- Email: [email protected]
The impact of symptoms and comorbidity on prognosis in stage IB cervical cancer
The impact of symptoms and comorbidity on prognosis in stage IB cervical cancer Jeffrey F. Peipert, MD, MPH,.· b Carolyn K. Wells, MPH,b Peter E. Schwartz, MD,c and Alvan R. Feinstein, MDb. d Providence, Rhode Island, and New Haven, Connecticut OBJECTIVE: Like other gynecologic malignancies, cervical cancer is classified by the anatomic location and extent of the tumor. Because clinical variables such as patients' symptoms, symptom severity, and comorbidity may indicate a cancer's biologic virulence and the host-tumor interaction, this study was performed to test the hypothesis that clinical variables will also affect survival of patients with stage IB cervical cancer. STUDY DESIGN: From medical records of 251 cases of invasive cervical cancer treated at Yale-New Haven Hospital between 1984 and 1988, information was extracted for patients' demographic characteristics, symptoms, symptom severity, comorbidity, physical findings, laboratory data, treatment, and subsequent course. RESULTS: In the 122 available cases of stage IB cervical cancer the overall 3-year survival rate was 79%. For a composite clinical predictive system on the basis of symptom status and comorbidity, the 3-year survival rates were as follows: symptomatic patients with comorbidity 58% (seven of 12), either symptomatic or comorbid but not both 77% (46/60), and asymptomatic patients without comorbidity 86% (43/50) (p = 0.02 for linear trend x2 ). When entered into a Cox proportional-hazard model along with other variables that might have an impact on prognosis, the composite symptom-comorbidity stage was the only variable that remained statistically significant. CONCLUSION: These findings demonstrate the importance of clinical variables in estimating prognosis in stage IB cervical cancer. Unless these variables are suitably analyzed, prognostic estimates based only on morphologic studies will be imprecise and therapeutic evaluations may be misleading. (AM J OBSTET GYNECOL 1993; 169:598-604.)
Key words: Cervical cancer, prognosis, symptoms, comorbidity A cancer has a biologic effect as well as an anatomic form. Although both the function of the tumor and its anatomic extensiveness will affect the prognosis, current staging systems rely exclusively on anatomic categories, ignoring the cancer's function and the hosttumor interaction. I. 2 The multiple anatomic or morphologic factors that have been shown to affect the prognosis include histologic type, tumor volume, vascular or lymphatic space involvement, and depth of stromal invasion,3-5 but systematic attention has not been given to factors such as the presence or absence of symptoms, symptom severity, clinical evidence of rate of From the Department of Obstetrics and Gynecology, Women and Infants' Hospital, Brown University School of Medicine: and the Departments of Internal Medicine,' Obstetrics and Gynecology,' and Epidemiology and Public Health," Yale University School of Medicine. Supported in part by the Robert Wood johnson Foundation, Princeton, New jersey a.F.p.). Received for publication December 18, 1992; revised April 13, 1993; accepted April 30, 1993. Reprint requests: jeffrey F. Peipert, MD, MPH, Department of Obstetrics and Gynecology, Women and Infants Hospital of Rhode Island, 101 Dudley St., Providence, RI02905. Copyright © 1993 by Mosby-Year Book, Inc. 0002-9378/93 $1.00 + .20 6/1/48335
598
growth, or the impact of comorbid ailments, in spite of the prognostic importance these factors have shown for diverse other cancers. 6-12 In 1993 in the United States about 13,500 new cases of invasive cervical cancer will be diagnosed, an estimated 4400 deaths will occur as a result of the disease. 13 The best method of managing women with localized disease and poor prognostic factors is controversial,14 and new protocols are evaluating the use of chemotherapy in advanced disease l5 and in patients with "bulky" or extensive localized tumors. 16-18 The choices for treatment of cervical cancer are regularly affected by features of clinical and comorbid severity. For example, women who receive radiation are older, generally "sicker," and include more "poor surgical candidates" than women chosen predominantly from groups with a more favorable prognosis who are treated with surgery. Because these two groups of patients are not prognostically equal before therapy, a selection bias will distort the results if these factors are ignored in the evaluation of therapeutic efficacy. A heterogeneous group of patients exists within each International Federation of Gynecology and Obstetrics (FIGO) stage. Careful attention must be paid to clinical
Volume 169, Number 3 Am J Obstet Gynecol
variables such as symptom status and the presence of medical comorbidity that can have an impact on survival to control for this heterogeneity, Incorporating these variables into estimates of prognosis will provide more accurate and individualized prognostic information for patients and should be useful for prognostic stratification in clinical trials of new forms of therapy, 19 This study was performed to test the hypothesis that the patient's clinical symptom status and comorbidity would have an impact on prognosis in invasive cervical cancer. The specific hypotheses were that survival would be adversely affected by the presence of symptoms (vs no symptoms), the presence of more severe symptoms (vs mild symptoms), and the concomitant presence of severe comorbid conditions, The study hypothesis was tested within a single FIGO stage (i.e., stage IB) to control for the effect of morphologic stage on survival. If a combination of the symptomatic and comorbid features formed suitable levels of prognostic stratification within FIGO stage IB, impact of clinical variables could later be tested within all FIGO stages. Material and methods
The population under study was identified from the Yale Tumor Registry as being diagnosed or treated for invasive cervical cancer at Yale-New Haven Hospital between Jan. 1, 1984, and Dec. 31, 1988. The original search yielded 259 cases of invasive cervical cancer, of which two were excluded because of equivocal evidence of primary site. Six other patients were excluded because of unavailable medical record data in one, diagnosis of carcinoma in situ in two, and treatment after January 1989 in three. From the remaining 251 patients 134 were identified as having stage IB cervical cancer. They are evaluated separately in this report. Data extraction. The medical record, tumor registry record, and tumor conference notes for each patient were extracted and recorded onto a specially designed form according to methods previously described. 20 . 21 For classification of chronometry and baseline status, zero time was defined as the date of the first antineoplastic therapy for the cervical cancer. Clinical information and events that preceded zero time were described in detail, including the presence or absence of symptoms, the type and duration of symptoms, the existence of other conditions that might account for the symptoms, and the presence and severity of other major medical comorbid ailments. Also included were age; race; marital status; menopausal status; parity; history of cigarette smoking; alcohol or drug abuse; year of first treatment; details of radiographic, pathologic and physical examination findings; treatment modality; and subsequent course. Histologic slides had been previously reviewed at the Division of Gynecologic Oncology Tumor Conference.
Peipert et al.
599
Tumors were classified according to pre zero time tumor size. When this information was unavailable, a qualitative size of small or large was determined from the physician's description of the cervical examination. For example, if the tumor size in centimeters was not noted in the medical record, a qualitative description was cited if available. A normal cervical examination or lesions described as "small" were classified as small, whereas "large," "bulky," or "barrel" lesions were classified as large. Depth of cervical penetration was not precisely stated in most of the medical records reviewed and thus was not available for analysis. Each patient'S duration of follow-up and final status as alive or dead was obtained from the Yale Tumor Registry, supplemented by data from clinician's records in the Department of Gynecologic Oncology or Radiation Therapy. The National Death Index was used to obtain survival duration for patients whose follow-up status was previously not known. Because 3-year survival status was known for only 122 (91%) of the 134 patients, the remaining 12 women who are known to be alive at intervals < 36 months are excluded from tabular data reporting survival status at 3 years but are included for Cox proportional-hazard analyses of multiple variables. Classification of prognostic information. Each patient's condition immediately before zero time was classified according to conventional anatomic stages and two new clinical axes. All classifications were made without prior knowledge of the duration of survival. The tumor's anatomic spread, classified according to FIGO guidelines, was listed as axis I. Patients had been examined and staged by a gynecologic oncologist, and a radiation oncologist was also involved in staging when radiation therapy was considered for therapy. Because the definition of stage IB has changed over time, patients with > 3 mm depth of invasion below the basement membrane were considered to have stage IB disease for this study. Classification of symptoms and comorbidity. Clinical manifestations related to the cancer were listed in axis II. A "clinical manifestation" was defined as a reported symptom that was attributable or possibly attributable to the cancer. To be designated as attributable or possibly attributable, the manifestations had to be consistent with the cancer's anatomic status and customary effects, such as abnormal vaginal bleeding, discharge, pain, or urinary symptoms. 22 Primary symptoms included local symptoms such as vaginal bleeding or discharge or possibly regional manifestations such as pain or urinary complaints. Systemic symptoms included fatigue or nondeliberate weight loss. Patients with profuse bleeding or nondeliberate weight loss of > 10 pounds were designated as having severe symptoms. Symptom duration was also recorded.
600
Peipert et al.
September 1993 Am.J Obstet Gynecol
Table I. Clinical classification system based on symptoms (axis II) and comorbidity (axis III) Axis Ill: Comorbidity Axis II: Symptoms
Absent
Absent Present
ex
Present
~
All patients are FICO stage IB (axis I).
abies. For interpretation of statistical significance at < 0.05, we used one-tailed p values for results where the directional hypotheses were established before the research began. To evaluate the impact of baseline predictive variables, the Cox proportional-hazard method,"5 which can account for differing durations of follow-up, was used for all cases, and a multiple logistic regression was used for cases with 3-year survival data. Results
Comorbidity was classified for diagnostic and prognostic purposes. The term diagnostic comorbidity was used for any concomitant conditions that might account for the local symptoms. For local manifestations such as bleeding and vaginal discharge, diagnostically comorbid conditions might be perimenopausal status, leiomyoma, or vaginal infection. To be classified as present, the comorbid diagnoses must have been noted at or before zero time. A special category, quasiasymptomatic, was established to include all patients who were asymptomatic or who simultaneously had diagnostically comorbid conditions that might account for all of the symptoms. All other symptomatic patients were categorized as having oncogenic symptoms. Axis III included host factors unrelated to the cancer that might threaten life directly or make the patient susceptible to fatal ailments. The term prognostic comorbidity was applied to other medical conditions, beyond the cervical cancer, that could be expected to have a major impact on the patient's survival. Among such conditions were other malignancies; significant pathophysiologic disease in the heart, kidneys, or liver; and cerebrovascular disease or stroke. Comorbidity was graded for severity according to the methods proposed by Kaplan and Feinstein 23 and by Charlson et aJ.2 4 Prognostic comorbidity was considered present when either the Kaplan-Feinstein or Charlson score was 2: 2. A composite clinical classification was created that incorporated clinical axes II and III. The Creek letters ct, 13, and "y were chosen for subgroup designation to avoid confusion with alphanumeric characters used in the FICO classification. Composite clinical stage ct contained quasiasymptomatic patients without prognostic comorbidity. Composite stage 13 included patients with either oncogenic symptoms or prognostic comorbidity but not both. Composite stage "y included patients with both oncogenic symptoms and prognostic comorbidity. The composite classification system is summarized in Table I. Statistical analysis was performed with the Statistical Analysis Software system (SAS Institute). Univariate analyses were used to evaluate baseline variables, and bivariate analyses used Fisher exact tests, X2 analysis, and chi-square for linear trend to examine the prognostic impact of anatomic, histologic, and clinical vari-
The demographic and reproductive characteristics of the 134 cases of stage IB invasive cervical cancer are shown in Table II. The dates of first treatment (zero time) were distributed fairly evenly over the 5-year interval. The mean age of 44.6 years (range 21 to 84 years) is similar to that cited in analogous studies! Most patients were married (76%) and premenopausal. In separate bivariate analyses not shown here, the demographic and reproductive characteristics were found to have no substantial affect on survival. The effect of anatomic and histologic features on 3-year survival rates is shown in Table III. The histologic features of the cancer did not create prognostic distinctions, although with only 28 cases of adenocarcinoma the sample size is inadequate to detect statistically significant differences. Histologic grade also did not create clear prognostic gradients; there was little difference in survival between well, moderately well, and poorly differentiated tumors. The medical record contained reports for tumor size in centimeters in 58% of cases and actual measurements or qualitative descriptions of the cancer in 89%. The 3-year survival rate of 77% was slightly lower in the larger lesions than was the 89% survival rate in the smaller lesions, but the group sizes were too small for statistical significance. Survival results of the new clinical classifications are shown in Table IV. Among the 26% of stage IB patients who were asymptomatic, with cancers usually diagnosed by routine Papanicolaou smear, the 3-year survival rates were highest at 91 %. Primary symptoms, usually bleeding, were noted in 66% of patients; 8% had systemic symptoms. The corresponding 3-year survival rates were 84% and 70%, respectively (p = 0.13, linear trend X2). Of the 90 patients with symptoms 23 had other gynecologic or medical conditions as diagnostic comorbidity that might account for the symptoms. Among the latter group the 3-year survival rate was 96%, compared with a rate of 78% in the remaining 67 patients who were unequivocally symptomatic (p = 0.04). The 55 patients in the quasiasymptomatic category (i.e., being either asymptomatic or having diagnostic comorbidity) had a 3-year survival rate of 93%, whereas the patients in the oncogenic symptom category had a survival rate of 78% (p = 0.02).
Peipert et al.
Volume 169, Number 3 Am J Obstet Gynecol
601
Table II. Demographic and reproductive characteristics of patients with stage IB cervical cancer (N = 134) Characteristic
Category
No. of patients in category
Percentage of cohort
Date of zero time (first treatment)
1984 1985 1986 1987 1988 <30 30-39 40-49 50-59 60-69 ;;:,70
27 25 24 28 30 18 39 34 23 12 8 12 76 9 36 31 62 28 13 85 39 10
20 19 18 21 22 13 29 25 17 9 6 9 57 7 27 23 46 21 10 63 29 8
Age (yr)
Marital status
Single Married Widowed Other
0 1-2 3-4 ;;:,5
Parity
Menopausal status
Premenopausal Postmenopausal Unknown
Table III. Effect of anatomic and pathologic features on 3-year survival rates in stage IB cervical cancer (N = 122)
Characteristic Histologic type
Grade
Tumor size (cm) Qualitative size
Category
No. of patients in category
Percentage of cohort in category
No. of 3 yr survIVors
3 yr survival rate
85 28 8 1 23 51
70 23 6 1 19 42
70 26 6 1 20 43
82 93 75 100 87 84
35 13 36 34 52 70 39 13
29 10 30 28 42 57 32
30 10 33 27 42 62 30
86 77 92 79 81 89 77 85
Squamous Adenocarcinoma Adenosquamous Other Well differentiated Moderately well differentiated Poorly Unknown
:53 >3
Unknown Small Large Unknown
11
11
N one of the differences was statistically significant.
The duration of symptoms and the symptom severity were also found to have a suggestive impact on survival in patients with symptoms attributable to the cancer (oncogenic symptoms). The 3-year survival rates were 68% in patients with short duration of symptoms (::; 2 months), 76% for a symptom duration from 2 to 12 months, 81 % for symptom duration from 6 to 12 months, and 100% for a long duration of symptoms (> 12 months). The 3-year survival rates were 57% and 80%, respectively, for the presence or absence of severe symptoms in patients with oncogenic symptoms and 76% and 86% for the presence or absence of prognostic comorbidity in the entire cohort.
For the composite classification system, which incorporated symptoms and prognostic comorbidity as described earlier, the 3-year survival results are displayed in Table V. The 3-year rates were 86% for composite stage ex, 77% for composite stage (3, and 58% for stage 'Y. Thus within anatomic stage IB the clinical composite staging system produced a prognostic gradient that was clinically and statistically significant (p = 0.02, linear trend X2 ). To demonstrate its independent contribution to survival, the composite clinical classification was entered into various Cox proportional-hazard models along with the patient's age, tumor histologic type, grade, size,
602
Peipert et al.
September 1993 Am J Obstet Gynecol
1.0 ...---TI::-;L----, I : L:
OJ
0.8
!.--, L.."""L Alpha !.-------, ---"1....""1.... _ _ _ _ ! : - - - - - ' - - - - - "1.... ______ ,
:
L ___ ,
C
... _---,
.S;
en
Gamma
I
10._-------------------------------
.~ 06 :::J
Beta
____ _
I I
I
•
+-'
C
~ 0.4 Q)
a..
0.2
10
20
30
40
50
60
70
80
90
Months Fig. 1. Kaplan-Meier survival curves for stage IB cervical cancer according to composite clinical classification (Wilcoxon X2 4.62, P = 0.03).
Table IV. Distribution of patients with stage IE and 3-year survival rates for clinical manifestations (N = 122)
Characteristic Symptom type
Category Asymptomatic Primary Systemic Present Absent
Diagnostic comorbidity (symptomatic patients) Present Quasiasymptomatic Present Oncogenic symptoms Symptom duration (months) ";;2.0 2.1-6.0 6.0-12.0 '" 12.0 Absent Severe symptoms Present Absent Prognostic comorbidity Present
Percentage of cohort in category
No. of 3 yr survIVors
32 80 10 23 67
26 66 8 19 55
29 67 7 22 52
91 84 70 96 78
55 67 19 25 16 4 60 7 105 17
45 55 16 21 13 3 49 6 86 14
51 52 13 19 13 4 48 4 90 13
93 78 68 76 81 100 80 57 86 76
No. of patients in category
3 yr survival rate
Significance NS
P=
0.04
P=
0.02
NS
NS NS
NS, Not statistically significant. The p values represent one-tailed interpretations. Numbers differ because of missing or unknown values.
symptom duration, and symptom severity. Individual Cox models were used for the additional variables that had unknown values. Regardless of which models were used, however, the only variable that remained statistically significant in the models was the clinical composite stage. Kaplan-Meier survival curves for composite stages ex, ~, and "yare shown in Fig. I. The study and analysis was mainly concerned with prognostic variables that are known before therapy. Other variables that affect subsequent outcome include lymph node status and type of therapy. The presence of metastatic disease in the lymph nodes is One of the most important prognostic indicators of survival in patients with cervical cancer.' In the 76 patients with known
lymph node status and 3-year survival data the 3-year survival rates were 95% (58/61) and 60% (nine of 15), respectively, in women with negative and positive lymph nodes (p < 0.001). Patients with negative lymph nodes had a 100% survival rate (27/27) in composite clinical stage ex and 90% (29/34) in composite stages ~ or "y. Patients with positive lymph nodes had a 75% (three of four) survival rate in composite clinical stage ex and a 55% (six of 11) survival rate in women with composite stages ~ or "y. When stratified according to therapeutic modality, the 3-year survival rates were 89% (59/66) for radical hysterectomy and 78% (32/41) for radiation therapy. For patients who had radical surgery, the survival rates
Peipert et al.
Volume 169, Number 3 Am J Obstet Gynecol
603
Table V. Composite symptom-comorbidity classification system and 3-year survival rates in FICO stage IB cervical cancer (N = 122) Composite clinical stage ex
13 'Y
Description
No. of patients in category
Percentage of cohort in category
No. of 3 yr survzvors
3 yr survival rate
50
41
43
86
60
49
46
77
12
10
7
Quasiasymptomatic and no prognostic comorbidity Oncogenic symptoms or prognostic comorbidity present Oncogenic symptoms and prognostic comorbidity present
(p
=
58 0.02*)
*Linear trend X2 , one-tailed interpretation.
of according to clinical composite stage were a (28/29), 13 86% (30/35), and 'Y 50% (one of two). corresponding 3-year survival rates for patients had radiation therapy were a 81% (13/16), 13 (12/15), and 'Y 70% (seven of 10).
97% The who 80%
Comment Within each of the symptomatic and comorbid axes, differential survival rates were noted in the direction hypothesized, but the differences were not statistically significant because of small numbers within each subgroup. When symptoms and comorbidity were incorporated into a composite clinical classification system, however, the clinically important distinctions also became statistically significant. The distinctions were retained for patients receiving different modes of therapy. The rationale for the proposed classification of clinical variables has been previously described. 1. 2 The biologic virulence of a cancer can be indicated by its clinical manifestations, which reflect the tumor's functional effects and temporal growth. Rapidly growing cancers are particularly likely to produce symptoms rather than no symptoms and severe rather than nonsevere symptoms. '. 2. 6 In prognostic classification primary symptoms can be attributed to the tumor at its site of origin, systemic symptoms occur remote from the cancer's original site but do not imply anatomic dissemination, and metastatic symptoms arise remote from the tumor and imply anatomic dissemination. Severe or systemic symptoms are often caused by a more aggressive tumor, and metastatic symptoms indicate a virulent cancer. This taxonomy is associated with a "functional" prognostic gradient. 2 In asymptomatic patients the tumor has been present for an unknown length of time and might have remained silent for even longer periods of time had the cancer not been accidentally detected. In addition, slow-growing "silent" cancers may be detected because of symptoms produced by other gynecologic
conditions.'2 The duration of symptoms indicate the minimal amount of time a cancer has been present, because the cancer has been present as long as the symptoms. For symptoms of short duration, however, no definite conclusions can be drawn about the tumor's rate of growth. It may be either growing rapidly or growing slowly with a late appearance of symptoms. For symptoms of long duration, however, the cancer is likely to be slow growing, because the tumor has been present for at least as long as its symptoms. In addition to patterns of symptoms, the prognostic comorbidity of associated medical conditions has been shown to adversely affect survival in a variety of chronic diseases!" 24 For example, in endometrial cancer'2 patients without prognostic comorbidity had a 5-year survival rate of 78%, compared with a survival rate of 27% in patients with prognostic comorbidity. In a large series of stage I cervical cancer Hopkins and Morley" 4 found that diabetes mellitus significantly influenced survival in a Cox proportional-hazard model. The authors commented that patients with diabetes were generally older and in poorer medical condition. Clinical staging systems incorporating patients' initial clinical manifestations have been developed and have shown prognostic gradients for cancers of the lung,6 larynx,7 rectum,8. 9 breast, '0 prostate," and endometrium. '2 For example, in endometrial cancer'2 patients with primary or local symptoms had a 5-year survival rate of 78%, whereas patients with referred, systemic, or metastatic symptoms had survival rates of 67%, 57%, and 25%, respectively. The composite symptom-severity staging system produced prognostic distinctions within each FICO stage of endometrial cancer. Current staging systems for cancer identify the anatomic form but not the biologic effect or clinical function of the tumor. Although the term "early" is often used to describe anatomic stages such as stage I cervical cancer, it refers to a dimension of time, which is not measured in anatomic systems of staging. Although the
604
Peipert et al.
customary staging systems describe the morphologic appearance and structural damage produced by the tumor, no attention is given to the tumor's duration and rate of growth, which· can be manifested as the functional effects of the cancer in structures or systems that mayor may not be anatomically invaded. I, 2 The results of the current study confirm the hypothesis that clinical variables such as symptoms and comorbidity have important prognostic value in stage IE cervical cancer. These variables indicate aspects of the cancer's function and rate of growth that are not considered in traditional morphologic staging systems. Nevertheless, the morphologic staging systems denote anatomic features of the cancer that cannot be discerned from the clinical variables alone. Consequently, both morphologic and clinical features should be considered to produce adequate descriptions of the biologic behavior of a tumor and effective appraisals of prognosis and therapy. Numerous investigations in chronic diseases and a variety of malignancies have now demonstrated that current methods of classification in cancer staging are incomplete. Anatomic and morphologic systems have classified what a cancer is, but not what it does. 2 Because clinical variables such as symptom status and comorbidity have been shown to influence survival, increase prognostic precision, and improve the evaluation of new therapeutic modalities, these clinical variables should be appraised in gynecologic malignancies. Unless the clinical variables are suitably included, prognostic estimates based only on morphologic condition will be imprecise, and therapeutic evaluations may be misleading. We hope that other investigators will test and validate the importance of clinical variables in cohorts of women with cervical cancer or other gynecologic malignancies. REFERENCES 1. Feinstein AR. Symptoms as an index of biologic behavior and prognosis in human cancer. Nature 1966;209:241-5, 2. Feinstein AR. A new staging system for cancer and a reappraisal of "early" treatment and "cure" by radical surgery. N Engl j Med 1968;279:747-53. 3. Hopkins MP, Morley GW. Stage IB squamous cell cancer of the cervix: clinicopathologic features related to survival. AMj OBSTET GYNECOL 1991;164:1520-9. 4. Hopkins MP, Morley GW. A comparison of adenocarcinoma and squamous cell carcinoma of the cervix. Obstet GynecoI1991;77:912. 5. Gauthier P, Gore I, Shingleton HM. Identification of his-
September 1993 Am J Obstet Gynecol
6. 7. 8.
9.
10. 11.
12. 13. 14. 15. 16. 17, 18. 19. 20. 21. 22. 23. 24.
25.
topathologic risk groups in stage IB squamous cell carcinoma of the cervix. Obstet Gynecol 1985;66:569. Feinstein AR, Wells CK. A clinical-severity staging system for patients with lung cancer. Medicine 1990;69: 1-30. Feinstein AR, SchimpffCR, AndrewsJRjr, et al. Cancer of the larynx: a new staging system and a re-appraisal of prognosis and treatment. j Chron Dis 1977;30:277-305. Feinstein AR, Schimpff CR, Hull EW. A reappraisal of staging and therapy for patients with cancer of the rectum. I. Development of two systems of staging. Arch Intern Med 1975; 135: 1454-62. Feinstein AR, Schimpff CR, Hull EW. A reappraisal of staging and therapy for patients with cancer of the rectum. II. Patterns of presentation and outcome. Arch Intern Med 1975;135: 1454-63. Charlson ME, Feinstein AR. The auxometric dimension: a new method for using rate of growth in prognostic staging of breast cancer. JAMA 1974;228:180-5. Clemens jD, Feinstein AR, Holabird N, Cartwright S. A new clinical-anatomic staging system for evaluating prognosis and treatment of prostate cancer. j Chron Dis 1986;39:913-28, Wells CK, Stoller jK, Feinstein AR, Horwitz RI. Comorbid and clinical determinants of prognosis in endometrial cancer. Arch Intern Med 1984;144:2004-9. Boring CC, Squires TS, Tong T. Cancer statistics, 1993. CA 1993;43:7-26. Photopulos GJ. Surgery or radiation for early cervical cancer. Clin Obstet Gynecol 1990;33:872-82. Welander CE, Homesley HD, Barrett RJ. Combined interferon-a and doxorubicin in the treatment of advanced cervical cancer. AM j OBSTET GYNECOL 1991; 165 :284-91. Panici BP, Scambie F, Baiocchi G, et al. Neoadjuvent chemotherapy and radical surgery in locally advanced cervical cancer. Cancer 1991;67:372-9. Sardij, Sananes C, Giaroli A, et al. Neoadjuvent chemotherapy in locally advanced carcinoma of the cervix uteri. Gynecol Oncol 1990;38:486-93. Coleman DL, Gallup DG, Wolcott HD, et al. Patterns of failure of bulky-barrel carcinomas of the cervix. AM j OBSTET GYNECOL 1992;166:916-20. Feinstein AR. Clinical biostatistics, XIV. The purposes of prognostic stratification. Clin Pharmacol Ther 1972;13: 285-97. Feinstein AR, Pritchett jA, Schimpff CR. The epidemiology of cancer therapy. III. The management of imperfect data. Arch Intern Med 1969;123:448-61. Feinstein AR, Pritchett jA, Schimpff CR. The epidemiology of cancer therapy. IV. The extraction of data from medical records. Arch Intern Med 1969;123:571-90. DiSaia Pj, Creasman WT, Invasive cervical cancer. In: Clinical gynecologic oncology. St. Louis: Mosby-Year Book, 1993:66. Kaplan M, Feinstein AR. The importance of classirying initial co-morbidity in evaluating the outcome of diabetes mellitus. J Chron Dis 1974;27:387-404. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classirying prognostic comorbidity in longitudinal studies: development and validation. j Chron Dis 1987;40:373-83. Cox DR. Regression models and life tables. J R Stat Soc 1972;34:187-220.
Key words: Cervical cancer, prognosis, symptoms, comorbidity A cancer has a biologic effect as well as an anatomic form. Although both the function of the tumor and its anatomic extensiveness will affect the prognosis, current staging systems rely exclusively on anatomic categories, ignoring the cancer's function and the hosttumor interaction. I. 2 The multiple anatomic or morphologic factors that have been shown to affect the prognosis include histologic type, tumor volume, vascular or lymphatic space involvement, and depth of stromal invasion,3-5 but systematic attention has not been given to factors such as the presence or absence of symptoms, symptom severity, clinical evidence of rate of From the Department of Obstetrics and Gynecology, Women and Infants' Hospital, Brown University School of Medicine: and the Departments of Internal Medicine,' Obstetrics and Gynecology,' and Epidemiology and Public Health," Yale University School of Medicine. Supported in part by the Robert Wood johnson Foundation, Princeton, New jersey a.F.p.). Received for publication December 18, 1992; revised April 13, 1993; accepted April 30, 1993. Reprint requests: jeffrey F. Peipert, MD, MPH, Department of Obstetrics and Gynecology, Women and Infants Hospital of Rhode Island, 101 Dudley St., Providence, RI02905. Copyright © 1993 by Mosby-Year Book, Inc. 0002-9378/93 $1.00 + .20 6/1/48335
598
growth, or the impact of comorbid ailments, in spite of the prognostic importance these factors have shown for diverse other cancers. 6-12 In 1993 in the United States about 13,500 new cases of invasive cervical cancer will be diagnosed, an estimated 4400 deaths will occur as a result of the disease. 13 The best method of managing women with localized disease and poor prognostic factors is controversial,14 and new protocols are evaluating the use of chemotherapy in advanced disease l5 and in patients with "bulky" or extensive localized tumors. 16-18 The choices for treatment of cervical cancer are regularly affected by features of clinical and comorbid severity. For example, women who receive radiation are older, generally "sicker," and include more "poor surgical candidates" than women chosen predominantly from groups with a more favorable prognosis who are treated with surgery. Because these two groups of patients are not prognostically equal before therapy, a selection bias will distort the results if these factors are ignored in the evaluation of therapeutic efficacy. A heterogeneous group of patients exists within each International Federation of Gynecology and Obstetrics (FIGO) stage. Careful attention must be paid to clinical
Volume 169, Number 3 Am J Obstet Gynecol
variables such as symptom status and the presence of medical comorbidity that can have an impact on survival to control for this heterogeneity, Incorporating these variables into estimates of prognosis will provide more accurate and individualized prognostic information for patients and should be useful for prognostic stratification in clinical trials of new forms of therapy, 19 This study was performed to test the hypothesis that the patient's clinical symptom status and comorbidity would have an impact on prognosis in invasive cervical cancer. The specific hypotheses were that survival would be adversely affected by the presence of symptoms (vs no symptoms), the presence of more severe symptoms (vs mild symptoms), and the concomitant presence of severe comorbid conditions, The study hypothesis was tested within a single FIGO stage (i.e., stage IB) to control for the effect of morphologic stage on survival. If a combination of the symptomatic and comorbid features formed suitable levels of prognostic stratification within FIGO stage IB, impact of clinical variables could later be tested within all FIGO stages. Material and methods
The population under study was identified from the Yale Tumor Registry as being diagnosed or treated for invasive cervical cancer at Yale-New Haven Hospital between Jan. 1, 1984, and Dec. 31, 1988. The original search yielded 259 cases of invasive cervical cancer, of which two were excluded because of equivocal evidence of primary site. Six other patients were excluded because of unavailable medical record data in one, diagnosis of carcinoma in situ in two, and treatment after January 1989 in three. From the remaining 251 patients 134 were identified as having stage IB cervical cancer. They are evaluated separately in this report. Data extraction. The medical record, tumor registry record, and tumor conference notes for each patient were extracted and recorded onto a specially designed form according to methods previously described. 20 . 21 For classification of chronometry and baseline status, zero time was defined as the date of the first antineoplastic therapy for the cervical cancer. Clinical information and events that preceded zero time were described in detail, including the presence or absence of symptoms, the type and duration of symptoms, the existence of other conditions that might account for the symptoms, and the presence and severity of other major medical comorbid ailments. Also included were age; race; marital status; menopausal status; parity; history of cigarette smoking; alcohol or drug abuse; year of first treatment; details of radiographic, pathologic and physical examination findings; treatment modality; and subsequent course. Histologic slides had been previously reviewed at the Division of Gynecologic Oncology Tumor Conference.
Peipert et al.
599
Tumors were classified according to pre zero time tumor size. When this information was unavailable, a qualitative size of small or large was determined from the physician's description of the cervical examination. For example, if the tumor size in centimeters was not noted in the medical record, a qualitative description was cited if available. A normal cervical examination or lesions described as "small" were classified as small, whereas "large," "bulky," or "barrel" lesions were classified as large. Depth of cervical penetration was not precisely stated in most of the medical records reviewed and thus was not available for analysis. Each patient'S duration of follow-up and final status as alive or dead was obtained from the Yale Tumor Registry, supplemented by data from clinician's records in the Department of Gynecologic Oncology or Radiation Therapy. The National Death Index was used to obtain survival duration for patients whose follow-up status was previously not known. Because 3-year survival status was known for only 122 (91%) of the 134 patients, the remaining 12 women who are known to be alive at intervals < 36 months are excluded from tabular data reporting survival status at 3 years but are included for Cox proportional-hazard analyses of multiple variables. Classification of prognostic information. Each patient's condition immediately before zero time was classified according to conventional anatomic stages and two new clinical axes. All classifications were made without prior knowledge of the duration of survival. The tumor's anatomic spread, classified according to FIGO guidelines, was listed as axis I. Patients had been examined and staged by a gynecologic oncologist, and a radiation oncologist was also involved in staging when radiation therapy was considered for therapy. Because the definition of stage IB has changed over time, patients with > 3 mm depth of invasion below the basement membrane were considered to have stage IB disease for this study. Classification of symptoms and comorbidity. Clinical manifestations related to the cancer were listed in axis II. A "clinical manifestation" was defined as a reported symptom that was attributable or possibly attributable to the cancer. To be designated as attributable or possibly attributable, the manifestations had to be consistent with the cancer's anatomic status and customary effects, such as abnormal vaginal bleeding, discharge, pain, or urinary symptoms. 22 Primary symptoms included local symptoms such as vaginal bleeding or discharge or possibly regional manifestations such as pain or urinary complaints. Systemic symptoms included fatigue or nondeliberate weight loss. Patients with profuse bleeding or nondeliberate weight loss of > 10 pounds were designated as having severe symptoms. Symptom duration was also recorded.
600
Peipert et al.
September 1993 Am.J Obstet Gynecol
Table I. Clinical classification system based on symptoms (axis II) and comorbidity (axis III) Axis Ill: Comorbidity Axis II: Symptoms
Absent
Absent Present
ex
Present
~
All patients are FICO stage IB (axis I).
abies. For interpretation of statistical significance at < 0.05, we used one-tailed p values for results where the directional hypotheses were established before the research began. To evaluate the impact of baseline predictive variables, the Cox proportional-hazard method,"5 which can account for differing durations of follow-up, was used for all cases, and a multiple logistic regression was used for cases with 3-year survival data. Results
Comorbidity was classified for diagnostic and prognostic purposes. The term diagnostic comorbidity was used for any concomitant conditions that might account for the local symptoms. For local manifestations such as bleeding and vaginal discharge, diagnostically comorbid conditions might be perimenopausal status, leiomyoma, or vaginal infection. To be classified as present, the comorbid diagnoses must have been noted at or before zero time. A special category, quasiasymptomatic, was established to include all patients who were asymptomatic or who simultaneously had diagnostically comorbid conditions that might account for all of the symptoms. All other symptomatic patients were categorized as having oncogenic symptoms. Axis III included host factors unrelated to the cancer that might threaten life directly or make the patient susceptible to fatal ailments. The term prognostic comorbidity was applied to other medical conditions, beyond the cervical cancer, that could be expected to have a major impact on the patient's survival. Among such conditions were other malignancies; significant pathophysiologic disease in the heart, kidneys, or liver; and cerebrovascular disease or stroke. Comorbidity was graded for severity according to the methods proposed by Kaplan and Feinstein 23 and by Charlson et aJ.2 4 Prognostic comorbidity was considered present when either the Kaplan-Feinstein or Charlson score was 2: 2. A composite clinical classification was created that incorporated clinical axes II and III. The Creek letters ct, 13, and "y were chosen for subgroup designation to avoid confusion with alphanumeric characters used in the FICO classification. Composite clinical stage ct contained quasiasymptomatic patients without prognostic comorbidity. Composite stage 13 included patients with either oncogenic symptoms or prognostic comorbidity but not both. Composite stage "y included patients with both oncogenic symptoms and prognostic comorbidity. The composite classification system is summarized in Table I. Statistical analysis was performed with the Statistical Analysis Software system (SAS Institute). Univariate analyses were used to evaluate baseline variables, and bivariate analyses used Fisher exact tests, X2 analysis, and chi-square for linear trend to examine the prognostic impact of anatomic, histologic, and clinical vari-
The demographic and reproductive characteristics of the 134 cases of stage IB invasive cervical cancer are shown in Table II. The dates of first treatment (zero time) were distributed fairly evenly over the 5-year interval. The mean age of 44.6 years (range 21 to 84 years) is similar to that cited in analogous studies! Most patients were married (76%) and premenopausal. In separate bivariate analyses not shown here, the demographic and reproductive characteristics were found to have no substantial affect on survival. The effect of anatomic and histologic features on 3-year survival rates is shown in Table III. The histologic features of the cancer did not create prognostic distinctions, although with only 28 cases of adenocarcinoma the sample size is inadequate to detect statistically significant differences. Histologic grade also did not create clear prognostic gradients; there was little difference in survival between well, moderately well, and poorly differentiated tumors. The medical record contained reports for tumor size in centimeters in 58% of cases and actual measurements or qualitative descriptions of the cancer in 89%. The 3-year survival rate of 77% was slightly lower in the larger lesions than was the 89% survival rate in the smaller lesions, but the group sizes were too small for statistical significance. Survival results of the new clinical classifications are shown in Table IV. Among the 26% of stage IB patients who were asymptomatic, with cancers usually diagnosed by routine Papanicolaou smear, the 3-year survival rates were highest at 91 %. Primary symptoms, usually bleeding, were noted in 66% of patients; 8% had systemic symptoms. The corresponding 3-year survival rates were 84% and 70%, respectively (p = 0.13, linear trend X2). Of the 90 patients with symptoms 23 had other gynecologic or medical conditions as diagnostic comorbidity that might account for the symptoms. Among the latter group the 3-year survival rate was 96%, compared with a rate of 78% in the remaining 67 patients who were unequivocally symptomatic (p = 0.04). The 55 patients in the quasiasymptomatic category (i.e., being either asymptomatic or having diagnostic comorbidity) had a 3-year survival rate of 93%, whereas the patients in the oncogenic symptom category had a survival rate of 78% (p = 0.02).
Peipert et al.
Volume 169, Number 3 Am J Obstet Gynecol
601
Table II. Demographic and reproductive characteristics of patients with stage IB cervical cancer (N = 134) Characteristic
Category
No. of patients in category
Percentage of cohort
Date of zero time (first treatment)
1984 1985 1986 1987 1988 <30 30-39 40-49 50-59 60-69 ;;:,70
27 25 24 28 30 18 39 34 23 12 8 12 76 9 36 31 62 28 13 85 39 10
20 19 18 21 22 13 29 25 17 9 6 9 57 7 27 23 46 21 10 63 29 8
Age (yr)
Marital status
Single Married Widowed Other
0 1-2 3-4 ;;:,5
Parity
Menopausal status
Premenopausal Postmenopausal Unknown
Table III. Effect of anatomic and pathologic features on 3-year survival rates in stage IB cervical cancer (N = 122)
Characteristic Histologic type
Grade
Tumor size (cm) Qualitative size
Category
No. of patients in category
Percentage of cohort in category
No. of 3 yr survIVors
3 yr survival rate
85 28 8 1 23 51
70 23 6 1 19 42
70 26 6 1 20 43
82 93 75 100 87 84
35 13 36 34 52 70 39 13
29 10 30 28 42 57 32
30 10 33 27 42 62 30
86 77 92 79 81 89 77 85
Squamous Adenocarcinoma Adenosquamous Other Well differentiated Moderately well differentiated Poorly Unknown
:53 >3
Unknown Small Large Unknown
11
11
N one of the differences was statistically significant.
The duration of symptoms and the symptom severity were also found to have a suggestive impact on survival in patients with symptoms attributable to the cancer (oncogenic symptoms). The 3-year survival rates were 68% in patients with short duration of symptoms (::; 2 months), 76% for a symptom duration from 2 to 12 months, 81 % for symptom duration from 6 to 12 months, and 100% for a long duration of symptoms (> 12 months). The 3-year survival rates were 57% and 80%, respectively, for the presence or absence of severe symptoms in patients with oncogenic symptoms and 76% and 86% for the presence or absence of prognostic comorbidity in the entire cohort.
For the composite classification system, which incorporated symptoms and prognostic comorbidity as described earlier, the 3-year survival results are displayed in Table V. The 3-year rates were 86% for composite stage ex, 77% for composite stage (3, and 58% for stage 'Y. Thus within anatomic stage IB the clinical composite staging system produced a prognostic gradient that was clinically and statistically significant (p = 0.02, linear trend X2 ). To demonstrate its independent contribution to survival, the composite clinical classification was entered into various Cox proportional-hazard models along with the patient's age, tumor histologic type, grade, size,
602
Peipert et al.
September 1993 Am J Obstet Gynecol
1.0 ...---TI::-;L----, I : L:
OJ
0.8
!.--, L.."""L Alpha !.-------, ---"1....""1.... _ _ _ _ ! : - - - - - ' - - - - - "1.... ______ ,
:
L ___ ,
C
... _---,
.S;
en
Gamma
I
10._-------------------------------
.~ 06 :::J
Beta
____ _
I I
I
•
+-'
C
~ 0.4 Q)
a..
0.2
10
20
30
40
50
60
70
80
90
Months Fig. 1. Kaplan-Meier survival curves for stage IB cervical cancer according to composite clinical classification (Wilcoxon X2 4.62, P = 0.03).
Table IV. Distribution of patients with stage IE and 3-year survival rates for clinical manifestations (N = 122)
Characteristic Symptom type
Category Asymptomatic Primary Systemic Present Absent
Diagnostic comorbidity (symptomatic patients) Present Quasiasymptomatic Present Oncogenic symptoms Symptom duration (months) ";;2.0 2.1-6.0 6.0-12.0 '" 12.0 Absent Severe symptoms Present Absent Prognostic comorbidity Present
Percentage of cohort in category
No. of 3 yr survIVors
32 80 10 23 67
26 66 8 19 55
29 67 7 22 52
91 84 70 96 78
55 67 19 25 16 4 60 7 105 17
45 55 16 21 13 3 49 6 86 14
51 52 13 19 13 4 48 4 90 13
93 78 68 76 81 100 80 57 86 76
No. of patients in category
3 yr survival rate
Significance NS
P=
0.04
P=
0.02
NS
NS NS
NS, Not statistically significant. The p values represent one-tailed interpretations. Numbers differ because of missing or unknown values.
symptom duration, and symptom severity. Individual Cox models were used for the additional variables that had unknown values. Regardless of which models were used, however, the only variable that remained statistically significant in the models was the clinical composite stage. Kaplan-Meier survival curves for composite stages ex, ~, and "yare shown in Fig. I. The study and analysis was mainly concerned with prognostic variables that are known before therapy. Other variables that affect subsequent outcome include lymph node status and type of therapy. The presence of metastatic disease in the lymph nodes is One of the most important prognostic indicators of survival in patients with cervical cancer.' In the 76 patients with known
lymph node status and 3-year survival data the 3-year survival rates were 95% (58/61) and 60% (nine of 15), respectively, in women with negative and positive lymph nodes (p < 0.001). Patients with negative lymph nodes had a 100% survival rate (27/27) in composite clinical stage ex and 90% (29/34) in composite stages ~ or "y. Patients with positive lymph nodes had a 75% (three of four) survival rate in composite clinical stage ex and a 55% (six of 11) survival rate in women with composite stages ~ or "y. When stratified according to therapeutic modality, the 3-year survival rates were 89% (59/66) for radical hysterectomy and 78% (32/41) for radiation therapy. For patients who had radical surgery, the survival rates
Peipert et al.
Volume 169, Number 3 Am J Obstet Gynecol
603
Table V. Composite symptom-comorbidity classification system and 3-year survival rates in FICO stage IB cervical cancer (N = 122) Composite clinical stage ex
13 'Y
Description
No. of patients in category
Percentage of cohort in category
No. of 3 yr survzvors
3 yr survival rate
50
41
43
86
60
49
46
77
12
10
7
Quasiasymptomatic and no prognostic comorbidity Oncogenic symptoms or prognostic comorbidity present Oncogenic symptoms and prognostic comorbidity present
(p
=
58 0.02*)
*Linear trend X2 , one-tailed interpretation.
of according to clinical composite stage were a (28/29), 13 86% (30/35), and 'Y 50% (one of two). corresponding 3-year survival rates for patients had radiation therapy were a 81% (13/16), 13 (12/15), and 'Y 70% (seven of 10).
97% The who 80%
Comment Within each of the symptomatic and comorbid axes, differential survival rates were noted in the direction hypothesized, but the differences were not statistically significant because of small numbers within each subgroup. When symptoms and comorbidity were incorporated into a composite clinical classification system, however, the clinically important distinctions also became statistically significant. The distinctions were retained for patients receiving different modes of therapy. The rationale for the proposed classification of clinical variables has been previously described. 1. 2 The biologic virulence of a cancer can be indicated by its clinical manifestations, which reflect the tumor's functional effects and temporal growth. Rapidly growing cancers are particularly likely to produce symptoms rather than no symptoms and severe rather than nonsevere symptoms. '. 2. 6 In prognostic classification primary symptoms can be attributed to the tumor at its site of origin, systemic symptoms occur remote from the cancer's original site but do not imply anatomic dissemination, and metastatic symptoms arise remote from the tumor and imply anatomic dissemination. Severe or systemic symptoms are often caused by a more aggressive tumor, and metastatic symptoms indicate a virulent cancer. This taxonomy is associated with a "functional" prognostic gradient. 2 In asymptomatic patients the tumor has been present for an unknown length of time and might have remained silent for even longer periods of time had the cancer not been accidentally detected. In addition, slow-growing "silent" cancers may be detected because of symptoms produced by other gynecologic
conditions.'2 The duration of symptoms indicate the minimal amount of time a cancer has been present, because the cancer has been present as long as the symptoms. For symptoms of short duration, however, no definite conclusions can be drawn about the tumor's rate of growth. It may be either growing rapidly or growing slowly with a late appearance of symptoms. For symptoms of long duration, however, the cancer is likely to be slow growing, because the tumor has been present for at least as long as its symptoms. In addition to patterns of symptoms, the prognostic comorbidity of associated medical conditions has been shown to adversely affect survival in a variety of chronic diseases!" 24 For example, in endometrial cancer'2 patients without prognostic comorbidity had a 5-year survival rate of 78%, compared with a survival rate of 27% in patients with prognostic comorbidity. In a large series of stage I cervical cancer Hopkins and Morley" 4 found that diabetes mellitus significantly influenced survival in a Cox proportional-hazard model. The authors commented that patients with diabetes were generally older and in poorer medical condition. Clinical staging systems incorporating patients' initial clinical manifestations have been developed and have shown prognostic gradients for cancers of the lung,6 larynx,7 rectum,8. 9 breast, '0 prostate," and endometrium. '2 For example, in endometrial cancer'2 patients with primary or local symptoms had a 5-year survival rate of 78%, whereas patients with referred, systemic, or metastatic symptoms had survival rates of 67%, 57%, and 25%, respectively. The composite symptom-severity staging system produced prognostic distinctions within each FICO stage of endometrial cancer. Current staging systems for cancer identify the anatomic form but not the biologic effect or clinical function of the tumor. Although the term "early" is often used to describe anatomic stages such as stage I cervical cancer, it refers to a dimension of time, which is not measured in anatomic systems of staging. Although the
604
Peipert et al.
customary staging systems describe the morphologic appearance and structural damage produced by the tumor, no attention is given to the tumor's duration and rate of growth, which· can be manifested as the functional effects of the cancer in structures or systems that mayor may not be anatomically invaded. I, 2 The results of the current study confirm the hypothesis that clinical variables such as symptoms and comorbidity have important prognostic value in stage IE cervical cancer. These variables indicate aspects of the cancer's function and rate of growth that are not considered in traditional morphologic staging systems. Nevertheless, the morphologic staging systems denote anatomic features of the cancer that cannot be discerned from the clinical variables alone. Consequently, both morphologic and clinical features should be considered to produce adequate descriptions of the biologic behavior of a tumor and effective appraisals of prognosis and therapy. Numerous investigations in chronic diseases and a variety of malignancies have now demonstrated that current methods of classification in cancer staging are incomplete. Anatomic and morphologic systems have classified what a cancer is, but not what it does. 2 Because clinical variables such as symptom status and comorbidity have been shown to influence survival, increase prognostic precision, and improve the evaluation of new therapeutic modalities, these clinical variables should be appraised in gynecologic malignancies. Unless the clinical variables are suitably included, prognostic estimates based only on morphologic condition will be imprecise, and therapeutic evaluations may be misleading. We hope that other investigators will test and validate the importance of clinical variables in cohorts of women with cervical cancer or other gynecologic malignancies. REFERENCES 1. Feinstein AR. Symptoms as an index of biologic behavior and prognosis in human cancer. Nature 1966;209:241-5, 2. Feinstein AR. A new staging system for cancer and a reappraisal of "early" treatment and "cure" by radical surgery. N Engl j Med 1968;279:747-53. 3. Hopkins MP, Morley GW. Stage IB squamous cell cancer of the cervix: clinicopathologic features related to survival. AMj OBSTET GYNECOL 1991;164:1520-9. 4. Hopkins MP, Morley GW. A comparison of adenocarcinoma and squamous cell carcinoma of the cervix. Obstet GynecoI1991;77:912. 5. Gauthier P, Gore I, Shingleton HM. Identification of his-
September 1993 Am J Obstet Gynecol
6. 7. 8.
9.
10. 11.
12. 13. 14. 15. 16. 17, 18. 19. 20. 21. 22. 23. 24.
25.
topathologic risk groups in stage IB squamous cell carcinoma of the cervix. Obstet Gynecol 1985;66:569. Feinstein AR, Wells CK. A clinical-severity staging system for patients with lung cancer. Medicine 1990;69: 1-30. Feinstein AR, SchimpffCR, AndrewsJRjr, et al. Cancer of the larynx: a new staging system and a re-appraisal of prognosis and treatment. j Chron Dis 1977;30:277-305. Feinstein AR, Schimpff CR, Hull EW. A reappraisal of staging and therapy for patients with cancer of the rectum. I. Development of two systems of staging. Arch Intern Med 1975; 135: 1454-62. Feinstein AR, Schimpff CR, Hull EW. A reappraisal of staging and therapy for patients with cancer of the rectum. II. Patterns of presentation and outcome. Arch Intern Med 1975;135: 1454-63. Charlson ME, Feinstein AR. The auxometric dimension: a new method for using rate of growth in prognostic staging of breast cancer. JAMA 1974;228:180-5. Clemens jD, Feinstein AR, Holabird N, Cartwright S. A new clinical-anatomic staging system for evaluating prognosis and treatment of prostate cancer. j Chron Dis 1986;39:913-28, Wells CK, Stoller jK, Feinstein AR, Horwitz RI. Comorbid and clinical determinants of prognosis in endometrial cancer. Arch Intern Med 1984;144:2004-9. Boring CC, Squires TS, Tong T. Cancer statistics, 1993. CA 1993;43:7-26. Photopulos GJ. Surgery or radiation for early cervical cancer. Clin Obstet Gynecol 1990;33:872-82. Welander CE, Homesley HD, Barrett RJ. Combined interferon-a and doxorubicin in the treatment of advanced cervical cancer. AM j OBSTET GYNECOL 1991; 165 :284-91. Panici BP, Scambie F, Baiocchi G, et al. Neoadjuvent chemotherapy and radical surgery in locally advanced cervical cancer. Cancer 1991;67:372-9. Sardij, Sananes C, Giaroli A, et al. Neoadjuvent chemotherapy in locally advanced carcinoma of the cervix uteri. Gynecol Oncol 1990;38:486-93. Coleman DL, Gallup DG, Wolcott HD, et al. Patterns of failure of bulky-barrel carcinomas of the cervix. AM j OBSTET GYNECOL 1992;166:916-20. Feinstein AR. Clinical biostatistics, XIV. The purposes of prognostic stratification. Clin Pharmacol Ther 1972;13: 285-97. Feinstein AR, Pritchett jA, Schimpff CR. The epidemiology of cancer therapy. III. The management of imperfect data. Arch Intern Med 1969;123:448-61. Feinstein AR, Pritchett jA, Schimpff CR. The epidemiology of cancer therapy. IV. The extraction of data from medical records. Arch Intern Med 1969;123:571-90. DiSaia Pj, Creasman WT, Invasive cervical cancer. In: Clinical gynecologic oncology. St. Louis: Mosby-Year Book, 1993:66. Kaplan M, Feinstein AR. The importance of classirying initial co-morbidity in evaluating the outcome of diabetes mellitus. J Chron Dis 1974;27:387-404. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classirying prognostic comorbidity in longitudinal studies: development and validation. j Chron Dis 1987;40:373-83. Cox DR. Regression models and life tables. J R Stat Soc 1972;34:187-220.