Poorly differentiated carcinomas of the thyroid with trabecular, insular, and solid patterns

A clinicopathologic study of 183 patients




The term poorly differentiated (PD) carcinoma was proposed 20 years ago to define aggressive, follicular-derived thyroid carcinomas with behavior intermediate between follicular/papillary and anaplastic carcinomas. Among the variable histologic patterns recognized in such tumors, trabecular-insular-solid (TIS) areas usually are predominant. Conversely, some authors pointed out that PD carcinomas are characterized by unequivocal, high-grade histology with atypias, high mitotic counts, and necrosis rather than by a specific growth pattern.


The clinicopathologic features of a series of 183 thyroid carcinomas with predominant (n = 165 tumors) or focal (n = 18 tumors) TIS growth patterns were studied by univariate and multivariate overall survival analyses and were compared with clinical outcomes. Subgroups included tumors with predominant oxyphilic features (n = 66 tumors) and (residual) papillary carcinoma features (n = 24 tumors). Control groups of papillary (n = 68 tumors), follicular (n = 71 tumors), and anaplastic (n = 35 tumors) carcinomas also were included for overall survival analysis.


TIS carcinomas had an intermediate behavior between papillary/follicular and anaplastic carcinomas (P < 0.0001). Univariate and multivariate statistical analyses demonstrated that age > 45 years (P = 0.007), the presence of necrosis (P < 0.0001), and a mitotic count > 3 per 10 high-power fields (P = 0.01) were associated with poor outcome. A simplified scoring system based on statistically significant parameters allowed the identification of three prognostic subgroups (P < 0.0001).


PD TIS carcinomas overall followed a more aggressive course compared with differentiated thyroid carcinomas, irrespective of the extent of the TIS component. However, a numeric scoring system applied to specific clinicopathologic parameters further may identify three prognostic categories of patients who have significantly different survival rates at 5 years and 10 years. Cancer 2004;100:950–7. © 2004 American Cancer Society.

In the past 20 years, several reports have described nonfollicular, nonpapillary thyroid carcinomas, which produced thyroglobulin and were unrelated to anaplastic carcinoma. Different terms were selected to refer to these tumors, including solid,1 trabecular,2 insular,3, 4 poorly differentiated (PD),1, 5, 6 intermediate type,7 primordial cell,8 less well differentiated,9 and follicular carcinoma with insular component.10 Although none of those terms appears to describe this group of tumors appropriately, they are referred to generally as “poorly differentiated thyroid carcinomas”,11–14 a term that defines malignant neoplasms with different histologic patterns of growth, basically lacking follicular or papillary areas, and with clinical and morphologic features that lie in between the well differentiated (follicular and papillary) carcinomas and the highly malignant, thyroglobulin negative, anaplastic carcinoma.1, 3, 8, 11, 15

Some authors have criticized the idea of identifying a specific tumor subtype based on the growth pattern only and have suggested that the observed aggressive behavior of PD carcinoma may be related to a higher disease stage at presentation or to the presence of high-grade features within the tumor (atypias, high mitotic count, extensive necrosis).11, 14, 16 Consequently, there are several different criteria that define PD carcinomas, and most published series are not compared easily.

Additional sources of controversies on PD carcinomas including the following.

Some studies,6, 8, 17 after the original observation of Carcangiu et al.,3 incorporated both follicular-derived tumors and papillary-derived tumors into the PD carcinoma group. Others18–20 considered solid papillary carcinomas separate from PD carcinomas unless necrosis was present within the tumor.20 Those authors showed that solid papillary carcinoma carried a better prognosis compared with PD carcinoma and had slightly more aggressive behavior compared with conventional (well differentiated) papillary carcinoma; therefore, solid papillary carcinomas were a variant of the latter (rather than part of the former). Therefore, trabecular-insular-solid (TIS) carcinomas with papillary carcinoma nuclear features either were or were not included in different published PD carcinomas series to date. Aggressive variants of papillary carcinoma (e.g., tall cell or columnar cell carcinoma) occasionally were incorporated into the group of PD carcinoma.11, 19 Oxyphilic (Hürthle cell) carcinomas long have been considered a separate entity based on their apparently higher malignant potential; therefore, they were excluded from several studies. Others21–23 showed that oxyphilic tumors differed in no way from conventional follicular carcinomas or even solid-trabecular carcinomas, except for their mitochondrion-rich cytoplasm. The extent of TIS areas required to label a tumor as a PD carcinoma differed remarkably in different series, thus making comparisons very hard among currently published studies. Although most authors agree that PD carcinoma is defined by the presence of ≥ 75% TIS areas, other studies in patients with PD carcinoma found that insular or solid components comprised as low as 10% of tumors.6, 10

In a former study from a high-risk mountain area, we analyzed 63 TIS carcinomas, approximately 50% of which were made of small, uniform cells, strongly reminiscent primordial cords of the developing thyroid.8 The growth patterns were heterogeneous, resembling either Sakamoto type solid carcinomas,1 trabecular carcinomas,2 insular carcinomas, or follicular carcinomas with variable degrees of insular component. Extensive vascular invasion, high mitotic count, and necrosis were common features, especially in patients with clinically aggressive disease. Recently, some authors24, 25 analyzed the influence of histologic parameters such as atypias, mitoses, and necrosis on the prognosis of patients with well differentiated papillary carcinoma. They showed that malignant behavior was associated significantly with the presence of such high-grade features rather than related to the pattern of growth. Because some PD carcinomas have a fatal outcome and others follow an indolent course, we decided to apply the same approach. The influence of these same high-grade histologic parameters on prognosis was analyzed in a large series of patients with TIS carcinomas of the thyroid, in which these patterns of growth were represented more or less extensively (165 patients) or were recognized only focally (18 remaining patients).

This report shows that, in patients with TIS carcinomas who are age > 45 years, the presence of necrosis and a mitotic index > 3 affected prognosis significantly. The classification of TIS carcinomas into a separate group (also encompassing solid-trabecular oxyphilic carcinomas) appears justified, irrespective of the label used for these tumors.


Patient Selection

In all, 183 patients with of nonmedullary thyroid carcinoma that had TIS growth patterns were selected from a series of 2900 patients with malignant thyroid tumors who underwent surgery during the same period (1960–2002) from the files of the Departments of Pathology of the University of Turin and of the Mauriziano Hospital. Some of these patients had been included in previous studies.8, 21 All selected patients had been reevaluated histologically by four independent investigators (M.V., M.P., M.M., and B.T.). The criteria of inclusion consisted included the presence of either focal (10–40%), extensive, or predominant TIS growth patterns in follicular-derived (thyroglobulin positive) malignant tumors (overt capsular and/or vascular invasion in peritumoral blood vessels). Four patients who met these criteria but who had an anaplastic carcinoma focus were excluded from the current study. In this series, 159 tumors demonstrated a more or less extensive follicular component, whereas 24 tumors were associated with or showed the cytologic criteria of papillary carcinoma. Four to 15 hematoxylin and eosin-stained slides were available from each patient. On review, the following pathologic parameters were evaluated and recorded: extent of TIS growth pattern, defined as the percentage of the tumor area comprised of a tumor cell population arranged in nests or islands separated by a thin, fibrovascular stroma and artifactual clefts (insular pattern) or in a solid, uniform, trabecular, or alveolar fashion (solid-trabecular pattern); absent, focal (< 5% of the tumor area), or extensive (> 5% of the tumor area) necrosis; the presence of oxyphilic (Hürthle cell) features, defined as the presence in the majority (> 75%) of tumor cells of deeply eosinophilic, granular cytoplasm; mitotic count per 10 high-power fields (HPF); small (uniform cells with scant and poorly defined cytoplasm) or large (cells with an abundant eosinophilic or clear cytoplasm) cell size; and minimal (present in < 3 blood vessels outside the tumor capsule) or extensive (present in > 3 blood vessels outside the tumor capsule) vascular invasion.

For all 183 patients, information regarding gender, age, tumor size and stage, therapy administered, and outcomes were collected. Disease persistence and/or recurrence were evaluated by means of either elevated serum thyroglobulin levels, imaging studies, or histopathologic examination of resected recurrent tumors. In addition, 68 patients with conventional papillary carcinomas, 71 patients with follicular carcinomas, and 35 patients with anaplastic carcinomas who had complete follow-up information were used as control groups for the overall survival analysis.

Statistical Analysis

Univariate survival analysis were based on the Kaplan–Meier product-limit estimate of overall survival distribution; differences between survival curves were tested using the generalized Wilcoxon test. The relative importance of multiple prognostic factors on survival was estimated using the Cox proportional hazards regression model. All data were analyzed with BMDP (Berkeley, CA) statistical software (programs 1L and 2L). A level of P < 0.05 was considered statistically significant.


A summary of the results concerning the main clinicopathologic parameters is provided in Table 1. The most relevant clinical and pathologic features are described briefly.

Table 1. Clinicopathologic Features of 183 Trabecular-Insular-Solid Thyroid Carcinomas
CharacteristicNo. of patients (%)
  1. TIS: trabecular-insular-solid; NED: no evidence of disease; AWD: alive with disease; DOC: died of other causes; DOD: died of disease.

Gender (male/female)56/127
Age (yrs) (mean)56.3
Tumor size (cm) (mean)5.34
Associated goiter110/183 (60)
Predominant pattern 
 Insular 92/183 (50)
 Solid-trabecular 91/183 (50)
 TIS pattern (%) 
  Focal (< 50%) 18/183 (10)
  Predominant (50–75%) 42/183 (23)
  Pure (> 75%)123/183 (67)
Presence of necrosis (focal or extensive) 86/183 (47)
Presence of oxyphilic features 66/183 (36)
Mean mitotic count (range)2.16 (0–37)
Cell size 
 Small 87/183 (47)
 Large 96/183 (52)
Wide vascular invasion 96/178 (54)
Radioiodine therapy 83/100 (83)
 NED 91 (50)
 AWD 37 (20)
 DOC 13 (7)
 DOD 42 (23)
5-yr overall survival rate (%)85
10-yr overall survival rate (%)67

Clinical Parameters

A remarkable female predominance was observed (male:female ratio, 2:26), as expected. The mean patient age was 57 years. The age of 45 years, as proposed in the sixth edition of the TNM classification system,26 was considered the cut-off age for statistical analyses. Nodular goiter was found in nearly 60% of patients. Radioiodine therapy was given (1–8 doses of 131I between 3700 and 7400 megabecquerels) after total thyroidectomy in 83% of patients for whom postsurgical treatment information was available.

Clinically aggressive tumors, including those that caused recurrences or metastatic spread and/or patient deaths, accounted for 79 of 183 tumors (43%). The group of patients with TIS carcinomas had a mean 5-year survival rate of 85% and a 10-year survival rate of 67%.

Pathologic Parameters

The mean tumor size was 5.34 cm; for statistical analyses, a greatest tumor dimension of 4 cm, representing the limit between pathologic T2 (pT2) and pT3 disease according to the American Joint Committee on Cancer staging manual,27 was considered the cut-off size. All patients had clear-cut signs of malignancy (i.e., vascular invasion that was either minimal [82 tumors] or wide [96 tumors], which was present in all but 5 patients). The latter five patients had extensive capsular penetration only. With regard to the TIS growth pattern, in the majority (165) of the tumors, it represented the vast majority of the tumor areas, with insular features found to be predominant in 50% of tumors and solid-trabecular architecture predominant in the other 50% of tumors. Follicles were found in variable amounts in 159 tumors (87%), whereas papillary carcinoma features were associated with the remaining 24 tumors (13%). Predominant oxyphilic features were observed in 66 tumors (36%). Necrotic areas were present in 86 tumors (47%) and were either focal (43 tumors) or extensive (43 tumors). The mean mitotic count (per 10 HPF) was 2.16 (range, 0–37). Half of the tumors were comprised of large cells, and the other half (mostly corresponding to insular carcinomas) contained medium to small-sized cells.

Statistical Analysis

Patients who had TIS carcinomas had an overall survival that was intermediate between the survival of patients in the well differentiated follicular/papillary carcinoma and anaplastic carcinoma control groups (P < 0.0001) (Fig. 1A). The data on univariate analysis of the considered clinicopathologic parameters are reported in Table 2, and the curves of statistically significant parameters are shown in Figure 2.

Figure 1.

(A) Overall survival distribution based on Kaplan–Meier product-limit estimates of trabecular-insular-solid carcinoma (TISC) compared with papillary carcinoma (PTC), follicular carcinoma (FC), and anaplastic carcinoma (AC) control groups. (B) Overall survival distribution based on Kaplan–Meier product-limit estimates of TISC grouped into three categories according to a proposed scoring system, which was based on clinicopathologic parameters that were relevant in multivariate analysis compared with PTC, FC, and AC control groups.

Table 2. Correlation between Clinicopathologic Parameters and Overall Survival in 183 Patients with Trabecular-Insular-Solid Thyroid Carcinomas
VariableNo. of patients5-yr survival rate (%)10-yr survival rate (%)P value
  1. TIS: trabecular-insular-solid; HPF: high-power fields.

 ≤ 45 yrs4595830.007
 > 45 yrs1348362
Year of diagnosis    
 ≤ 19919279650.08
 > 1992919372
Greatest tumor dimension    
 ≤ 4 cm6792660.09
 > 4 cm1108167
TIS pattern    
 Focal (< 50%)181001000.23
 Predominant (50–75%)428454
 Pure (> 75%)1238468
Associated component    
Oxyphilic features    
 ≤ 3/10 HPF15388710.01
 > 3/10 HPF307340
Cell size    
Vascular invasion    
Radioiodine therapy    
Figure 2.

Overall survival distribution based on Kaplan–Meier product-limit estimates of trabecular-insular-solid carcinomas according to (A) age, (B) the presence of necrosis, and (C) mitotic index. HPF: high-power fields.

Briefly, a statistically significant correlation with poor outcome was observed in patients age > 45 years who showed the presence of necrosis and had a mitotic index > 3 per 10 HPF. The presence of necrosis was associated with aggressive tumor behavior, irrespective of disease extent, and also represented a bad prognostic indicator in the oxyphilic carcinoma and papillary carcinoma subgroups when they were analyzed separately. Multivariate analysis confirmed these parameters as independent prognostic factors (Table 3).

Table 3. Multivariate Analysis of Clinicopathologic Parameters using with a Cox Proportional Hazards Model
VariableChi-square statisticP value
  1. HPF: high-power fields.

Presence of necrosis18.77< 0.0001
Mitotic count > 3/10 HPF6.450.01
Age > 45 yrs5.430.02

To define different prognostic categories in the group of TIS carcinomas on the basis of significant parameters, a simplified scoring system was designed that scored the presence of necrosis (the most influent parameter) 3 points and scored the remaining factors (age > 45 years and mitoses > 3 per 10 HPF) 1 point each. The three resulting subgroups (Group 1 score, 0–1; Group 2 score, 2–3; Group 3 score, 4–5) had remarkably different survivals, as defined by overall survival distribution based on Kaplan–Meier product-limit estimates (P < 0.0001) (Fig. 1B). The 5-year survival rates were 94%, 89%, and 74%, for Groups 1–3, respectively. At 10 years, these rates were 83%, 74%, and 48% for the same groups, respectively.


In the current study, we confirmed that PD TIS carcinomas of the thyroid have a clinical behavior intermediate between follicular/papillary carcinoma and anaplastic carcinoma. This is in agreement with data from the literature3, 12, 28, 29 and with our previously reported findings.8 However, three different prognostic categories were be identified specifically by using a scoring system applied to the clinicopathologic parameters of patient age, the presence of necrosis, and mitotic count. In more general terms, in agreement with the findings of LiVolsi et al.,25, 30 the current study data suggest that the grading of thyroid carcinoma is of high prognostic and clinical value. A scheme defining the proposed spectrum of follicular-derived thyroid carcinomas, comparing their differentiation and their clinical behavior, is shown in Figure 3. The impact of relevant clinicopathologic parameters is discussed briefly.

Figure 3.

Schematic representation of the spectrum of follicular-derived thyroid tumors. Follicular carcinoma (FTC) and papillary carcinoma (PTC) are represented on the left, and anaplastic carcinoma (AC) is represented on the other extreme of the spectrum of differentiation (right). Trabecular-insular-solid carcinomas (TISC) lie in an intermediate position; and prognostic parameters, which separate indolent from aggressive tumors, are identified according to the present data.

Gender, Age, Tumor Size, and Radioiodine Therapy

No statistically significant differences were observed when comparing patient gender and overall survival. Although male patients reportedly had the worst prognosis among patients with papillary carcinoma of the thyroid, gender did not appear to be relevant in patients with TIS carcinoma. Conversely, an older age at the time of diagnosis was found to be associated significantly with poor outcome in patients with TIS carcinomas, as also was reported in patients with papillary carcinoma.13 Using an age of 45 years as a cut-off age, a significant difference (P = 0.007) was observed with regard toin overall survival. Tumor size > 4 cm, which reflects pT3 disease according to the greatest tumor dimension, was not found to be associated significantly with more aggressive tumor behavior. In the vast majority of patients, a homogeneous therapeutic approach was adopted, including total thyroidectomy and radioiodine therapy. The impact of radioiodine treatment on the prognosis of patients with TIS carcinomas to our knowledge is not well established. In a previous study,8 we showed that radiometabolic therapy was effective in many patients, including those who had metastases at the time of diagnosis. Iodine uptake was much higher in patients who were treated shortly after surgery compared with the uptake in patients who were treated at the time of local recurrence or distant spread. This was not the experience of other authors, who reported poor iodine uptake in a variable percentage of patients.3, 28, 31 In a recent study, it was found that the only patient with insular carcinoma who benefited from radioiodine treatment carried an activating mutation of the thyroid-stimulating hormone receptor.28 The heterogeneous response to radioiodine in patients with TIS carcinomas may be related to dosage and timing of the treatment. In the current series, patients with TIS carcinomas were treated with the same dosage and timing as patients with well differentiated carcinomas who had disease of a similar stage and with similar clinical features.

Extent of TIS Pattern of Growth

It was proposed originally that a PD carcinoma was diagnosed when the majority of the tumor (e.g., > 75% of the area) was made of trabecular, or insular, or solid patterns.1, 3, 4, 8 Relating the extent of the TIS component with clinical behavior, we did not find any statistically significant difference in the overall survival of patients who had carcinomas with only a minor TIS component (10–50%) compared with patients who had carcinomas with a well represented component (50–75%) and patients who had virtually “pure” TIS carcinomas (> 75%), although a better prognosis was observed in patients who had tumors with a minor TIS component compared with the latter 2 patient groups considered together (P = 0.08). The fact that even a minor insular component is associated with a worse prognosis compared with the prognosis for patients in the control group who had well differentiated follicular carcinoma was reported previously by Nishida et al..6 Other authors did not confirm the impact of a minor insular component on the prognosis of patients with follicular carcinoma.10 From a practical point of view, we suggest that the presence of a TIS component must appear in the final pathologic report of a thyroid carcinoma. If the tumor also demonstrates aggressive features (large tumor size, necrosis, and mitoses), then the diagnosis of PD carcinoma with TIS patterns is appropriate.

Extent of Vascular Invasion

The vast majority of patients in the current series (178 of 183 patients) had neoplastic thrombi in at least some peripherally located blood vessels. The remaining five patients met the criteria of malignancy, with disease that had extensive capsular penetration. However, no significant differences in behavior were observed between minimally and widely invasive TIS carcinomas.

Associated Well Differentiated Components

Since the original report of Carcangiu et al.,3 it has been made clear that solid and insular carcinomas have small follicles included in the tumoral insulae. However, the general architecture was not that of a well differentiated follicular carcinoma. It also was clear that some tumors had insular (or, in general, PD) features intermingled with otherwise classic follicular carcinoma and—in rare tumors—with papillary carcinoma.32, 33 Although the results of the current study have shown that the relative proportion of well differentiated and PD (TIS) components does not affect prognosis, the recognition of such follicular or papillary features may be of interest from a histogenetic point of view. A solid tumor with clear nuclei, grooves, and nuclear holes may be interpreted as a solid variant of papillary carcinoma20 if necrosis and a high mitotic index are absent. In the spectrum of follicular-derived thyroid carcinomas, it is possible that there is a degree of overlap in the interpretation of these solid tumors. Indeed, in the current series, 9 of 24 patients with papillary carcinoma-associated features had tumors that lacked necrosis and mitoses. However, in all patients, prominent vascular invasion was present, and clinically aggressive tumor behavior was recorded in seven of these patients. This is at variance with the “slightly worse prognosis” (compared with classic papillary carcinoma) reported by Nikiforov et al.20 in their series of solid papillary carcinomas. In the current series, overall, solid carcinomas with papillary carcinoma-associated features were found to have an more aggressive tumor behavior compared with the control group of differentiated carcinomas. It was shown that a link between TIS carcinoma and papillary carcinoma exists also from a genetic point of view, because the RET/PTC1 rearrangement (typical of papillary carcinoma) also was found in a fraction of TIS carcinomas that had the nuclear features of papillary carcinoma and/or some (residual) foci of papillary carcinoma.34 This indicates that TIS PD carcinoma may dedifferentiate from well differentiated, conventional, follicular or even papillary carcinoma. It was not shown that the type of associated component affected the prognosis of patients with TIS carcinomas.

Oxyphilic Features and Cell Size

After our past experience in patients with solid-trabecular oxyphilic carcinomas21 and the experience of others in Hürthle cell carcinomas in general,22, 23 in this study, we confirmed that the histopathologic features and the clinical behaviors of the 66 oxyphilic carcinomas with TIS growth patterns were similar to those of conventional PD TIS carcinomas. Although oxyphilic (Hürthle) cell carcinomas generally have been considered much more aggressive than nonoxyphilic carcinomas,12, 13, 35 this was not confirmed in the current series of oxyphilic carcinomas with predominant solid-trabecular architecture. In the group of solid-trabecular oxyphilic carcinomas, it was suggested that tumors made of small, deeply oxyphilic cells had a more aggressive course21 compared with tumors that contained the classic large, granular, eosinophilic cells. However, no statistically significant difference was observed in the current series of 66 TIS oxyphilic carcinomas (36 of 66 tumors were made of small cells).

Necrosis and Mitotic Count

Finally, the presence of necrosis and a mitotic count > 3 per 10 HPF appeared to be the most relevant parameters for predicting prognosis in patients with TIS carcinomas. With regard to necrosis, its presence was a sufficient criterion to predict an aggressive behavior (P < 0.0001), irrespective of its extent (focal vs. extensive). Necrotic foci were observed either in the middle of solid or insular areas or as small clusters containing few or single necrotic cells. The presence of necrosis was found to have a statistically significant impact on prognosis, even considering separately the groups of tumors with predominant oxyphilic features or with a papillary carcinoma-associated component. The mitotic index was much higher in TIS carcinomas, irrespective of their oxyphilic or conventional features. This is in agreement with the literature data regarding Ki-67-detected proliferative activity in PD carcinomas.4, 17, 21 Choosing 3 mitoses per 10 HPF as a cut-off value allowed the separation of a group of patients who had low proliferating carcinomas and, thus, a relatively better prognosis, from a group of patients who had aggressive carcinomas (P = 0.01).

Scoring System

Overall, these findings indicate that, by combining the statistically significant parameters into a scoring system, TIS carcinomas can be grouped into three prognostic categories. Group 1 (score, 0–1) showed a survival overlapping that of well differentiated papillary and follicular carcinomas. Group 3 (score, 4–5) showed a clinical course closer to that of anaplastic carcinoma, whereas Group 2 (score, 2–3) had a survival curve intermediate between the former 2 groups. From a practical point of view, because, together, all TIS carcinomas represent a clinically heterogeneous group of tumors, we suggest that this scoring system may be useful for identifying tumors that carry a higher risk of an unfavorable clinical outcome.

The results of the current study demonstrated that: 1) the group of 187 TIS carcinomas studied in the current report followed a more aggressive course than the control groups of well differentiated follicular and papillary carcinomas, irrespective of the extent of TIS areas; 2) solid/trabecular foxyphilic carcinomas have behavior similar to that of conventional (nonoxyphilic) TIS carcinomas; 3) age > 45 years, the presence of necrosis, and a mitotic count > 3/10 HPF are associated significantly with clinically aggressive tumors; and 4) a scoring system using a combination of such clinicopathologic parameters may help to define TIS carcinomas that carry a higher risk of aggressive behavior.


The authors are deeply indebted to Prof. G. Bussolati (University of Turin) for his suggestions and for critically reviewing the article.