1. on behalf of the ACOSOG Z4032 Investigators
Development of a Nomogram for Predicting Outcomes after Sublobar Resection for Lung Cancer
An Analysis of ACOSOG Z4032
I would like to thank the association for the privilege of presenting our work, entitled “Development of a Nomogram for Predicting Outcomes after Sublobar Resection for Lung Cancer: An Analysis of ACOSOG Z4032
Michael Kent, MD
on behalf of the ACOSOG Z4032 Investigators
AATS Annual Meeting, 2015

2. Disclosures
None
I have no financial disclosures

3. Introduction ACOSOG Z4032 Z4032 was a randomized, prospective trial
Compared sublobar resection to sublobar resection with brachytherapy
Accrual (n=224)
Included 41 centers and 48 surgeons
The American College of Surgeons Z4032 trial was a randomized, prospective study comparing sublobar resection alone, to sublobar resection plus brachytherapy.
This study was open from 2006 to 2010, and enrolled 224 patients.
The study involved 41 centers and 48 surgeons.

4. Introduction Z4032 Biopsy-proven, clinical stage I tumors 3cm or less
Considered high-risk on the basis of cardiopulmonary disease
Wedge or segmentectomy
VATS or open
Enrollment in the study was limited to those with biopsy proven, stage I lung cancers.
Only patients considered high-risk on the basis of cardiopulmonary disease were eligible.
Patients underwent wedge or segmentectomy, though either VATS or an open approach, at the discretion of the operating surgeon

5. One major or two minor criteria required
Z4032
Entry Criteria
Major Criteria
FEV1<50%
DLCO<50%
Minor Criteria
Age ≥ 75 yrs
FEV %
DLCO 51-60%
Pulmonary HTN
Poor LVEF
pO2≤55%
pCO2≥45mmHg
Dyspnea score ≥3
Enrollment in the trial required the presence of either one major, or two minor criteria, as shown.
One major or two minor criteria required

6. Z4032 Results No difference in OS, DFS or local recurrence rates
Any recurrence: 26%
5-yr OS: 59%
40% of deaths were due to cancer
Results of the trial were published this year.
No differences in overall survival, disease-free survival or local recurrence rates were seen
Overall, 26% of patients experienced a recurrence, and the 5-year overall survival was 59%. Of note, only 40% of the deaths in the trial were due to cancer
Fernando, J Clin Oncol 2015

7. Purpose To develop a nomogram to predict: Overall survival
Local recurrence-free survival
Any recurrence-free survival
Using data from ACOSOG Z4032
The purpose of the present study was to develop a nomogram to predict overall and recurrence-free survival, using data from ACOSOG Z4032

8. Methods Data Source Secondary analysis of Z4032
Data from both groups were combined
17 clinical variables were evaluated to construct survival and recurrence models
Only patients with all variables were included (n=173)
This was a secondary analysis of data from Z4032.
As there was no difference in any outcome measure between groups, data from both arms of the trial were combined for this analysis
We evaluated 17 clinical variables to model survival and recurrence

9. Methods Statistical Methodology
Factors significant in the univariate model (p≤.10) were included in the multivariate model
Final model used factors significant at the .05 level
Concordance index and calibration plots obtained using internal validation
Only factors significant on univariate analysiswere included in the multivariate model.
Furthermore, we only used patients for whom all clinical data was available.
The concordance index and calibration plots were obtained using internal validation.

10. Clinical Variables Baseline Factors Arm: SRB vs SR Age BMI
Baseline Performance Status
Race: White vs. Others
Method of payment: Uninsured/Medicaid vs. Others
ASA class: III/IV vs. I/II
Baseline DLCO%
Baseline FEV1%
Here are the variables used in development of the survival and recurrence models.
Shown are baseline factors, such as age, performance status and pulmonary function.

11. Clinical Variables Surgical and Tumor Variables
Surgery Approach: VATS vs. Thoracotomy
Type of Resection: Wedge vs. Segmentectomy
Clinical Nodule Size: >2 cm vs.≤ 2 cm
Actual Margin Size (cm)
Margin Tumor Ratio
Maximum Tumor Diameter (cm)
Lymph Node Evaluation: MLND/Sampling vs. None
Histology Type
Surgical and tumor-related variables were also recorded.
Operative notes and pathology reports were specifically reviewed to determine the method of resection, parenchymal margin and degree of lymph node evaluation.

12. CONSORT Diagram
Shown here is the CONSORT diagram for the present analysis.
224 patients were randomized to Z4032.
Of those, complete clinical information was available for 173 patients, and these patients formed the basis of the present study
n=173

13. Results Baseline Demographics
Median age: 70
Median F/U: 4.4 years
Mean DLCO: 46% predicted
Wedge resection: n=129 (74.6%)
No LN sampling: n=61 (35.3%)
Shown here are the baseline demographics for the patients included in this analysis
The median age was 70, with a follow-up period of 4.4 years
Pulmonary function was severely limited, with a median diffusion capacity of 46% of predicted.
Approximately two-thirds of patients underwent a VATS approach, and the majority of patients were treated with a wedge resection
Of note, a third of patients had no lymph nodes evaluated at the time of surgery

14. Results Baseline Demographics
Tumor diameter: 1.8 cm ( cm)
Margin size: 1.0 cm ( cm)
SCCA and AC equal
Overall 5-year survival: 58.4%
The median tumor diameter was 1.8cm, with a parenchymal margin of 1cm
Squamous cell and adenocarcinoma histologies were equally represented, and the overall five-year survival was 58%

15. Overall Survival Univariate Analysis
Factors
Hazard Ratio
p-value
Age
1.03 (1.00, 1.06)
0.06
Baseline DLCO%
0.97 (0.96, 0.99)
<0.01
Margin Tumor Ratio
0.66 (0.44, 0.98)
0.04
Maximum Tumor Diameter
1.37 (1.06, 1.78)
0.02
Histology Type
1.63 (1.02, 2.60)
Shown here are the factors which were significant predictors of overall 5-year survival in univariate analysis
Age, baseline DLCO, margin-tumor ratio, maximum tumor diameter and histology were all statistically significant

16. Overall Survival Multivariate Analysis
Factors
Hazard Ratio
p-value
Age
1.03 (1.00,1.06)
0.04
Baseline DLCO%
0.97 (0.95,0.99)
<0.01
Margin Tumor Ratio
0.83 (0.53,1.28)
0.39
Maximum Tumor Diameter
1.29 ( )
0.05
Histology Type
1.24 (0.76,2.02)
Shown here are the factors which were significant predictors of overall 5-year survival in multivariate analysis
Only age, baseline DLCO, and maximum tumor diameter were found to be statistically significant

17. Overall Survival Nomogram
Shown here is the nomogram to predict overall survival.
Patients are assigned points based on their age, baseline diffusion capacity and tumor diameter. The factor with the greatest effect size is assigned a maximum value of 100 points.
As one can see, diffusion capacity had the greatest effect on overall survival.
The total number of points for age, diffusion capaciity and tumor diameter are calculated, and a straight line is drawn to calculate the predicted overall 5-year survival. The greater the number of points, the lower predicted survival.
For example, a patient with 80 points would be predicted to have a 5-year survival of 68%.
The C-index is a measure of the predictive power of the nomogram, with a range from 0 to 1.0.
C-index: 0.622

18. Overall Survival Calibration Plot
Shown here is the calibration plot for the nomogram, demonstrating good agreement between predicted and observed overall survival.
The wide confidence intervals reflect the relatively small sample size in the study

19. Recurrence-Free Survival Nomogram
Nomograms also constructed for local recurrence, and any recurrence-free survival
Age, DLCO and tumor diameter remained the three significant factors on multivariate analysis
C-index for LRFS: 0.606
C-index for RFS: 0.591
We also performed a similar analysis for local recurrence-free survival, and any recurrence-free survival.
Similar to overall survival, only age, DLCO and tumor diameter remained statistically significant predictors on multivariate analysis.

20. Summary
Using data from ACOSOG Z4032, OS, LRFS and RFS were predicted by:
Age
Diffusion capacity
Tumor diameter
Parenchymal margin, LN sampling and anatomic resection were not predictive
In summary, in a secondary analysis of data from ACOSOG Z4032, overall survival, local-recurrence free survival, and overall recurrence-free survival were predicted by age, diffusion capacity and maximum tumor diameter.
Surprisingly, factors traditionally associated with appropriate oncologic surgery, such as the parenchymal margin, degree of lymph node evaluation and performance of a segmectectomy over a wedge resection, were not predictive of survival or recurrence.

21. Limitations Unique patient population
Many of the deaths were not due to cancer
Not all relevant datapoints available (e.g. SUVmax)
Small data set and no external validation
We acknowledge that there are important limitations to this study.
First, we should emphasize the unique patient population which we studied. Specifically, only high-risk patients with clinical stage I cancer were enrolled in ACOSOG Z4032.
Because of significant comorbidities, many of the deaths among these patients were not due to cancer recurrence
Also, certain data points, such as the SUV from the PET scan, were not collected in the case report forms and were therefore not considered in the model.
In addition, the sample size was relatively small. With a larger dataset it is possible that factors such as margin and lymph node sampling would have been statistically significant.
Finally, the nomogram has not been subjected to validation with an external dataset. At present, this would be difficult, given the unique patient population which we studied.

22. Applicability
May not be valid in lobectomy (standard- risk) population
May not be valid in patients treated with other modalities (e.g. SBRT)
The current nomograms are therefore considered exploratory
In addition, there are limitations to the applicability of the nomogram.
Specifically, we would not consider the model to be valid for standard-risk patients who are fit to undergo lobectomy.
In addition, the model has not been studied in patients undergoing non-operative therapy such as CyberKnife.

23. Strengths Multicenter, prospective database
Central review of all operative and pathology reports
Important cohort of patients in whom there is no consensus on treatment
However, there are some important strengths to this analysis.
First, the data was obtained from a multicenter prospective trial with pre-defined outcome measures
All operative and pathology reports were centrally reviewed
Importantly, the analysis focuses on a important cohort of patients in whom the optimal treatment is undefined

24. Future Directions
Plan to validate in an upcoming study comparing SRS to surgery in high-risk patients (Stablemates study, formerly ACOSOG Z4099)
Consider use of nomogram to stratify patient groups in future studies
We are hoping validate the nomogram in a upcoming study which aims to compare surgery to stereotactic radiosurgery in high-risk patients.
Furthermore, we hope that if this nomogram is validated, it may be ultimately be useful to determine the optimal method of treatment in high-risk patients.