Prognostic factors in resected stages I and II adenocarcinoma of the lung

J THORAC

CARDIOVASC SURG

1990;99:218-26

Prognostic factors in resected stages I and II adenocarcinoma of the lung A multivariate regression analysis of 137 consecutive patients The prognostic factors for sunival after radical resection of stages I and II adenocarcinoma of the lung were investigated in 137 consecutive patients. Seventeen variables were investigated by use of Cox's multivariate regression model. Factors predicting a poor prognosis were low status of Kamofsky performance, histologic subtype of solid carcinoma with formation of mucus, and T2 and Nl disease. The staging according to American Joint Committee on Cancer Staging did not provide significant information on sunival when the variables describing T and N status were included in the Cox model. The new international staging system carried prognostic information because the new stages I and II are identical with the NO and Nl groups. Groupings of patients according to the prognostic variables defined in this study allow for more detailed prediction of sunival than can be achieved solely by T and N groupings. The prognostic information may be used in the design and interpretation of clinical trials and in the selection of patients who might be considered potential candidates for studies of adjuvant treatment.

Jens B. S~rensen, MD,a and Jens H. Badsberg, MSC,b Copenhagen, Denmark

h e role of surgery in patients with localized non-small cell lung cancer (NSCLC) is well established, and it is clear that patient selection plays a critical role in the interpretation of the results from surgical trials. I Knowledge of prognostic factors is essential for selection of the best treatment for individual patients, for stratification in randomized trials and interpretation of the results, and for comparison between studies. At present a large number of the patients who have resections die within 5 years of operation. Stage of the disease is an important predictor ofsurvival. 2,3 Other factors have been attributed a prognostic impact among patients with stage I or II NSCLC, including the major histologic types (squamous cell carcinoma, adenocarcinoma, or

From the Department of Oncology ONK, Finsen Institute; and the Statistical Research Unit, Panum Institute," University of Copenhagen, Copenhagen, Denmark. Supported by grants from the Danish Cancer Society. Received for publication Oct. 13, 1988. Accepted for publication May I, 1989. Address for reprints: Jens Benn Serensen, MD, Department of Oncology ONK, Finsen Institute, 49 Strandboulevarden, DK-2IDO Copenhagen, Denmark.

12/1/14373

218

large cell carcinoma), performance status, leukocyte count, age and sex of patient, and extent of operation.v" Different prognostic subgroups within the stage I group defined by the American Joint Committee on Cancer Staging? have been reported.S 8, 9 and a new classification accounting for this has recently been proposed. 10 There is little available literature with respect to other potential factors predicting survival, such as pretreatment laboratory parameters, although such parameters have recently been evaluated among patients with stage III adenocarcinoma of the lung (ACL).ll Histologic subtyping of ACL included some prognostic information among patients with stages I and II ACL in univariate analyses, 12 but the prognostic impact in relation to other variables has not been explored, Knowledge regarding prognostic factors can be obtained by a multivariate regression analysis, 13 and among a number of theoretical models," Cox's multivariate regression model 15 is the most frequently applied when duration of survival is the measure of interest. On the basis of this model, the dominating prognostic factors can be characterized, and results from this analysis permit the estimation of survival probability. Accordingly, to clarify factors of dominating prognostic influence and to establish a prognostic index for pre-

Volume 99 Number 2

Adenocarcinoma of lung

February 1990

Table I. Influence of clinical features and surgical procedures on median duration ofsurvival

Variable

Sex Male Female Age :s57 yr >57 yr Performance status 90%-100% 700/0-80%

50%-60% Pulmonary symptoms No Yes Weight loss :s5%

>5% Operation type Lobectomy or less Operation on two lobes Pneumonectomy AJCstage" I II

New stage] I II

Lymph nodes NO NI Tumor size and location TI T2

No. of patients

Table II. Influence of histopathologic type and laboratory parameters on median duration ofsurvival

Median survival (wk)

68 69

112 130

51 86

150 107

77 53 7

134 98 35

43 94

149 114

116 19

127 114

95 7 35

128 131 99

III 26

135 56

107 30

131 75

107 30

131 75

55 82

145 101

p

0.07 0.05 0.02

0.30 0.69 0.06

0.0003 0.0007 0.0007 0.0001

'Stage according to the American Joint Committee on Cancer Staging.' tStage according to new international classification,to

diction of the survivalexpectancy in patients with stages I and II ACL, we performed a multivariate regression analysis on retrospectively collecteddata from all patients withradical resectionof stages I or II ACL in a defined geographic area and time period.

Material and methods We reviewed all records in the eastern region of Denmark (population 2.7 million) for patients with lung cancer who had resections during the period February 1981 through July 1985. Onlypatientswith macroscopicand microscopic indicationsfor radical pulmonary resection and regional lymph node biopsies and with surgical-pathologic stages I and II ACL according to the American Joint Committee? were included. Complete dissection of mediastinal lymph nodes was not done. The clinical record of each patient was reviewedfor the following information: age, sex, performance status, pulmonary

2 19

Variable

Histopathologic subtyping Acinar Papillary Bronchiole-alveolar Solid carcinoma Degree of differentiation Well Moderate Poor Hemoglobin "?7.5 mmoljL <7.5 rnrnol/L WBC ::s9x 109/L >9x 109/L AST :s40 U/L >40 U/L Alkaline phosphatase :s275 U/L >275 U/L LDH :s450 U/L >450 U/L

No. of patients

Median survival (wk)

p

0.20 78 16 24 19

122 126 150 37

29 36 72

144 135 113

118 19

128 49

25 40

218 114

98 6

131 112

99 21

138 107

70 6

131 52

0.28

0.02 0.10 0.86 0.42 0.18

symptoms, and weight lossduring the previous 6 months. Preoperative values of hemoglobin, total white blood cell count (WBC), serum alkaline phosphatase, serum aspartate aminotransaminase (AST) , and serum lactate dehydrogenase (LDH) were also included. Descriptions of operations were studied for typeofoperation,and this information along with the pathologicexamination wasusedto assigntumor (T) and lymph node (N) categories according to the American Joint Committee definitionofTNM categories(AJC staging), which in stages I and II disease are similar to the definitions in the new international staging system.'? In the AJC staging stage I is definedas Tl NO,T2 NO,or Tl Nl disease,whereas stage II is T2 NI disease. In the new staging system stage I comprisesT 1-2 NOand stage II is defined as Tl-2 N 1 disease. All histologic materials were reviewedby one pathologist and classified accordingto the World Health Organization criteria 16 (see Appendix). This histologic evaluation was performed without knowledge of the clinical results. Survivalfiguresare estimated accordingto the Kaplan-Meier method!"; the log rank test" is used for survival comparisons. Evaluationof the parameters for independentimpact on survival is made by Cox's multivariate regression model.P Further details on the statistical procedures are given in the Appendix.

Results Patient characteristics. A total of 137 patients with ACL had resectionsduring the study period. Assessment

The Journal of Thoracic and Cardiovascular

2 2 0 S¢rensen and Badsberg

Surgery

%

100 -

All patients (n ",137)

Stage I

80

Stage n

(n

~107)

(n;: 30)

OJ

c:

's .:;:

:;

.,

60

c.

-,

UJ

L

C Q) o

0;

Table III. Prognostic factors in radically resected stages I and II ACL: Results of Cox regression analysis in 137 patients

_,

Variable

Regression coefficient

Standard error

Performance status Tumor Node Solid carcinoma

0.45 0.69 0.63 0.68

0.19 0.25 0.25 0.32

p

0.02 0.006

om

0.03

'-l

~

40

L,

o,

Table IV. Relative death hazards according to different levels of tumor and node. estimated from the final Cox model in Table III

20

Relative death hazard o+----r--.,..---...,--~-__,r__-__,__+

o

400

800

1200

1600

2000

2400 Days

Fig. 1. Survival for patients with resected ACL, both overall and according to the new international staging.

of lymph nodes was done intraoperatively as well as by mediastinoscopy in 101 patients; in some of the remaining patients mediastinoscopy was replaced by computed tomographic scan. Details of the patient characteristics and treatment are given in Tables I and II, which include p values from univariate survival analysis. A total of 84 patients (61%) have died. The median observation time for the remaining 53 patients was 47 months (range, 25 to 80 months), with overall median survival of 127 weeks. Eight patients (6%) died within 30 days of operation: Four of these patients had pneumonectomies, three had lobectomies, and one had a wedge resection. Cause of death in three of the eight patients who died within 30 days of operation was acute myocardial infarction. The remaining five patients died of pulmonary embolism, septicemia, bleeding, respiratory insufficiency, and disseminated intravascular coagulation, respectively. Among the patients surviving more than 30 days, 59 died of clinically documented recurrence of ACL, four died of other cancers (one patient each with small cell lung carcinoma, pancreatic cancer, primary brain tumor [astrocytoma], and esophageal cancer [squamous cell carcinoma], three died of myocardial infarction, and two patients died of pneumonia. One patient each died of thoracic empyema, bleeding from an aortic aneurysm, unregulated diabetes, lung embolism, and suicide. The cause of death was unknown in three patients. Survival of patients is shown in Fig. 1. Multivariate regression analysis. The final Cox model was based on complete information for all 137 patients and showed a significant and independent prognostic im-

NO NI

Tl

T2

1.00 1.49

1.90 4.24

pact for performance status, T, N, and histologic subtyping (Table III). The variables describing performance status and hemoglobin level were negatively correlated; fewer patients with poor performance had hemoglobin levels above 7.5 mmol/L, In additional, there was a positive correlation between performance status and solid carcinoma with formation of mucus: More patients in the poor performance group had this histologic subtype.' . The AlC stage of disease was not significant when information on tumor and node staging was included in the Cox model. The relative death hazards according to different levelsofT and N and based on the final Cox model are demonstrated in Table IV. The probablity of death was nearly doubled with the presence of each of the following characteristics: T2, N 1, or solid carcinoma with formation of mucus. In addition, the probability of death increased by a factor 1.6 when performance status decreased from a range of90%-1 00% to 70%-80%or from 70%-80% to 50%-60%. Risk groups based on the regression models. Study of the variables analyzed herein revealed that the final Cox model should be stratified into two groups according to histologicdegree of differentiation (see Appendix). The two groups included patients with well-differentiated or moderately differentiated tumors and patients with poorly differentiated tumors. To illustrate the impact of the prognostic index, the survival curves for patients with well-differentiated or moderately differentiated tumors were grouped according to values of prognostic indexes from 0 to 0.25, 0.25 to 0.75, and >0.75 (Fig. 2). For patients with poorly differ-

Volume 99 Number 2

Adenocarcinoma of lung 2 2 1

February 1990

CJ

z

100 80

IZ

60

w a: w

No.of pts.

< 0.25 - - 0.25-0.75

s>a:

::) if)

Prog. Index

> 0.75

19 22 24

0

0...

,.,

40

c",

c __

,

'------

20

,

O+--.---,.-.,----r---r-,--,--,--,......--r----r---,

o

400

800

1200

1600

2000

2400

DAYS

Fig. 2. Survival for patients with well-differentiated or moderately differentiated adenocarcinoma of the lung, groupedaccording to prognostic indexes.

CJ Z

s

100

> a: 80 ::J

L _ r

Prog. Index

No.of pts,

< 0.5

17

0.5-1.0

15

1.0-1.5

23

> 1.5

17

C/)

l-

Z

W

e.

60

L.,

0

t - - - - - -

a: W 0...

,,

--I

t_

1 I

40

c.

:

-----------'-

--- I

1_ _ _ _ _ _ _ _ _

20 0

0

400

800

1200

1600

2000

2400 DAYS

Fig. 3. Survival for patients with poorlydifferentiated adenocarcinoma of the lung, grouped accordingto prognostic indexes. entiated tumors, the intervals of prognostic indexes employed were 0 to 0.5,0.5 to 1.0, 1.0 to 1.5, and> 1.5 (Fig. 3). The number of patients in each of these groups, their median survival time, and number of l-year, 2-year, and 5-year survivors are shown at Table V.

Discussion Several studies have evaluated prognostic factors by univariate analyses in patients with NSCLC who had

resections, especially with respect to groupings of T, N, and major histologic types. The two former variables are important predictors for postoperative survival,2-4,6,8, 19-22 although controversy exists with respect to the prognostic impact of the major histologic type. In three studies, each with fewer than 300 patients, there was no prognostic impact of major histologic type.2, 23, 24 In contrast, Mountain and colleagues'? observed a significantly better survival expectancy for 549 patients with

The Journal of Thoracic and Cardiovascular

2 2 2 StJrensen and Badsberg

Surgery

Table V. Groupings ofprognostic index, stratified according to histologic degree of differentiation Prognostic index

No. ofpatients (total/dead)

Median survival (days)

1 yr

2 yr

5 yr*

19/8 22/12 24/19

>2291 991 801

95% 86% 83%

79% 73% 58%

52% 41% 15%

17/4 15/10 23/18 17/13

>1802 1026 701 234

94% 67% 52% 44%

82% 60% 48% 19%

76% 19% 22% 19%

Proportion surviving

WeU- or moderately differentiated <0.25 0.25-0.75 >0.75

Poorlydifferentiated <0.5 0.5-1.0 1.0-1.5 > 1.5 *Kaplan-Meier estimate.

squamous cell lung cancer compared with 572 patients with ACL (p = 0.014). The Ludwig Lung Cancer Study Group" found a significantly poorer survival for 44 patients with large cell carcinoma compared with 255 patients with squamous cell carcinoma and 88 patients with ACL (p = 0.03); the two latter groups had a uniform prognosis. Other extensively evaluated variables include age, sex, and type of operation which have been significant" although not uniformv-" determinants of survival. Few reported studies haveused multivariate modelsfor evaluation of variables with independent prognostic information among patients with NSCLC who had resections. In two important studies Gail and colleagues" used a multivariate Weibull survival model-" to evaluate 392 patients with stage I disease who had resectionsand who were randomly assignedto receiveadjuvant intrathoracic immunotherapy or placebo, whereas Lipford and colleagues- used the Cox regression model to study 173 patients with stage I or II disease who had resections. None of thesestudiesexamined the simultaneouseffectof both T and N classifications, performance status, pretreatment laboratory parameters, histologic subtyping, and degree of differentiation in ACL; these among other potential variables were the focus of this study. Performance status is a major predictor of survival among patients with nonresectablestage III ACL. 11 Only Gail and colleagues" have evaluated the impact of performance status in less advanced stages of disease. They reported that Karnofsky performance status above 80% is an independent predictor of survival. This finding is in accordance with our study results. It is interesting to note that although it is a somewhat rough estimate of symptom status and ambulatory function, the impact of . performance status on survival is comparable to that of both T and N categories.

Data on laboratory parameters were not complete in our retrospective study. However, information concerning these variables was available in most patients (Table II) and was included to achievethe maximal information from the study. LDH, AST, and WBC have previously been determined by Cox analysis to have a significant impact on survivalamong patients with stage III ACL. 11 None of these laboratory parameters provided prognostic information for patients with less advanced ACL. However, because of the limited information about LDH and WBC, values for which were available in 76 and 65 patients, respectively, these two variables are currently not adequately evaluated. Gail and colleagues" also evaluated laboratory values, but only values for hemoglobin and serum calcium were significant in univariate testing, and none of the lO laboratory values examined achieved significance in the multivariate analysis. All histologic types of NSCLC were included in the latter study, and this may blur the possible prognostic importance for the individual histologic types. An impact on prognosis for histologic subtyping and degree of differentiationamong patients with stage I or II ACL has previously been reported.P based on results from univariate analyses and on the most recent World Health Organization classification. There were fewer 1year survivors among patients with solidcarcinoma with formation of mucus compared with all other subtypes, even though the overall survivalcurves for the four subtypes were not significantly different.'? Employing an earlier histologic classification by Kreyberg.I' Larsson-f observeda trend toward better survival for four patients with bronchoalveolar carcinoma compared with 54 patients with acinar and papillary adenocarcinoma; however, solid carcinoma with formation of mucus was not included.

Volume 99 Number 2 February 1990

In this multivariate analysis patients with bronchoalveolar carcinoma, acinar adenocarcinoma, and papillary adenocarcinoma could be combined into one group in the final Cox model. This group had significantly better survivalthan patients with solid carcinoma with formation of mucus. The latter group of patients represented 14% of the population included in the study, and the results indicate a consideration for subtyping of radically resected ACL tumors for more accurate prediction of prognosis and decisions regarding further therapy. Because no other multivariate studies have reported data on subtyping by the most recent WHO classification among patients with radically resected ACL, this subject should be further explored for confirmation or refutation of false positive findings. The degree of differentiation of the tumors had an impact on survival during the first 2 years after operation, but from that point, survival expectancies were similar.'? The well-differentiated and moderately differentiated tumors could be combined into one group, a finding that also has been reported among patients with stage III ACL. 29 Separating moderately differentiated from well-differentiated ACL tumors is difficult, and this may partly explain why these two degrees of differentiation have a similar prognostic impact. However, neither histologic subtyping nor degree of differentiation has been evaluated for interobserver variability in a study of radically resected tumors. The prognostic significance of both tumor classification and the presence or absence oflymph node metastases has repeatedly been demonstrated among patients with NSCLC,4.5,22 and in this study is also demonstrated solely among patients with ACL. These factors and the presence of distant metastases form the basis of staging. Previousstudies have documented differences in survival between the four tumor-node groups T1 NO, T1 Nl, T2 NO, and T2 Nl,5,22 which together form stages I and II. Combination of these groups into stages should reflect

major differences in prognosis and possibly in treatment

options as well. Whereas the TNM classification by the American Joint Committee pools T1 NO, T1 Nl, and T2 NO together to form stage I and leaves T2 N 1 for stage 11,7 others have recently recommended other approaches. Mountain'? defined a new stage I that includes patients with T1 or T2 tumors and no evidence of lymph node metastases (NO), whereas stage II disease includes patients with a primary tumor (classified as T1 or T2) and metastasis to the intrapulmonary (including hilar) lymph nodes(Nl). Other groups also observed that the survival for patients with Nl involvement differed considerably from survival in patients without nodal involvement, re-

Adenocarcinoma of lung 2 2 3

gardless of whether the primary tumors were classified as Tl or T2. 30,31 A study by Lipford and colleagues' indicated that patients whose nodal status and tumor size were both favorable (i.e., T1 and NO) should be considered prognostically separate from those in whom one or both of these features were unfavorable, particularly if lymph node metastasis is present. Also, Roeslin and colleagues" observed significantly longer survival for 108 patients with T1 NO disease compared with 171 patients with either T1 N 1 or T2 NO disease; the latter two groups of patients had similar survival. Our study supports and extends these observations and identifies three prognostically separate groups, according to a lack of or the presence of one or two of these negative prognostic signs: T1 NO defined the most favorable group, T1 Nl and T2 NO defined the intermediate range for prognosis, and T2 N 1 carried the worst prognosis (Table IV). Similarly, in a recent study by Mountain and colleagues'? the T1 Nl and T2 NO subsets patients with ACL had a significantly worse prognosis than patients with T1 NO disease; whereas in patients with squamous cell carcinoma the Tl N 1 subset indicated an intermediate prognosis that was not significantly different from either T1 NO or T2 NO status. Thus in patients with ACL the T1 NO subset carries a prognosis distinctly more favorable than other T N constellations. The 5-year survival rate was 58%, 24%, and 10% for patients with T1 NO, T1 Nl or T2 NO, and T2 Nl status, respectively. Only four patients (2.9%) in this study were classified in the T1 Nl subset. A comparable low number of patients qualifying for this subgroup has also been reported by others, ranging from 3.3% to 8.1%.3,4,20,22 It is interesting that although the AlC staging did not contribute independent information on survival in the Cox model, the opposite was noted for the new international staging system. The new stages I and II are identical to the NO and Nl groups, with subsets NO and Nl carrying significant prognostic weight in the final Cox model, and thereby contribute information on survival when combined with performance status, tumor classification, and histologic subtyping. Our study indicates that at least four variables offer independent prognostic information for patients with radically resected, stage I or II ACL. In addition to the well-recognized impact of tumor and node classifications, preoperative performance status and histologic subtyping of the resected tumor also contributed prognostic information, allowing for. more detailed information than achieved solely by T and N groupings. On the basis of these variables, it is possible to estimate prognosis, characterize study populations for optimal interpretation of

2 2 4 S(Jrensen and Badsberg

the results, and select patient groups who might be considered potential candidates for studies of adjuvant treatment. We express our appreciation to the staff of the departments of thoracic surgery and pathology at Rigshospitalet, Bispebjerg Hospital, Gentofte Hospital, and Nykpbing Falster Hospital, who treated the patients in this study and provided the histologic material for revision. We also thank Jens Olsen, MD, Department of Pathology, Rigshospitalet, Copenhagen, for revision of the histologic material. REFERENCES 1. Mulshine JL, Glatstein E, Ruckdeschel Jc. Treatment of non-small-cell lung cancer. J Clin Oneal 1986;4:1704-15. 2. Giedl J, Hohenberger W, Meister R. The pTNM classification of carcinomas of the lung, and its prognostic significance. Thorac Cardiovasc Surg 1983;31:71-5. 3. Shields TW. Classification and prognosis of surgical treated patients with bronchial carcinoma: analysis of VASOG studies. Int J Radiat Oncol Bioi Phys 1980;6:1021-7. 4. Gail MH, Eagan RT, Feld R, et al. Prognostic factors in patients with resected stage I non-small cell lung cancer. A report from the Lung Cancer Study Group. Cancer 1984;54: 1802-13. 5. Lipford EH, Sears DL, Eggleston JC, Moore GW, Lillemoe KD, Baker RR. Prognostic factors in surgically resected limited-stage, non small cell carcinoma of the lung. Am J Surg Pathol 1984;8:357-65. 6. Williams DE, Pairolero PC, Davis CS, et al. Survival of patients surgically treated for stage I lung cancer. J THORAC CARDIOVASC SURG 1981;82:70-6. 7. American Joint Committee for Cancer Staging and End Results Reporting, Task Force on Lung Cancer. Chicago, 1979:1-23. 8. Martini N, Beattie EJ. Results of surgical treatment in stage I lung cancer. J THORAC CARDIOVASC SURG 1977; 74:499-505. 9. Feld R, Rubinstein LV, Weisenberg TH, et al. Sites of recurrence in resected stage I non-small cell lung cancer: a guide for future studies. J Clin Oneal 1984;2:1352-8. 10. Mountain CF. A new international staging system for lung cancer. Chest 1986;89(suppl):225s-33s. 11. S¢rensen JB, Badsberg JH, Olsen J. Prognostic factors in inoperable adenocarcinoma of the lung: a multivariate regression analysis of259 patients. Cancer Res 1989;49:574854. 12. S¢rensen JB, Olsen JE. Prognostic implications of histopathologic subtyping in patients with surgically treated stages I and II adenocarcinoma of the lung. J THORAC CARDIOVASC SURG 1989;97:245-51. 13. Sather HN. The use of prognostic factors in clinical trials. Cancer 1986;58:461-7. 14. Kalbfleisch JD, Prentice RL. The statistical analysis of failure time data. New York: John Wiley & Sons, Inc, 1980.

The Journal of Thoracic and Cardiovascular Surgery

15. Cox DR. Regression models and life-tables. J R Stat Soc 1972;34:187-220. 16. World Health Organization. Histological typing of lung tumours. 2nd ed. Geneva: World Health Organization, 1981. 17. Kaplan EL, Meier P. New parametric estimation from incomplete observations. J Am Stat Assoc 1958;58:457-81. 18. Peto R, Pike MC, Armitage P, et al. Design and analysis of randomized clinical trials requiring prolonged observation of each patient. Br J Cancer 1977;35:1-39. 19. Shields TW, Humprey EW, Matthews M, Eastridge CE, Keehn RJ. Pathological stage grouping of patients with resected carcinoma of the lung. J THORAC CARDIOVASC SURG 1980;80:400-5. 20. Kayser K, Bulzebruck M, Probst G, Vogt-Moykopf I. Retrospective and prospective tumor staging evaluating prognostic factors in operated bronchus carcinoma patients. Cancer 1987;59:355-61. 21. Roeslin N, Chalkidakis G, Dumont P, Witz JP. A better prognostic value from a modification oflung cancer staging. J THORAC CARDIOVASC SURG 1987;94:504-9. 22. Mountain CF, Lukeman JM, Hammar SP, et al. Lung cancer classification: the relationship of disease extent and cell type to survival in a clinical trials population. J Surg Oncol 1987;35:147-56. 23. Vincent RG, Takita H, Lane WW, Gutierrez AC, Pickren JW. Surgical therapy of lung cancer. J THORAC CARDIaVASC SURG 1976;71:581-91. 24. Little AG, DeMeester TR, Ferguson MK, et al. Modified stage I (TlNOMO,T2NOMO), nonsmall cell lung cancer: treatment results, recurrence patterns, and adjuvant immunotherapy. Surgery 1986;100:621-8. 25. Ludwig Lung Cancer Study Group. Adverse effect of intrapleural Corynebacterium parvum as adjuvant therapy in resected stages I and II non-small cell carcinoma of the lung. J THORAC CARDIOVASC SURG 1985;89:842-7. 26. Gauless JF. Statistical models and methods for lifetime data. New York: John Wiley & Sons, Inc, 1982. 27. Kreyberg L. Histological typing of lung tumours. International histological classifications of tumours, no. 1. Geneva: World Health Organization, 1967. 28. Larsson S. Pretreatment classification and staging ofbronchogenic carcinoma. Scand J Thorac Cardiovasc Surg 1973;(suppllO):1-147. 29. S~rensen lB, Hirsch FR, Olsen l. The prognostic implication of histopathologic subtyping of pulmonary adenocarcinoma according to the classifications of the World Health Organization. An analysis of 259 consecutive patients with advanced disease. Cancer 1988;62:361-7. 30. Immerman SC, Vanecko RM, Fry WA, et al. Sites of recurrence in patients with stages I and II carcinoma of the lung resected for cure. Ann Thorac Surg 1981;32:23-7. 31. Ferguson MK, Little AG, Golomb HM, et al. The role of adjuvant therapy following resection ofTl Nl MOand T2 N 1 MOnon-small cell lung cancer. J THORAC CARDIOVASC SURG 1986;91:344-9.

Volume 99 Number 2

Adenocarcinoma of lung

February 1990

32. Saijo N, Niitani H, Tominaga K, et al. Comparison of survival in nonresected well differentiated and poorly differentiated adenocarcinoma of the lung. J Cancer Res Clin Oncol 1980;97:71-9. 33. Andersen PK. Testing goodness of fit of Cox's regression and life model. Biometrics 1982;38:67-77.

Appendix Histologic evaluation. The tumors were classified according to the predominant histologic type into four groups: acinar adenocarcinoma, papillary adenocarcinoma, bronchoalveolar carcinoma, and solid carcinoma with formation of mucus. The acinar and papillary subtypes were further described as well differentiated, moderately differentiated, and poorly differentiated according to the most highly differentiated tissue present. In the analysis of data the group with poorly differentiated acinar and papillary adenocarcinomas, together with solid carcinoma with formation of mucus, formed a group of patients with poorly differentiated tumors. The group with well-differentiated and moderately differentiated acinar and papillary adenocarcinomas, together with the bronchoalveolar adenocarcinoma, formed a patient group with well-differentiated tumors. Such grouping facilitated comparison with other studies 12.29.32 (see list of references following text). Statistical analysis. The survival times of the patients were calculated from date of operation until death or the most recent update (October 1987). No patients were lost to follow-up. Seventeen pretreatment variables were chosen for analysis either because previous studies had indicated a possible effect on prognosis or because such an effect seemed likely (see text Tables I and II). The univariate effect of individual variables on prognosis was examined by stratifying the patients on the basis of each variable in turn and by computing the Kaplan- Meier plots17 and the survival times of the respective groups by log rank test.!" All median survival values were estimated according to Kaplan-Meier plots.

c

-

.Q

All variables were to be evaluated for prognostic significance in a regression analysis performed with Cox's proportional hazards model.'> A stepwise procedure based on partial likelihood" ratios was used in the analysis. Survival related to individual variables. Results of the univariate analyses of survival among the 137 patients with radically resected ACL are summarized in text Tables I and II. On the basis of these nonparametric univariate analyses, a statisically significant (p < 0.05) prognostic influence was observed for performance status, stage, lymph node classification, tumor classification, and hemoglobin level. Several other variables approached significance, such as age (p = 0.05), solid carcinoma with formation of mucus versus other subtypes (p = 0.06), type of operation (p = 0.06), and sex of patient (p = 0.07). No other variables provided significant information in univariate analyses. Multivariate regression analyses. Multivariate regression analyses were used to examine the relative impact of each variable on survival. The model was based on complete information from 137 patients for all variables apart from laboratory tests and on histologic degree of differentiation simplified into the two groups mentioned above. The variables that supplied independent information in this model were performance status, tumor and node classification, and histopathologic subtyping. The prognostic index was calculated for all patients according to the coefficients in text Table III. In the histopathologic subtyping a model dividing the material into a group with solid carcinoma with formation of mucus and a group having other subtypes resulted in a significantly better fit to the survival data than was derived from other groupings of the subtypes. Verification of the model was done according to the graphic and numeric methods described by Andersen.P and it was found that the proportionality assumption was not reasonably fulfilled for groupings based on degree of differentiation (Appendix Fig. 1). Patients with well-differentiated or moderately differentiated tumors could be combined into one group that did not have frequency of death proportional to that in patients with poorly

0.0

- - - - 1- - - - - -

1- -

-

-

-

-

o

C :::J

- -1.0

~ .;?:

>

~

:::J

(/)

-2.0

/-'

,

OJ

..Q

(/) -3.0

..

;.: ..

Well differentiated (n; 29) Moderately differentiated (n = 36)

:::J

c

'E OJ

o

Poorly differentiated (n; 72)

-4.0

-l

o

400

800

225

1200

1600

2000

2400 DAYS

Appendix Fig. 1. Log minus log survival function stratified according to degree of differentiation.

2 26

S(Jrensen and Badsberg

differentiated tumors (p = 0.(04). Thus these two patient groups should not be combined into one model. Therefore the final regression model was stratified into two categories: patients with well-differentiated or moderately differentiated tumors and patients with poorly differentiated tumors. Variables describing continuous data were dichotomized according to the limits in text Tables I and II before they were included in the model. Neither age or weight loss, hemoglobin, WBC, alkaline phosphatase, nor AST or LDH provided signif-

TheJournal of Thoracic and Cardiovascular Surgery

icant and independent information. The five laboratory parameters were also examined, as dichotomized at the median levels, as divided into quartiles, as continuous variables, and with logarithmic transformation. However, values for hemoglobin, LDH, and AST approached significance when dichotomized according to the normal limits with p values of 0.08, 0.12, and 0.12. No other transformations revealed significant information, and these variables were therefore not included in the final model.

Bound volumes available to subscribers Bound volumes ofTHE JOURNAL OF THORACIC AND CARDIOVASCULAR SURGERY are available tosubscribers (only) forthe 1990 issues fromthe Publisher, at a costof$58.oo($75.00 international) forVol. 99 (January-June) and Vol. 100(July-December). Shipping charges are included. Each bound volume contains a subject and author index and all advertising is removed. Copies are shipped within 60 days after publication of the last issue of the volume. The binding is durable buckram with the JOURNAL name, volume number, and year stamped in gold on the spine. Payment must accompany all orders. ContactThe C.V. Mosby Company, Circulation Department, 11830 Westline Industrial Drive, St. Louis, Missouri 63146-3318, USA; phone (800) 325-4177, ext. 351. Subscriptions must be in forceto qualify. Bound volumes are not available in place of a regular JOURNAL subscription.