemo- therapy for some of these patients, how do you see this scoring system or classification system affecting advice for or against postoperative chemotherapy for these patients?
Dr Elliot A. Asare: Although this was not within the scope of our work, the data on receipt of
chemotherapy in the National Cancer Database for this cohort was really sparse, so we did not assess the effect of chemotherapy. I know that there is a risk stratification system that calls for incorporation of Ki67. However, we know that Ki67 assays are not available at many institutions and there is also wide variability in interpretation of the assay results by pathologists. So, if a patient is of a particular age, based on our staging system, if she has had a com- plete tumor resection and the clinician feels that offering a chemotherapy might offer some benefit to hopefully decrease the recurrence, I think that is left to the clinician’s judgment.