papillary and follicular thyroid cancer: prognostic factors in 1,578 patients

Papillary and Follicular Thyroid Cancer

Prognostic Factors in 1,578 Patients

W. J, SIMPSON, M.D., F.R.C.P.(C)

S. E. McKINNEY, M.Math.

J. S. CARRUTHERS, M.D.

M. K. GOSPODAROWICZ, M.D.,

F.R.C.P.(C)

S. B. SUTCLIFFE, M.D., F.R.C.P.(C)

T. PANZARELLA, M.Sc.

Toronto, Ontario, Canada

From the Departments of Radiation Oncology and Pathology, Princess Margaret Hospital, the Departments of Radiology and Pathology, Uni- versity of Toronto, and the Department of Biosta- tistics, Ontario Cancer Institute, Toronto, Ontar- io, Canada. Requests for reprints should be ad- dressed to Dr. W. J. Simpson, Princess Margaret Hospital, 500 Sherbourne Street, Toronto, On- tario, M4X lK9. Manuscript submitted June 20, 1986, and accepted April 22, 1987.

This report from the Canadian survey of thyroid cancer describes 1$74 patients with papillary thyroid cancer and 504 with follicular thyroid cancer followed for four to 24 years. The study groups included more patients with “advanced” disease and fewer with “early” disease than in the general population because these patients were referred to radiotherapy cancer centers, sometimes routinely, but often because referring physiclans believed that certain clinical features indicated the need for additional treatment. Although this report is subject to all the problems of retrospective studies, a careful assess- ment of the pretreatment extent of disease combined with a long follow-up period has allowed an analysis of prognostic factors with considerable confidence. Unlvariate analysis of 12 possible prognostic factors (excluding treatment) demonstrated that nine of them were of statistical significance: postoperative status, age at diagnosis, extrathyroidal invasion, distant metastases, nodal involvement, differentiation, sex, tumor size, and pathologic type (in descending order of importance). Multivariate analysis was carried out using cause-specific survival rates. Independently important prognostic factors at initial treatment were age at diagnosis, extrathyroidal invasion, and degree of differentiation histologically for papillary cancers; and extrathyroidal invasion, distant metastases, primary tumor size, nodal involvement, age at diagnosis, and postoperative status for follicular cancers. The prognostic factors for tumor recurrence were quite different for the papillary and follicular cancers and ranked diierently for the two groups.

Papillary and follicular cancers are the most common malignancies arising in the thyroid gland, and yet their optimal management remains controversial. There are strong proposals for one or another approach, often emphasizing one treatment modality, ignoring the effects of others, and not taking into account the natural history of these diseases (which also remains controversial) or the importance of prognostic factors, which remain, to a large extent, uncertain or unproved.

The Canadian survey of thyroid cancer treatment and outcome, com- missioned by the Canadian Association of Radiologists in 1981, has provided a wealth of data regarding these cancers. Our initial report [I] concentrated on the prognostic factors for all thyroid cancers; this report examines the prognostic factors for papillary and follicular cancers in greater detail.

PATIENTS AND METHODS

Our report is based on 1,074 patients with papillary cancers and 504 patients with follicular cancers, augmented by more detailed information available from the records of 334 patients seen at Princess Margaret

September 1987 The American Journal of Medicine Volume 83 479

PROGNOSTIC FACTORS IN THYROID CANCER-SIMPSON ET AL

TABLE I Age and Sex Distribution for 1,578 Patients with Thyroid Cancer, Subdivided into Those with Grossly Total Tumor Removal and Those with Unresectabie or Incompletely Resected Tumors

Papillary Cancer Follicular Cancer Age Age

No residual plus microscopic residual disease (papillary, 838 patients; follicular, 379 patients)

Gross residual plus unresectable disease (papillary, 79 patients: follicular, 49 patients)

Range/Median Male:Female (years) Ratio

6-90141 1:3.2

lo-90/62.5 1:1.3

Range/Median Wars)

7-90152

25-90/65

Male:Female Ratio

1:2.8

1:1.7

Hospital and included in the larger groups. Although pathologic review was not required for the Canadian survey of thyroid cancer, more than 90 percent of the cases from Princess Margaret Hospital have been reviewed by one of us (J.S.C.). The papillary cancers include both pure papillary tumors and those with a mixed papillary and follicuiar pattern, even when almost ail of the sections examined demonstrated a follicular pattern, provided that other histologic features such as ground-glass nuclei and psammoma bodies were consistent with this diagnosis.

Postoperative status was an important feature that referred to the primary tumor in the thyroid gland, not to nodal disease or distant metastases. “No residual disease” was recorded when the primary tumor was excised with a good margin of uninvolved tissue histologically. “Microscopic residual disease” was noted when the surgeon stated that all visible tumor had been removed but the pathologist reported that tumor was present at, or within 2 mm of, a

TABLE ii Extent of Disease at Diagnosis in 1,578 Patients with Thyroid Cancer

Papillary Follicular Cancer Cancer

(n = 1,074) (n = 504)

Primary tumor size <l cm 11% 5% l-4 cm 67% 53% >4 cm 22% 42% (not stated in 192 and 75 patients respectively)

Extrathyroidal invasion Absent 70% 69% Present 30% 31% (not stated in 148 and 90 patients, respectively)

Nodal involvement Absent 73% 88% Present 27% 12 % (not stated in 266 and 120 patients, respectively)

Distant metastases Absent 94% 87% Present 6% 13% (not stated in 117 and 75 oatients, resoectivelv)

resection margin. In most patients (97 percent of both papillary and follicular groups), cancers that were entirely intrathyroidal were completely removed (both visibly and histologically), but those that invaded adjacent tissues (fat, muscle, nerve, esophagus, trachea, etc.) were much more often considered to have microscopic, residual disease histologically (in 23 percent of papillary and 26 percent of follicular tumors with extrathyroidal invasion). “Gross residual disease” indicated that the surgeon left visible tumor behind, sometimes very small in amount (less than 10 g) but more often very large, or there was no resection at all (biopsy only).

It should be noted that “microscopic residual disease” usually referred to the primary thyroid tumor, judged by the surgeon and pathologist, whereas “gross residual disease” was assessed by clinical examination, chest and skeletal radiography, radioiodine body scanning, and bone scanning in addition to the surgical observations. However, the extent and sensitivity of investigations varied considerably during this long time period. Patients and Treatment. The sex and age distributions are shown in Table I, and Table ii shows the extent of disease at the time of diagnosis for the two groups. Surgical resection was carried out in 94 percent and 90 percent of the patients with papillary and follicular cancers, respectively. Radioio- dine was used in 12 percent and 17 percent, external radiation therapy was used in 14 percent and IO percent, and both modalities (radioiodine plus radiotherapy) were used in 6 percent and 6 percent following the initial surgical procedure.

The extent of the initial surgical procedure(s) was very variable, ranging from “enucleation” of the primary tumor, through lobectomy only, to total or near-total thyroidecto- my. Similarly, nodal surgery ranged from none, to biopsy of one or two nodes, to functional neck dissection or radical neck resection. The use of radioiodine was also variable, ranging from none to the postoperative ablation of residual thyroid tissue, to its therapeutic use in the management of nodal or distant metastases, sometimes without tracer studies to demonstrate that the metastases concentrated radioiodine. Because many patients were treated with radioiodine outside the cancer centers, the details of radioiodine treatment (dates and doses, whether ablative or

480 September 1987 The American Journal of Medicine Volume 83

PROGNOSTIC FACTORS !N THYROID CANCER-SIMPSON ET AL

therapeutic, results of tracer studies) were unavailable in a high proportion of the patients. The details of radioiodine treatment of the patients at Princess Margaret Hospital help to fill this gap. Radiotherapy was also variable, with doses ranging from less than 3,500 cGy over three weeks to more than 6,000 cGy over five or six weeks, delivered to the thyroid bed only, or in volumes so large that they included all cervical and mediastinai lymph nodes. Some patients had a “boost” dose to areas of known or suspected residual gross disease. Chemotherapy, employing a variety of agents, was used in a small number of patients with very advanced disease (local and/or metastatic).

Patients were followed for periods ranging from four to 24 years from the date of registration at one of the radiotherapy centers (January 1958 to December 1978), with the median duration of follow-up from six a@ a half to 10 years for the various treatment groups. For some patients, the interval from diagnosis was much longer, for they were referred at the time of tumor recurrence or after radioiodine therapy was no longer effective. Only 830 of the patients with papillary cancer and 333 of the patients with follicular cancer were referred at the time of initial diagnosis. A separate analysis of those patients who were referred at the time of initial treatment did not result in any significant changes in the ranking of prognostic factors, although their survival was better than that of patients referred after tumor recurrence, For both papillary and follicular cancer, the treatment of recurrent disease was extremely variable and included further surgery (especially lymph node resection), radioiodine, radiotherapy, and (infrequently) chemotherapy. Statistical Methods. Data were stored on the Princess Margaret Hospital’s VAX 8200 computer. Analyses were performed with jn-house programs and with BMDP and SAS packages of statistical programs [2,3]. For tabulated data, the likelihood ratio [4] and Pearson’s chi-square statistics [5] were used. Survival and cause-specific survival curves were estimated by the actuarial [S] and Kaplan-Meier [7] methods. Survival curve comparisons were performed with the log rank statistic and the Cox proportional hazard likelihood ratio statistic [8]. A stepwise selection procedure [9] with the Cox proportional hazards regression model was used to rank prognostic factors by multivariate analyses. All analyses are univariate unless stated otherwise.

RESULTS

Prognostic Factors. Our previous analyses for all types of thyroid cancer reported only on the effect of age at diagnosis, sex, pathologic type, and extrathyroidal invasion as prognostic factors. This report examines the importance of these and other factors specifically for papillary and follicular thyroid cancers as separate groups.

Of the 12 prognostic factors (for overall survival) examined by univariate analysis for all 1,578 patients with papillary and follicular cancers, ranking by p values showed that the most important was postoperative status, as defined in the Patients and Methods section (a factor not reported by others). Age at diagnosis and extrathyroi-

TABLE III Ranking of Prognostic Factors (univariate analysis) for Overall Survival of 1,578 Patients with Thyroid Cancer

Postoperative status (no residual disease versus microscopic residual disease versus gross disease)

Age at diagnosis (<45 versus 45-60 versus >60) Extrathyroidal invasion Distant metastases Nodal involvement Differentiation (well differentiated versus poorly differentiated) Sex Tumor size (<1 cm versus 1-4 cm versus >4 cm) Pathology (papillary versus follicular)

TABLE IV Multivariate Analysis of Prognostic Factors for Cause-Specific Death and Recurrence (local, nodal, or systemic) In 885 Patients with Papillary Thyroid Cancer

Cause-Specific Death (n = 23) Recurrence (n = 89) Prognostic P Prognostic P

Factor

Age* Extrathyroidal

invasion Differentiation

Value

<0.00001 0.0028

0.047

Factor

Age” Nodal

involvement Tumor size*

Value

<0.00001 <0.00001

0.046

Nodal involvement

Metastases Tumor size * Postoperative

status * Sex

0.3

0.44 0.54 0.61

0.76

Metastases

Sex Differentiation Postoperative

status’ Extrathyroidal

invasion

0.18

0.54 0.72 0.89

>O.QO

l age = 560 years versus >60 years; size = 14 cm versus >4 cm; postoperative status = gross disease absent versus present.

dal invasion by tumor were next most important-more so than the presence of distant metastases, which ranked fourth in importance. These four factors proved to be of major importance, with nodal involvement, differentiation, and sex being of intermediate importance, and tumor size and pathologic type of least importance, as indicated in Table III. Multicentricity, bilaterality, and prior irradiation proved to be of no prognostic significance.

However, when we re-examined our data by multivariate analysis using cause-specific death as the end-point, a different picture emerged. As shown in Table IV, only age at diagnosis, extrathyroidal invasion, and differentiation were independent prognostic factors for survival among patients with papillary cancer; however, age, nodal involvement, and primary tumor size were the only independent prognostic indicators of tumor recurrence



TABLE V Multivariate Analysis of Prognostic Factors for Cause-Specific Death and Recurrence in 253 Patients with Follicular Thyroid Cancer

Cause-Specific Death (n = 28) Recurrence (n = 45) Prognosfic P Prognosfic P

Factor Value Factor Value

Extrathyroidal invasion

Metastases Size

Nodal involvement

Age Postoperative

status (gross)

Postoperative status (microscopic)

Differentiation Sex

<O.OOOOl Metastases

0.0001 0.0008

0.002

0.014 0.025

Age Postoperative

status (microscopic)

Nodal involvement

Size

0.19 Extrathyroidal invasion

0.73 0.86

Sex Differentiation Postoperative

status (gross)

<0.00001

0.0002 0.0002

0.037

0.036

0.25

0.43 0.82 0.90

Note the different ranking and the different factors involved in survival compared with recurrence (as in patients with papillary cancer).

TABLE VI Relationship of Extrathyroidal Invasion and Postoperative Status

No Invasion Invasion Total

Papillary cancer No residual disease 455 (89 %) 83 (49 %) 538 Microscopic residual disease 55 (11%) 85 (51%) 140

510 168 678 Follicular cancer

No residual disease 199 (91%) 40 (64%) 239 Microscopic residual disease 20 (9 %) 23 (36%) 43

219 63 282

p = 0.0005 for both groups.

(local, nodal, or distant metastases). A similar analysis among patients with follicular cancer showed that there were still six independent prognostic factors for survival, the most important being extrathyroidal invasion, closely followed by distant metastases and primary tumor size, with nodal involvement, age at diagnosis, and the presence of gross residual disease of lesser importance, but still significant and independent of the other prognostic factors. Again, factors predictive of recurrence ranked differently in this multivariate analysis (Table V).

Prognostic factors were inter-related to some extent,

but not entirely. For example, older patients were more likely to have tumors with extrathyroidal invasion and residual disease postoperatively than younger patients, and those with extrathyroidal invasion more often had residual disease postoperatively than those without (Table VI).

Figure 1 illustrates the influence of postoperative status on cause-specific survival: patients with papillary cancer who had no residual disease had better survival rates than those who had microscopic residual disease, and those who had gross disease but no distant metastases fared worse (p = 0.0002). Figure 2 demonstrates a similar pattern for patients with follicular cancer. The presence of distant metastases increased the probability of a fatal outcome (p -CO.0005 for both papillary and follicular cancers); nevertheless, 20 to 30 percent of patients who had distant metastases appeared to be “cured” (usually by radioiodine) as shown by the flat survival curves beyond IO years.

Age at diagnosis was a potent prognostic factor for cause-specific survival (including only deaths due to thyroid cancer or treatment complications, “censoring” patients who died of other causes), especially in patients with papillary cancer (Figure 3) but also in those with follicular cancer (Figure 4). The overall survival rates among patients with papillary cancer were 92 percent, 64 percent, and 12 percent for patients younger than 40 years, 40 to 59 years old, and more than 60 years old, respectively at 20 years, whereas the 20-year cause- specific rates were 95 percent, 84 percent, and 39 percent. Among patients with follicular cancer, the overall survival rates were 85 percent, 57 percent, and 13 percent, whereas the cause-specific rates were 88 percent, 6 1 percent, and 32 percent. Although these curves show only three age groupings, we found a steadily worsening survival rate after the age of 40 years when groupings by decade were examined.

Extrathyroidal invasion was also a potent prognostic factor, as illustrated for papillary and follicular cancers in Figures 5 and 6; patients whose tumors invaded adjacent tissues-fat, muscle, esophagus, trachea, major nerves, or arteries-had much poorer survival rates than those whose cancers did not. These figures also indicate that nodal involvement was of importance, although much less so than extrathyroidal invasion. When analyzed by the presence or absence of extrathyroidal invasion, involvement of regional lymph nodes predicted a poorer outcome of statistical significance in patients with follicular cancer and those with papillary cancer who had no extrathyroidal invasion. The presence of distant metastases (Figures 1 and 2) was also a major indicator of poor survival, although not nearly as strong a factor as it is in most other malignancies, for which distant metastases almost invari- ably predict a fatal outcome.



---- Micro. residual dwase 80 - (153 pts)

2 1

: . Gross residual disease

$ : i (no mefastases, 23 pts)

Y) 60' ‘0

p’ 0.00001

,* .$

5 40-

i

I , J 0 5 IO 15 20

Time iyears) ISSlR

ivgure 7. mfluence of postoperative status on the cause- specific survival of patients with papillary thyroid cancer (highly significant differences).

rgure 3. influence of age on the cause-specific survival Figure 4. Influence of age on the cause-specific survival of patients with papillary thyroid cancer. of patients with follicular thyroid cancer.

Comparison of the survival among patients with papillary and follicular cancer who had well differentiated tumors and those who had moderately or poorly differentiated tumors (Figures 7 and 8) demonstrated improved survival for those who had well differentiated tumors (p <0.0005). In papillary cancers, differentiation proved to be an important prognostic factor (Table VII), even after adjustment for age, extrathyroidal invasion, and postoperative status (p = 0.003). Similarly, females had better survival rates than males (Figures 9 and 10).

The size of the primary tumor had only a slight influence on survival, with poorer survival rates as tumors

No res,duol d,sease (263 pts)

i

I i I 0 5 IO I5 20

Time (years) I6WR L J

Figure 2. Influence ofpostoperative status OR the survival of patients with follicular thyroid cancer (highly significant differences).

increased in size (Figures 11 and 12). The poor survival of patients with follicular cancer who had very small primary tumors (less than 1 cm in diameter) will be discussed later. When other factors were taken into account, the pathologic type (papillary versus follicular) also had only a small influence on survival (Figure 13), rating least important among the nine factors studied by univariate analysis.

Microscopic vascular invasion was not one of the items requested in the Canadian survey, but this information is available for the patients in the survey from Princess Margaret Hospital. When patients with papillary and follic-

Sqptember 1997 The American Journal of Medicine Volume 93 433


Figure 5. Influence of extrathyroidal invasion and nodal involvement on the cause-specific survival of patients with papillary thyroid cancer. Note the much greater effect of extrathyroidal invasion compared with nodal involvement.

Well differentmted (623ptsl

4 3

60.

4 Mod. or poorly differentiated

b (85 pts1 d B 40.

P Y

d 20.

L Figure 8. Influence of extrarhyroidal invkion and nodal involvement on the cause-specific survival of patients with follicular thyroid cancer.

Well differentiated (224 pts)

Mod. + poorly differentiated _ (95 *ts)

I I I 0 5 IO 15 20

Time (years) NS61R

Figure 7. Influence of differentiation on the cause-specific survival of patients with papillary thyroid cancer.

Figure 8. Influence of differentiation on the cause-sDecif- ic survival of patients with follicolar thyroid cancer.

L I I I 0 5 IO 15 20

Time (years1 /IS//)

Figure 9. Influence of sex on the cause-specific survival of patients with papillary thyroid cancer.

Females (369 pisl _

Moles (133 ptsl

I 0 5 IO 15 20

Time (years) 16611

Figure 10. Influence of sex on the cause-specific survival of patients with follicular thyroid cancer.


Figure 11. Influence of primary tumor size on the cause- specific survival of patients with papillary thyroid cancer.

TABLE VII Influence of Differentiation on Cause- Specific Survival In Patients with Papillary Cancer, Adjusted for Age, Extrathyroidal Invasion, and Postoperative Status

Number of Deaths/ Number of Patients

Well differentiated 29146 1 Moderately or poorly differentiated 9146

p = 0.033.

TABLE VIII Influence on Vascular Invasion on Cause-Speclflc Survival In Patients from Princess Margaret Hospital

Number of Deaths/Number of Patients Papillary Follicular Cancer Cancer

Invasion absent 14/117 (12%) 7/33 (21%) Invasion present 9/40 (23 %) 29167 (43 %)

p = 0.17 for papillary cancer and 0.049 for follicular cancer.

ular were pooled, better survival of patients without vascular invasion was found (p = 0.004). When papillary and follicular cancers were analyzed separately (Table VIII), this difference disappeared for papillary cancers (p = 0.38) but not for follicular cancers (p = 0.03). It should be noted that only 25 percent of papillary cancers exhibited vascular invasion, whereas 67 percent of follicular cancers were found to have vascular invasion. When adjust- ments were made for both pathologic type and differentiation (well versus moderately or poorly differentiated), there were no significant survival differences (p = 0.25). It appears that microscopic vascular invasion is not a significant prognostic factor as it is intimately associated with pathologic type and degree of differentiation.


0 5 IO 15 20

Time (years) /11.* A

Figure 72. Influence of primary tumor size on the cause- specific survival of patients with follicular thyroid cancer. Note the very small number of patients with small tumors (less than 7 cm) (see text).

100 7: :*. ‘----I- . . . Pure Papillary (39 ptsl

:.

p : so- :..... L. :.....

s :i..

‘-\-+- ----

‘0 6. p<o.o005 Papillary-Follicular (165pls)

3 * :.....

60- : . Foliicular * . :~~:.K (129 pts)

8 B

:....:......

x i.

2 p”

40- i...

20 ” s “‘I ‘E 0 r c ” c “1 1 0 5 IO 15 20

Time (years1 ,.I,8 -I

Figure 13. Influence of pathologic type on cause-specific survival.

The foregoing results were obtained by univariate analysis of all patients for whom the specific attributes were recorded. Thus, for different analyses, the number of patients varied considerably, a drawback encountered in most retrospective studies. For some analyses, e.g., age and sex, the frequency of omission of these attributes was less than 1 percent; however, for others (tumor size, extrathyroidal invasion, etc.), information was lacking for 10 to 20 percent of patients. Thus, when we attempted multivariate analysis of all nine factors shown to be of statistical significance prognostically, only a small proportion of patients (about 10 percent) had recorded information for every factor being analyzed. Accordingly, we used a statistical method [9] that included patients whose re-



cords contained most of the needed information, placing the patients with missing information in the “neutral” or “median” position with respect to the missing data. This resulted in the inclusion of two thirds of the patients with papillary cancer and one half of the patients with follicular cancer in our multivariate analysis of prognostic factors, shown in Tables IV and V. It is noteworthy that prognostic factors were different or ranked differently not only for papillary and follicular tumors, but also for death compared with recurrence within each pathologic type. Treatment and Treatment Complications. Treatment consisted of surgery, thyroid hormone, radioiodine, radiotherapy, and in a few patients, chemotherapy. A detailed analysis of the various treatment modalities, their complications, and their influence on local control and survival is reported separately. Cause of Death. As we reported previously, the majority of deaths were due to thyroid cancer: 55 percent of the 226 deaths among patients with papillary cancer and 66 percent of the 198 deaths among patients with follicular cancer. An additional four deaths in each group were the result of treatment complications. Occasional deaths caused by thyroid cancer occurred as long as 20 years or more after initial diagnosis and treatment.

COMMENTS

Because our report is based on patients referred to radiotherapy centers, it includes a much higher proportion of patients with a poor prognosis than is found in reports based on all patients with thyroid cancer. This is reflected in the older median age at diagnosis of our patients and the high proportion of patients with a follicular histologic type. Thus, direct comparisons with results in other studies are impossible; on the other hand, the higher proportion of patients with a poor prognosis (a relatively small proportion of all patients with thyroid cancer) permitted us to analyze prognostic factors with greater ease and reli- ability than if most patients had had good prognostic factors.

For many years, thyroidologists have recognized the increasingly adverse effect of advancing age at diagnosis on the prognosis of patients with papillary or follicular cancer. In recent years, this opinion has been confirmed by reports based on large numbers of patients: Mazzaferri and co-workers [ 10,111, Tubiana et al [ 121, Carcangiu et al [13], and others. Our data not only confirm this stand but also demonstrate that age at diagnosis is the most important influence on cause-specific survival among patients with papillary cancer, although it is much less important (but still statistically significant) among patients with follicular cancer. However, we could not substantiate the claim that there was an abrupt worsening of prognosis after the age of 45 years, as claimed by Tubiana et al [ 121.

When analyzed decade by decade, our patients with papillary cancer demonstrated a steadily worsening survival with increasing age, with a much better survival of patients younger than 40 years of age compared with those in their 5Os, and with a second point of markedly worse survival after the age of 60 years. For patients with follicular cancer, the cut-off points occurred about a decade earlier. In other reports [ 12-151, sex was reported to be a prominent prognostic factor. We also found sex to be of prognostic importance-males fared poorly compared with females-but it was not of major importance, ranking only seventh in importance among nine prognostic factors in univariate analysis and not reaching statistical significance in multivariate analysis,

As others have reported [ 13-171, extrathyroidal invasion proved to be an important prognostic factor. It was the most important prognostic factor for survival among patients with follicular cancer and the second most important among patients with papillary cancer. Interestingly, extrathyroidal invasion was not a predictor of recurrence-perhaps because a higher proportion of these patients are never free of local and/or nodal disease or distant metastases compared with those with no extrathyroidal invasion, or because adjuvant treatment pre- vents local recurrence but not distant metastases. We recognized that patients with extrathyroidal invasion also had a higher frequency of other poor prognostic factors: larger primary tumor size, greater age, less frequent responses to specific therapies, more frequent distant metastases, etc. We agree with Tsumori [ 181 and others that these patients require more aggressive treatment initially than those without extrathyroidal invasion.

In contrast, nodal involvement was a much weaker prognostic factor. By univariate analysis, it did not influence the survival of patients with papillary cancer and ranked lower than extrathyrokfal invasion, distant metastases, and primary tumor size for patients with follicular cancer. The fact that nodal involvement was a potent predictor of tumor recurrence but unimportant for survival suggests that the management of nodal recurrence, usually surgical, is very successful, especially for papillary thyroid cancers.

When we compared patients with and without nodal involvement, we found no significant survival differences for papillary or follicular cancers, but when we analyzed patients with and without extrathyroidal invasion by tumor as separate groups, nodal involvement demonstrated a deleterious influence (Figures 5 and 6). This observation may explain the discrepancies between reports claiming that nodal involvement did [ 12,14,19] or did not [10,11,13,15,17]influencesurvival;Cadyetal[16]even reported that nodal involvement conferred a survival ad- vantage, no doubt reflecting a large proportion of young patients (with an excellent prognosis) presenting with

48% September 1987 The American Journal of Medicine Volume 83

nodal me&stases from papillary primary tumors. Groups of patients in which equal proportions of tumors with extrathyroidal invasion were included among patients with and without nodal involvement would show better survival rates for the patients without nodal involvement; however, if there were more patients with extrathyrokfal invasion but no nodal involvement than patients without extrathyroidal invasion but with involved nodes, the two influences would negate each other, resulting in no significant survival differences. It is likely that Cady et al’s [ 161 study group included a large proportion of young patients without extrathyroidal invasion presenting with nodal disease.

Distant metastases proved not to be significant for cause-specific survival among patients with papillary cancer. This may reflect infrequent distant metastases (which are usually pulmonary when they do occur), the ability of younger patients to live with them for many years (beyond the 20-year follow-up period of this study), and the eradi- cation of pulmonary metastases by radioiodine in some patients. In contrast, distant metastases were the second most important predictor of survival among patients with follicular cancer, in whom they are more frequent and for whom long-term survival in the presence of distant metastases is unusual unless the metastases are eradicated by radio&line-an uncommon event. In differentiated thyroid cancers, distant metastases may be eradicated by radioiodine if the tumor burden is not great and the tumor concentrates radioiodine well; otherwise, symptoms may be relieved temporarily, but survival may not be prolonged significantly. Even when gross tumor disappears completely on clinical, radiologic, and radioisotopic study, the tumor will recur eventually in most patients except when the metastatic tumor is microscopic (e.g., normal chest radiographic appearance but diffuse lung concentration of radioiodine by tumor, usually papillary). The ability of radioiodine to eradicate metastatic disease in some patients may explain the low ranking of metastatic disease among the prognostic factors for papillary cancers.

The degree of tumor differentiation was of prognostic significance for papillary cancers (p = 0.047) but not for follicular cancers (p = 0.82). Moderately or poorly differentiated elements were found only in the “follicular” components of mixed papillary-follicular tumors, never in the “papillary” components, and were noted in only 10 percent of all mixed papillary-follicular tumors. It might therefore be argued that the distinction between “papillary” and “mixed papillary-follicular” tumors is valid. However, since the prognostic importance of poor differentiation is barely significant statistically, we doubt the need to make this distinction since the majority of “mixed” tumors behave clinically much more like “papillary” than like “follicular” cancers (Figure 13). In contrast to the major importance given to poorly differentiated


tumors by Sakamoto et al [20], we found it to be a relatively weak prognostic factor compared with the others we examined. It is interesting to speculate on why differentiation is not a significant prognostic factor for follicular cancers, although half of them are moderately or poorly differentiated. Perhaps it is because the poorly differentiated tumors concentrate radioiodine more often: of the 11 well differentiated tumors with demonstrable nodal or distant metastases in the patients we studied from Princess Margaret Hospital, seven concentrated radio&line sufficiently to be useful, whereas four of the five poorly differentiated tumors did so. On the other hand, it may be because radioiodine rarely eradicated the disease permanently in any of these patients.

In most reports [ 12,14,15] on differentiated thyroid cancer, males have a poorer survival then females. Our data confirmed this on univariate analysis, but gender was of much less importance than other factors and ranked last on multivariate analysis (p = 0.76 and 0.86). The influence of the size of the primary tumor on survival among patients with papillary cancer (Figures 11 and 12) was negligible for tumors less than 4 cm in diameter. There was no difference between those less than 1 cm and those 1 to 4 cm, probably because the former usually present with neck node metastases in young patients, whereas the latter usually come to medical attention because of the primary tumor. The poorer survival of patients with cancers larger than 4 cm compared with smaller ones is consonant with experience with other malignancies. The poor survival among the very small number of patients with tiny (less than 1 cm) follicular cancers probably reflects the high proportion presenting with symptoms caused by distant metastases, especially bone pain or spinal cord compression, but impalpable primary tumors. The survival of patients with tumors larger than 4 cm in diameter is significantly worse than that of patients with smaller tumors, as expected.

The distinction between papillary and follicular thyroid cancers may be of little more than academic interest, contrary to the view expressed by Fransilla [2 11, because this factor rated lowest in our univariate analyses (ranked by p values). However, we believe that the differences in importance of other prognostic factors make it essential for us to continue to regard them as separate entities until future data suggest a m-examination of their biologic behavior for similarities and differences.

These analyses indicate the need for a better system of assessing patients with differentiated thyroid cancer in order to improve treatment recommendations Byar and colleagues [ 141 attempted to do this by providing a numer- ical score for each of the prognostic factors they identified. Their system has not been verified in a different group of patients [22]; moreover, it is so cumbersome that it is unlikely to enjoy widespread use. In its place, we



will present a “decision tree” in the secortd part of this report, in which the most importartt prognostic factors we identified are used to determine further treatment.

In conclusion, the Canadian survey of thyroid Cancer has produced new information about the prognostic factors affecting patients with papillary and follicular thyroid Cancer. Although we have confirmed the importance of most of the prognostic factors reported previously, we have also determined their relative importance. For in- stance, we have reaffirmed the importance of age at diagnosis and extrathyroidal invasion as prognostic factors of great importance, whereas sex is very much less

1.

2.

3.

4.

5.

6.

7.

8.

9.

IO.

11.

12.

important. Similarly, nodal involvement influences sunfiv- al mainly in the absence of extrathyroidal involvement, which influences survival to a much greater extent th&n does nodal involvement. All of these prognostic factors and their relative importance should be kept in mind when postoperative treatment is under consideration.

ACKNOWLEDGMENT

We thank our Colleagues in 12 Canadian centers for providing the data on which this report is based, and Mrs. Karen Rice, Ms. Evelyn Eisenreich, and Ms. Diane Cart- wright for their secretarial and technical assistance.

REFERENCES

Simpson WJ, McKinney S: Canadian survey of thyroid cancer. Can Med Assoc J 1985; 132: 925-931.

Dixon WJ, ed: BMDP statistical software, Berkeley, Califor- nia: University of California Press, 1981; 143-206, 555- 594.

Joyner SP, ed: SAS user’s guide: statistics, 5th ed. Cary, North Carolina: SAS Institute, 1965; 529-558.

Kalbfleisch JG: Probability and statistical inference, vol 2. New York: Springer-Verlag. 1979; 146-170.

Mendenhall W, Schaeffer R: Mathematical statistics with applications, North Scituate, Massachusetts: Duxbury Press, 1973; 501-516.

Cutler SJ, Ederer F: Maximum utilization of the life table method in analysing survival. J Chronic Dis 1958; 8: 699- 712.

Peto R, Pike MC, Armitage P, et al: Design and analysis of randomized clinical trials requiring prolonged observation of each patient. Part 2. Analysis and examples. Br J Cancer 1977; 35: l-39.

Kalbfleisch JD, Prentice RL: The statistical analysis of failure time data. New York: Wiley, 1980; 70-84.

Dixon WJ, ed: BMDP statistical software. Berkeley, Califor- nia: University of California Press, 1983; 587-590.

Mazzaferri EL, Young RL, Oertel JE, et al: Papillary thyroid carcinoma: the impact of therapy in 576 patients. Medi- cine (Baltimore) 1977; 56: 171-196.

Young RL, Mazzaferri EL, Rahe AJ, Dorfman SG: Pure follicular thyroid carcinoma: impact of therapy in 214 patients. J Nucl Med 1960; 21: 733-737.

Tubiana M, Schlumberger M, Rougier P, et al: Long term results and prognostic factors in patients with differentiat-

13.

14.

15.

16.

17.

16.

19.

20.

21.

22.

ed thyroid carcinoma. Cancer 1965; 55: 794-604. Carcangiu ML, Zampi G, Pupi A, et al: Papillary carcinoma of

the thyroid. Cancer 1985; 55: 605-628. Byar DP, Green SB, Dor P, et al: A prognostic index for

thyroid carcinoma. A study of the E.O.R.T.C. thyroid cancer cooperative group. Eur J Cancer 1979; 15: 1033- 1041.

Woolner LB: Thyroid carcinoma: pathologic classification with data on prognosis. Semin Nucl Med 1971; 1: 461- 502.

Cady B, Sedgwick CE, Meissner WA, et al: Risk factor analysis in differentiated thyroid cancer. Cancer 1979; 43: 810-820.

Samaan NA, Mahesh Wary YK, Nader S, et al: Impact of therapy for differentiated carcinoma of the thyroid: an analysis of 706 cases. J Clin Endocrinol Metab 1983; 56: 1131-I 137.

Tsumori T, Nakao K, Miyata M, et al: Clinicopathologic study of thyroid carcinoma infiltrating the trachea. Cancer 1985; 56: 2643-2848.

Harwood J, Clark OH, Dunphy JE: Significance of lymph node metastasis in differentiated thyroid cancer. Am J Surg 1978; 136: 107-110.

Sakamoto A, Kasai N, Sugano H: Poorly differentiated carcinoma of the thyroid. Cancer 1983; 52: 1849-1855.

Franssila KO: Is the differentiation between papillary and follicular thyroid carcinoma valid? Cancer 1973; 32: 853- 864.

Tennvall J, Biorklund A, Moller T, et al: Is the E.O.R.T.C. prognostic index of thyroid cancer valid In differentiated thyroid carcinoma? Cancer 1986; 57: 1405-1414.


papillary and follicular thyroid cancer: prognostic factors in 1,578 patients

Documents