Classification of Parathyroid Cancer

ORIGINAL ARTICLE – ENDOCRINE TUMORS

Classification of Parathyroid Cancer

Klaus-Martin Schulte, MD, FRCS1, Anthony J. Gill, FRCPA2, Marcin Barczynski, MD3, Elias Karakas, MD4,

Akira Miyauchi, MD5, W. T. Knoefel, MD6, Celestino Pio Lombardi, MD7, Nadia Talat, BSc Hons1,

Salvador Diaz-Cano, FRCPath8, and Clive S. Grant, MD9

1Department of Endocrine Surgery, King’s College Hospital, King’s Health Partners, London, UK; 2Department of

Anatomical Pathology, Royal North Shore Hospital, St. Leonards, NSW, Australia; 3Department of Endocrine Surgery,

Jagiellonian University College of Medicine, Krakow, Poland; 4Department of Visceral, Thoracic, and Vascular Surgery,

University Hospital Marburg, Marburg, Germany; 5Department of Surgery, Kuma Hospital, Kuma, Japan; 6Department of

General and Visceral Surgery, Heinrich-Heine University, Dusseldorf, Germany; 7Division of Endocrine Surgery,

Department of Surgery, Universita Cattolica del Sacro Cuore, Rome, Italy; 8Department of Pathology, King’s College

Hospital, King’s Health Partners, London, UK; 9Department of Surgery, Mayo Clinic, Rochester, MN

ABSTRACT

Purpose. Parathyroid cancer is rare and often has a poor

outcome. There is no classification system that permits pre-

diction of outcome in patients with parathyroid cancer. This

study was designed to validate two prognostic classification

systems developed by Talat and Schulte in 2010 (‘‘Clinical

Presentation, Staging and Long-term Evolution of Parathyroid

Cancer,’’ Ann Surg Oncol 2010;17:2156–74) derived from a

retrospective literature review of 330 patients.

Methods. This study contains 82 formerly unreported

patients with parathyroid cancer. Death due to disease was

the primary end point, and recurrence and disease-free sur-

vival were the secondary end points. Data acquisition used a

questionnaire of predefined criteria. Low risk was defined by

capsular and soft tissue invasion alone; high risk was defined

by vascular or organ invasion, and/or lymph node or distant

metastasis. A differentiated classification system further

classified high-risk cancer into vascular invasion alone (class

II), lymph node metastasis or organ invasion (class III), and

distant metastasis (class IV). Statistical analyses included

risk analysis, Kaplan-Meier analysis, and receiver–operating

characteristic (ROC) analysis.

Results. Follow-up ranged 2–347 months (mean 76 months).

Mortality was exclusive to the high- risk group, which also

predicted a significant risk of recurrence (risk ratio 9.6;

95% confidence interval 2.4–38.4; P \ 0.0001), with sig-

nificantly lower 5-year disease-free survival (v2 = 8.7;

P \ 0.005 for n = 45). The differentiated classification also

provided a good prognostic model with an area under the

ROC curve of 0.83 in ROC analysis, with significant

impairment of survival between classes (98.6%, 79.2%,

71.4%, 40.0%, P \ 0.05 between each class).

Conclusions. This study confirms the validity of both

classification systems for disease outcome in patients with

parathyroid cancer.

Parathyroid cancer is rare, but it is an important cause of

death in the group of patients with primary hyperparathy-

roidism.1–4 It accounts for 0.7–2.1% of all cases of primary

hyperparathyroidism, with an incidence of 0.4 to 0.6 per

million based on Surveillance, Epidemiology, and End

Results data.5,6 Because there is no agreed-upon staging

system for parathyroid cancer, it is currently not possible to

provide patients with a solid prognosis, and trial designs

have had a lack of stratification.

A formerly proposed staging system was insufficient to

achieve meaningful outcome prediction in a large retro-

spective cohort.1,6 Small cohort studies have highlighted

the relevance of histopathologic criteria, and our recent

detailed literature analysis of 330 patients also showed

their significant predictive power for survival and recur-

rence (Tables 1, 2).1,7,8 The current retrospective study was

Electronic supplementary material The online version of thisarticle (doi:10.1245/s10434-012-2306-6) contains supplementarymaterial, which is available to authorized users.

� Society of Surgical Oncology 2012

First Received: 16 December 2011

K.-M. Schulte, MD, FRCS

e-mail: [email protected]

Ann Surg Oncol

DOI 10.1245/s10434-012-2306-6

http://dx.doi.org/10.1245/s10434-012-2306-6

designed to validate two predefined classification systems

in an independent validation cohort.

PATIENTS AND METHODS

Endocrine surgeons in centers with expertise in parathy-

roid cancer were approached to provide data of their

complete cohort of patients with parathyroid cancer. Patients

included in the cohort underlying the initial analysis and

derivation of the novel classification schemes, referred to as

the reference cohort, were excluded to avoid duplication.1

This validation study is based on 82 patients with complete

information received in questionnaires from eight expert

institutions. (See Supplemental Table 1 for questionnaire.)

This cohort is referred to as the validation cohort.

Inclusion criteria included the following: unambiguous

histopathologic diagnosis of parathyroid cancer by World

Health Organization 2004 criteria, defined by infiltrative

growth combing capsular invasion combined with soft

tissue invasion as minimal criteria or—in their absence—

histologic proof of vascular invasion (major criteria) with

or without invasion of vital organs (trachea, oesophagus,

major vessel); presence of locoregional or distant metas-

tasis; a complete data set; known margin status; and

follow-up of at least 2 months.9 The pathologic features

were extracted from the original pathology reports of the

eight participating institutions.

The 82 patients we describe here were identified by

strict application of the inclusion and exclusion criteria.

Twenty-seven additional patients were initially considered

for entry onto the study but were excluded from analysis

because of failure to meet all the inclusion criteria.

As described in detail elsewhere, the reference cohort

initially underwent multinominal regression analysis taking

into account gender, capsular invasion, soft tissue invasion,

vascular invasion, invasion of vital organs, lymph node

metastases, and margin status.1 From this analysis, we

derived two classification schemes, which are summarized

in Table 1. The term differentiated classification relates to

a scheme created in alignment with analogous schemes

provided in the framework of tumor, node, metastasis

system (TNM) classification by the Union for International

Cancer Control.10 The second classification scheme is

referred to as high-risk/low-risk classification. It is simpler

and falls into only two categories (Table 1).

The 82 patients reported here comprise the validation

cohort. They have been stratified according to both above

classification schemes, and outcome differences between

classes are reported throughout.

Surgical intervention was grouped into two categories:

local excision only, and en bloc excision. Local excision

comprised only pericapsular excision of the parathyroid

lesion. En bloc excision included en bloc and oncologic

resection. En bloc excision describes excision of the

parathyroid with circumferential soft tissue as the minimal

criterion; oncologic resection additionally includes further

surgery such as ipsilateral thyroid lobectomy, centrocer-

vical lymphadenectomy, or further locoregional excision.

Outcome was defined before requesting data from study

centers in accordance with results of the underlying study.1

Adverse outcome events were defined as death due to

disease or recurrence. These adverse events were used to

define primary and secondary end points.

The primary end point was to assess death due to dis-

ease. This was done by calculation of risk ratio, odds ratio,

and overall survival by Kaplan-Meier analysis. Secondary

end points were recurrence (defined as any event where

locally or distantly recurring disease was identified during

follow-up) and disease-free survival (DFS). This was

assessed by calculation of risk ratio, odds ratio, and overall

survival by Kaplan-Meier analysis.

We defined death and recurrence as the absence or

presence of reaching the primary end point of death due to

TABLE 1 Classification scheme

Classification Histology criteria

Differentiated

T

(Tx) No information available

T1 Evidence of capsular invasion

T2 Invasion of surrounding soft tissues excluding the vital

organs of trachea, larynx, and oesophagus

T3 Evidence of vascular invasion

T4 Invasion of vital organs—hypopharynx, trachea,

oesophagus, larynx, recurrent laryngeal nerve,

carotid artery

N

(Nx) Lymph node not assessed

N0 No regional lymph node metastases

N1 Regional lymph node metastases

M

(Mx) Distant metastases not assessed

M0 No evidence of distant metastases

M1 Evidence of distant metastases

Differentiated

I T1 or T2 N0M0

II T3 N0 M0

III Any T, N1 M0, or T4

IV Any N, M1

High and low risk

Low Capsular invasion combined with invasion of

surrounding soft tissue

High Vascular invasion and/or lymph node metastases and/

or invasion of vital organs and/or distant metastases

K.-M. Schulte et al.

disease or the secondary end point of recurrence, in

alignment with the definition of DFS. This was assessed by

calculation of risk ratio, odds ratio, and overall survival by

Kaplan-Meier analysis.

Five-year-disease-free survival was calculated for the

subgroup of patients with at least 60 months’ follow-up

(n = 45).

The tertiary end point was model quality of the high-risk

versus low-risk classification system. This was assessed by

receiver–operating characteristic (ROC) and basic proba-

bility statistics.

Data were analyzed by SPSS 16.0 software (SPSS,

Chicago, IL). All P values are two-sided. All analyses were

carried out on the entire cohort of 82 patients, with

exception of calculation of 5-year DFS, for which data

were available for only 45 patients.

RESULTS

Contributing authors provided data on a total of 109

patients. A total of 27 patients not fulfilling the study’s

inclusion criteria were excluded: 17 data sets did not contain

sufficient or any histology information, 5 data sets described

capsular invasion as the sole criterion of malignancy and

were therefore excluded, and 5 data sets related to cancer but

did not report follow-up data. All remaining 82 patients in the

validation cohort satisfied strict histologic criteria (Table 1)

for the diagnosis of parathyroid cancer (defined as infiltrative

growth, invasion of blood vessels, perineural spaces, soft

tissues, thyroid gland, or other adjacent structures or tumors

with documented metastases).11,12

Patients (n = 82) originated from eight centers: Mayo

Clinic, Rochester, Minnesota (n = 28); Royal North Shore

TABLE 2 Comparison of cohorts used for developing and validating the classification system

Factor Reference cohorta Validation cohort Total P(n = 140) (n = 82) (n = 222)

Gender (M/F) 62/78 (44.3%) 40/42 (48.9%) 102/120 (45.9%) NS

Age, y NS

Mean ± SD 52.1 ± 14.9 52.4 ± 14.5 52 ± 15.0

Median 54.0 51.0 53.0

Range 14.0–78.0 25.0–84.0 14.0–84.0

Size, mm NS

Mean ± SD 36.4 ± 24.4 35.0 ± 17.7 36.0 ± 22.0

Median 30.0 30.0 30.0

Range 12.0–140.0 15.0–120.0 12.0–140.0

PTH, times upper limit of normal \0.05

Mean ± SD 9.4 ± 10.9 14.2 ± 13.8 11.6 ± 12.5

Median 5.2 10.7 8.0

Range 1.0–76.1 1.2–86.0 1.0–86.0

Ca, mM NS

Mean ± SD 3.5 ± 0.7 3.4 ± 0.5 3.5 ± 0.6

Median 3.4 3.4 3.4

Range 2.3–6.2 2.3–4.7 2.3–6.2

Local excision 32 (22.9%) 18 (22.0%) 50 (22.5%) NS

Oncologic resection 108 (48.6%) 64 (78.0%) 172 (77.4%) \0.01

Follow-up, mo \0.01

Mean ± SD 55.0 ± 58.3 76.3 ± 74.0 63.0 ± 65.0

Median 34.0 50.0 36.0

Range 3.0–302.0 2.0–347.0 2.0–347.0

Local recurrence (yes/no) 20/120 (14.3%) 18/64 (22.0%) 28/184 (12.6%) NS

Distant recurrence (yes/no) 34/106 (24.3%) 11/71 (13.4%) 45/177 (20.3%) \0.05

All recurrence (yes/no) 50/90 (35.7%) 27/57 (32.9%) 75/147 (33.8%) NS

Death (yes/no) 27/113 (19.3%) 11/71 (13.4%) 38/184 (17.1%) NS

PTH parathyroid hormone, Ca calciuma The reference cohort of 140 patients was obtained from the study by Talat and Schulte.1 We analyzed clinical presentation in 330 patients with

parathyroid cancer; we have only included 140 patients because they have complete or consistent data sets with regard to the study factors


Hospital, Sidney, Australia (n = 16); King’s College

Hospital, London, UK (n = 11); Jagiellonian University,

Medical College, Krakow, Poland (n = 9); University of

Marburg, Germany (n = 8); Kuma Hospital, Kobe, Japan

(n = 4), University of Dusseldorf, Germany (n = 3); and

Universita Cattolica Del Sacro Cuore, Rome, Italy (n = 3).

Demographic features, macroscopic findings, and bio-

chemical parameters of the validation group did not in

principle differ between the reference and validation

cohorts (Table 2). Histopathologic features differed sig-

nificantly (Table 3). Both cohorts had undergone similar

treatment (Table 2), with significantly more oncologic

resections relating to a higher number of patients with vital

tissue invasion. Accordingly, local and total recurrence was

similar, while distant metastasis was more common in the

reference group, in accordance with significantly more

common vascular invasion. Margin status was available in

the validation cohort and in none of the reference cohort

(Table 3).

Disease-free resection margins were achieved in 87.3%.

Margin status did not differ by surgical approach. En bloc

or oncologic resections achieved 81.2% disease-free mar-

gins, compared to 88.9% in local excision (P = NS).

Follow-up ranged 2–347 months with a median of

50.0 months (mean 76.3 ± 74.0 months). A total of 11

patients died during follow-up after a median follow-up of

37 months (range 9–90 months; mean 46 ± 27 months),

and 18 patients experienced locoregional recurrence after a

median follow-up of 10 months (range 2–192 months; mean

32 ± 47 months). Of the 18 patients experiencing locore-

gional recurrence, 9 experienced recurrence before

12 months (at 2, 3, 6, 6, 7, 8, 8, 8, and 8 months). Distant

metastases were found in 11 patients after a median

follow-up of 48 months (range 2–300 months; mean 74 ±

75 months). Thirty patients died or experienced disease

recurrence after a median follow-up of 36 months (range

5–312 months; mean 75 ± 89 months).

A cross table of outcomes against classification is pro-

vided in Table 4. The progressive risk to experiencing

adverse events associated with class progression is shown

in Supplemental Digital Content 2.

Prediction of Primary and Secondary End Points Using

the High-risk Versus Low-risk Classification

The relative risk to die due to disease is infinitely higher

in high-risk patients, in fact exclusive to them. The relative

risk to experience recurrence is 12.8 times higher in high-

risk patients (P \ 0.0001). This is confirmed by Kaplan-

Meier analysis (Fig. 1a, b). Overall DFS (v2 = 15.4; 95%

confidence interval (2.7–29.1); P \ 0.0001) (Fig. 1c) and

5-year DFS are significantly higher in patients classified

as high-risk (93.3% vs. 46.7%; v2 = 8.7, P \ 0.003)

(Fig. 1d).

Prediction of Primary and Secondary End Points Using

the Differentiated Classification

The relative risk to die from disease is infinitely higher

in class II disease compared to class I and class III com-

pared to class I (P \ 0.0001). There is no difference

between class II disease versus class III (P = NS). Kaplan-

Meier analysis confirms the class-related progression of

mortality (Fig. 2a), and recurrence (Fig. 2b), with lower

DFS in class I compared to II and III.

Assessment of Model Quality

The factors derived from the reference cohort and used

to build the classification schemes were again entered into

multinominal regression analysis for the primary and sec-

ondary outcomes. This showed good model quality.

Omnibus tests of model coefficients showed good perfor-

mance for the low-risk versus high-risk system and the

differentiated staging system, with v2 of 13.0, P \ 0.0001,

and v2 of 16.3, P \ 0.001, respectively. The Hosmer-

Lemeshow goodness of fit test showed a v2 value of 0 with

a significance level of 1, supporting the value of the model.

TABLE 3 Comparison of histology features in cohorts used for developing and validating the classification system

Histology criteria Reference cohorta Validation cohort Total P(n = 140) (n = 82) (n = 222)

Capsular invasion (yes/no) 108/32 (77.1%) 62/20 (75.6%) 170/52 (76.6%) NS

Vascular invasion (yes/no) 94/46 (67.1%) 33/49 (40.2%) 127/95 (57.2%) \0.0001

Soft tissue invasion (yes/no) 89/78 (63.6%) 63/19 (76.8%) 152/97 (68.4%) 0.0003

Vital organ infiltration (yes/no) 9/131 (6.4%) 18/64 (22.0%) 27/195 (12.1%) 0.006

Lymph node metastasis (yes/no) 21/119 (15.0%) 8/74 (9.8%) 29/193 (13.1%) NS

Margin status (positive/negative) Not available 12/70 (17.1%) 12/70 (17.1%) –

a The reference cohort of 140 patients was obtained from the study by Talat and Schulte.1 We analyzed clinical presentation in 330 patients with

parathyroid cancer; we have only included 140 patients because they have complete or consistent data sets with regard to the study factors


The area under the ROC curve (AUC) was determined

for both models, showing good performance in ROC

analysis. (See Supplemental Figs. 3 and 4.) The AUC of

the high-risk versus low-risk model was 0.74 for the entire

cohort, 0.82 for the group with local excision only, and

0.74 for the en bloc group. The AUC of the differentiated

classification was 0.83 for the entire cohort, 0.91 for the

group with local excision only, and 0.79 for the en bloc or

oncologic surgery group. This makes it unlikely that sur-

gical choice confounded the model construction.

Predictive performance in terms of probability statistics

is optimal for the primary outcome and very good for the

secondary outcomes (Table 5).

Exclusion of Margin Status as a Confounder in Model

Construction

It is well established in cancer that disease-positive

resection margins are associated with increased local

recurrence rates.13,14 Because margin status was unknown

in most patients included in the reference cohort, it could

have been a major confounder in the initial model con-

struction.1 Margin status was known for all patients of the

validation cohort.

Because margin status was unknown, we therefore repe-

ated the outcome analysis in the subcohort of the validation

cohort with disease-negative resection margins (n = 70).

This confirms the unchanged validity of the high-risk versus

low-risk classification for outcome prediction (Supplemen-

tal Fig. 5). Likewise, class-related outcomes using the

differentiated classification scheme did not change (Sup-

plemental Fig. 6). In the same subcohort, ROC analysis

continued to show good performance (AUC 0.74) for the

high-risk versus low-risk model, and for the differentiated

classification (AUC 0.77). The slight drop compared to the

whole group was almost certainly related to the total number

of patients and events available for analysis.

Performance of the Differentiated Classification

Scheme in a Fused Cohort

Fusion of the reference cohort with the validation cohort

was undertaken to confirm consistency of results in a larger

cohort of 222 patients. Kaplan-Meier analysis found signif-

icant differences across classes. It demonstrated significant

progressive deterioration of prognosis with increasing dis-

ease class (Fig. 3).

Vascular Invasion Is Uniquely Associated with Distant

Metastasis

Class II and III disease predicted significant differences

in survival but not in recurrence. The hypothesis tested wasTA

BL

E4

Cro

ss-t

able

anal

ysi

so

fo

utc

om

es:

dea

tho

rre

curr

ence

alo

ne

and

ov

eral

lD

FS

ind

iffe

ren

tcl

asse

so

fp

arat

hy

roid

can

cer

in8

2p

atie

nts

Ch

arac

teri

stic

Dea

thR

ecu

rren

ceD

FS

Yes

,n

(%)

No

,n

(%)

Ris

kra

tio

,P

Yes

,n

(%)

No

,n

(%)

Ris

kra

tio

,

P(9

5%

con

fid

ence

inte

rval

)

No

,n

(%)

Yes

,n

(%)

Ris

kra

tio

,P

(95

%co

nfi

den

cein

terv

al)

Cla

ss

I0

34

(10

0.0

%)

14

.0,\

0.0

01

2(5

.1%

)3

2(9

4.9

%)

17

.5,\

0.0

01

4(1

0.3

%)

30

(89

.7%

)1

7.4

,\0

.00

01

II3

(12

.5%

)2

1(8

7.5

%)

13

(54

.2%

)1

1(4

5.8

%)

13

(54

.2%

)1

1(4

5.8

%)

III

7(3

1.8

%)

15

(68

.2%

)9

(40

.9%

)1

3(5

9.1

%)

11

(50

.0%

)1

1(5

0.0

%)

IV1

(50

.0%

)1

(50

.0%

)1

(50

.0%

)1

(50

.0%

)2

(10

0.0

%)

0

Ris

k

Lo

w0

34

(10

0.0

%)

Infi

nit

y2

(5.1

%)

32

(94

.9%

)8

.1,\

0.0

00

1(2

.1–

32

.3)

4(1

0.3

%)

30

(89

.7%

)8

.9,\

0.0

00

1(2

.7–

29

.1)

Hig

h1

1(2

2.9

%)

37

(77

.1%

)2

3(4

7.9

%)

25

(52

.1%

)2

6(5

4.2

%)

22

(45

.8%

)

DF

Sd

isea

se-f

ree

surv

ival

DF

Sw

asd

efin

edas

tim

eto

dea

than

d/o

rre

curr

ence


that vascular invasion the defining feature of class II dis-

ease conferred a risk to distant metastasis, as opposed to

local and distant recurrence associated with class III dis-

ease. Figure 4 confirmed this hypothesis. Vascular invasion

was present in all patients with distant metastases (risk

ratio = infinity; P \ 0.0001).

DISCUSSION

The analysis of this independent data set of hitherto

unreported patients confirms the validity of both formerly

proposed staging systems for parathyroid cancer. Sharp

outcome differences between classes are not due to diag-

nostic confusion caused by atypical parathyroid adenoma

being overdiagnosed as parathyroid carcinoma because we

used World Health Organization 2004 criteria for definition

of parathyroid cancer.9,11,15–17 For inclusion as cancer,

lesions needed to be infiltrative and to show a combination

of at least capsular and soft tissue invasion in the absence

of other determining features. Margin status as principal

confounder of model construction is unlikely because

model performance remained good in a subcohort defined

by disease-free resection margins. On the basis of this

reference cohort, we proposed a system discriminating

between high-risk and low-risk cancer (Table 1), and a

differentiated classification system aligned with the per-

spectives of TNM classification (Table 2).1

Performance for the Primary End Point Death Due

to Disease

The primary end point was well predicted by both

models. The high-risk versus low-risk system differentiated

two roughly equally sized groups with sharply distinct

outcomes. There were no deaths in the low-risk group. The

differentiated classification system delivered identical

a b

c d

FIG. 1 High-risk and low-risk

classification predict primary

and secondary outcomes.

a Death due to disease.

b Recurrence-free survival.

c Overall DFS. d Five-year

DFS. a–c Includes a total of 82

patients, of whom 34 were in

the low-risk group (41.5%) and

48 (58.3%) were in the high-risk

group. d Five-year DFS for 45

patients with complete follow-

up data at 5 years


results with regard to class I disease, which is identical to

the low-risk class. Although there is a difference between

survival in class II and III disease, this did not reach sig-

nificance in the validation cohort (Fig. 2a). This was likely

due to cohort size: outcome analysis for the fused cohort

containing patients of the reference cohort clearly dem-

onstrated significant outcome variation (Fig. 3). The

validation study had not credibly approached any issues

regarding class IV disease because the number of patients

was insufficient to address this. Corresponding to our

experience in this and other cancers, it remains highly

plausible that cancer presenting with metastasis at the

outset has a poor outcome.

Performance for the Secondary End Points Recurrence

and DFS

The secondary end point of recurrence was well predicted

by both models. Again, the high-risk versus low-risk system

differentiated groups with sharply distinct recurrence rates,

overall survival, and 5-year DFS (Fig. 2) The differentiated

classification system revealed that there was no difference

between classes II and III with regard to recurrence. This

highlights that vascular invasion is an important independent

predictor of recurrence and therefore defines a class of its

own.1 This concurs with observations in other endocrine

pathologies, where vascular invasion is an independent

adverse feature associated with hematogenous spread. Vas-

cular invasion is an established risk marker in adrenocortical

cancer, pheochromocytoma, and medullary thyroid can-

cer.18–21 It is debated whether it confers an adverse outcome

in papillary thyroid cancer.22–25 The clinical importance of

vascular invasion, the defining feature of class II disease,

resides in its formerly unknown unique association with

distant recurrence in this cancer (Fig. 4). This rivals the risk

of locoregional recurrence caused by lymphatic spread or

extensive local invasion. In fact, vascular invasion and

locoregional extended disease contribute about equally to

the overall appearance of recurrence defined as the sum of

locoregional and distant recurrence.

a b c

FIG. 2 Differentiated classifications predict primary and secondary

outcomes. a Death due to disease. b Recurrence-free survival.

c Overall DFS. Data are from 80 patients, of whom 34 (42.5%) had

class I disease, 24 (30.0%) class II, and 22 (27.5%) class III. Class IV

was excluded from analysis because only 2 patients had class IV

disease

TABLE 5 Statistical performance indicators of the classification system

Outcome High risk vs. low risk Differentiated classification system

I vs. II I vs. III II vs. III

Sensitivity (%) NPV (%) Sensitivity (%) NPV (%) Sensitivity (%) NPV (%) Sensitivity (%) NPV (%)

Death 100.0 100.0 100.0 100.0 100.0 100.0 70.0 87.5

Recurrence 92.0 94.1 86.7 94.1 81.8 94.1 40.9 44.8

Death and recurrence 86.7 88.2 76.5 88.2 73.3 88.2 45.8 45.8

True negative (TN) indicates low risk without adverse event; truly positive (TP), high risk with adverse event; false negative (FN), low risk with

adverse event; false positive (FP), high risk without adverse event. Sensitivity = TP/(TP ? FN); negative predictive value (NPV) = TN/

(FN ? TN)


Tertiary Outcome: Model Quality

The quality of both models is good as demonstrated by

ROC rank-order statistics.26 The models have high sensi-

tivity and negative predictive power for adverse outcomes

defining by primary and secondary outcomes.

Margin Impact Has Not Likely Acted as a Major

Confounder in Model Construction

Positive status of resection margins is known to increase

local recurrence and adversely affect overall outcome in

many cancers. It is not usually included in any staging or

classification systems because it is not a feature of the

cancer itself, but rather a result of the surgical

approach.10,27 More aggressive cancers might be associated

frequently with positive margins. In parathyroid cancer,

there are few data supporting this general notion.28,29 A

systematic retrospective analysis showed significantly

increased numbers of repeat surgeries when local excision

only was performed.1 Margin status does not affect model

quality in a reiterated analysis in a cohort with known

negative margin status composed of those patients of the

former and validation cohort, where this information was

available (n = 70). Although margin status has demon-

strable impact on outcome, it has not been a significant

confounder of the current validation analysis. This may be

because the choice of surgery was inconsistent throughout

the cohorts. More radical surgery greatly reduces the need

for repeat surgery and improves survival outcome.1,30 In

the absence of a stage-related surgical strategy, and as a

result of the considerable inhomogeneity of the chosen

approach, confounding effects of margin status and surgi-

cal choice will have rendered both effects unrecognizable.

The distinction between cancer recurrence and para-

chromatosis can be notoriously difficult; further, it can lead

to the wrong diagnosis of parathyroid cancer made after the

fact, when faced with an unexpected or multifocal source

of recurrent hyperparathyroidism. This problem, among

others, led us to reanalyze outcomes in the subset of 70

patients with known negative resection margins (Supple-

mental Figs. 5 and 6). Lesions that underwent R0 resection

could not have caused parachromatosis. This analysis

produced identical results.

It is possible that the prognostic significance of vascular

invasion depends on its extent—that is, minimal versus

extensive. The current study does not address this issue or

the potential implications of grading of other histopathol-

ogy features. This limitation relates to two main arguments.

The current study was planned and performed as a revali-

dation of a previously established classification system.1

The addition of further subclassifying features would dis-

tort this approach. Also, the group of patients with vascular

invasion contains only 33 of 82 patients. The size of the

present cohort, which to our knowledge is the largest

cohort ever reported relating to individualized patient data

rather than registry or national database entries, is insuffi-

cient to support reliable statistical analysis of prognostic

implications after further splitting the group according to

vascular invasion.

FIG. 3 Fusion of the reference cohort with the validation cohort

predict primary and secondary outcomes in differentiated classifica-

tion system. Analysis included the reference cohort of 140 patients

from the study by Talat and Schulte and 82 from the validation cohort

(n = 222).1 Class I, 69 (31.1%); class II, 106 (47.7%); class III, 42

(18.9%); class IV, 5 (2.3%). Overall survival is 98.6% for class I,

79.2% for class II, 71.4% for class III, and 40.0% for class IV

FIG. 4 Vascular invasion confers a risk of distant metastasis.

Analysis included 82 patients, of whom 49 (59.8%) had no vascular

invasion and 33 (40.2%) had vascular invasion


The current study provides substantial evidence that

histopathologic criteria are key outcome predictors in

parathyroid cancer. We are currently collecting additional

case material and plan to subject all specimens to central

histopathologic review against an enhanced list of graded

criteria. This will overcome limitations of the current

study, including those possible deriving from the extraction

from former pathology reports, but data may not be

available in the very near future.

In summary, this study confirms the validity of two

novel classification systems to achieve reliable outcome

prediction for patients with parathyroid cancer.

REFERENCES

1. Talat N, Schulte K. Clinical presentation, staging and long-term

evolution of parathyroid cancer. Ann Surg Oncol. 2010;17:

2156–74.

2. Rodgers SE, Perrier ND. Parathyroid carcinoma. Curr OpinOncol. 2006;18:16–22.

3. Koea JB, Shaw JH. Parathyroid cancer: biology and management.

Surg Oncol. 1999;8:155–65.

4. Delellis RA. Challenging lesions in the differential diagnosis of

endocrine tumors: parathyroid carcinoma. Endocr Pathol. 2008;

19:221–5.

5. Lee PK, Jarosek SL, Virnig BA, Evasovich M, Tuttle TM. Trends

in the incidence and treatment of parathyroid cancer in the United

States. Cancer. 2007;109:1736–41.

6. Shaha AR, Shah JP. Parathyroid carcinoma: a diagnostic and

therapeutic challenge. Cancer. 1999;86:378–80.

7. Schantz A, Castleman B. Parathyroid carcinoma. A study of 70

cases. Cancer. 1973;31:600–5.

8. Chang YJ, Mittal V, Remine S, et al. Correlation between clinical

and histological findings in parathyroid tumors suspicious for

carcinoma. Am Surg. 2006;72:419–26.

9. Bondeson L, Grimelius L, Delellis RA. World Health Organisa-

tion classification of tumours, pathology and genetics: tumour of

endocrine organs. Lyon, France: IARC Press, 2004.

10. Sobin LH, Gospodarowicz MK, Wittekind C; International Union

Against Cancer. TNM classification of malignant tumours. 7th

ed. Oxford: Wiley-Blackwell, 2009.

11. DeLellis RA. Parathyroid carcinoma: an overview. Adv AnatPathol. 2005;12:53–61.

12. Carlson D. Parathyroid pathology: hyperparathyroidism and

parathyroid tumors. Arch Pathol Lab Med. 2010;134:1639–44.

13. Agrawal S, Belghiti J. Oncologic resection for malignant tumors

of the liver. Ann Surg. 2011;253:656–65.

14. Renton SC, Gazet JC, Ford HT, Corbishley C, Sutcliffe R. The

importance of the resection margin in conservative surgery for

breast cancer. Eur J Surg Oncol. 1996;22:17–22.

15. Juhlin CC, Nilsson IL, Johansson K, et al. Parafibromin and APC

as screening markers for malignant potential in atypical para-

thyroid adenomas. Endocr Pathol. 2010;21:166–77.

16. Elgoweini M, Chetty R. Hyalinizing parathyroid adenoma and

hyperplasia: report of 3 cases of an unusual histologic variant.

Ann Diagn Pathol. 2011;15:329–32.

17. Howell VM, Gill A, Clarkson A, et al. Accuracy of combined

protein gene product 9.5 and parafibromin markers for immu-

nohistochemical diagnosis of parathyroid carcinoma. J ClinEndocrinol Metab. 2009;94:434–41.

18. Volante M, Bollito E, Sperone P, et al. Clinicopathological study

of a series of 92 adrenocortical carcinomas: from a proposal

of simplified diagnostic algorithm to prognostic stratification.

Histopathology. 2009;55:535–43.

19. Kimura N, Watanabe T, Noshiro T, Shizawa S, Miura Y. Histo-

logical grading of adrenal and extra-adrenal pheochromocytomas

and relationship to prognosis: a clinicopathological analysis of 116

adrenal pheochromocytomas and 30 extra-adrenal sympathetic

paragangliomas including 38 malignant tumors. Endocr Pathol.2005;16:23–32.

20. Gao Y, Shimizu M, Yamada S, Ozaki Y, Aso T. The effects of

chemotherapy including cisplatin on vitamin D metabolism.

Endocr J. 1993;40:737–42.

21. Rıos A, Rodrıguez JM, Acosta JM, et al. Prognostic value of

histological and immunohistochemical characteristics for pre-

dicting the recurrence of medullary thyroid carcinoma. Ann SurgOncol. 2010;17:2444–51.

22. Akslen LA. Prognostic importance of histologic grading in pap-

illary thyroid carcinoma. Cancer. 1993;72:2680–5.

23. Lang BH, Lo CY, Chan WF, Lam KY, Wan KY. Staging systems

for papillary thyroid carcinoma: a review and comparison. AnnSurg. 2007;245:366–78.

24. Dralle H, Machens A. Surgical approaches in thyroid cancer and

lymph-node metastases. Best Pract Res Clin Endocrinol Metab.2008;22:971–87.

25. Sugitani I, Toda K, Yamamoto N, Sakamoto A, Fujimoto Y.

Re-evaluation of histopathological factors affecting prognosis of

differentiated thyroid carcinoma in an iodine-sufficient country.

World J Surg. 2010;34:1265–73.

26. Harrell FE. Regression modeling strategies: with applications to

linear models, logistic regression, and survival analysis. New

York: Springer, 2011.

27. Birbeck KF, Macklin CP, Tiffin NJ, et al. Rates of circumfer-

ential resection margin involvement vary between surgeons and

predict outcomes in rectal cancer surgery. Ann Surg. 2002;235:

449–57.

28. Kassahun WT, Jonas S. Focus on parathyroid carcinoma. Int JSurg. 2011;9:13–9.

29. Wiseman SM, Rigual NR, Hicks WL Jr, et al. Parathyroid car-

cinoma: a multicenter review of clinicopathologic features and

treatment outcomes. Ear Nose Throat J. 2004;83:491–4.

30. Schulte KM, Talat N, Miell J, Moniz C, Sinha P, Diaz-Cano S.

Lymph node involvement and surgical approach in parathyroid

cancer. World J Surg. 2010;34:2611–20.


Classification of Parathyroid Cancer

Documents

Transcript of Classification of Parathyroid Cancer