Prognostic value of p27^kip1 expression in adenocarcinoma of the pancreatic head region

Abstract

Background. p27^kip1 is a tumour suppressor gene, functioning as a cyclin-dependent kinase inhibitor, and an independent prognostic factor in breast, colon, and prostate adenocarcinomas. Conflicting data are reported for adenocarcinoma of the pancreas. The aim of this study was to establish the prognostic value of p27^kip1 expression in adenocarcinoma of the pancreatic head region. Patients and methods. The study included 45 patients (male/female ratio 2:1; mean age 59, range 38–82 years) with adenocarcinomas of the pancreatic head region: 24 – pancreatic head, 18 – periampullary and 3 – uncinate process. The patients underwent the Kausch-Whipple pancreatoduodenectomy (n=39), pylorus-preserving pancreatoduodenectomy (n=5), or nearly total pancreatectomy (n=1). Eight patients received adjuvant chemotherapy postoperatively. Follow-up time ranged from 3 to 60 months. Tumours were staged according to the pTNM classification (UICC 1997). Immunohistochemistry was done on paraffin-embedded blocks from tumour sections. Quantitative determination of p27^kip1 expression was based on the proportion of p27^kip1 -positive cells (< 5% = negative). Survival analysis was carried out using the Kaplan-Meier method and Cox regression model. Results. Positive p27^kip1 expression was detected in 22 tumours (49%), whereas 23 tumours (51%) were p27^kip1-negative. There were no significant correlations between p27^kip1 index and stage or lymph node involvement. Median survival time in patients with p27^kip1-positive tumours was 19 months, whereas in patients with p27^kip1-negative tumours it was 18 months (p=0.53). A significant relationship was found between p27^kip1-negative tumours and radical resection (p=0.04). Multivariate survival analysis revealed that the localization of the tumour (pancreatic head/uncinate process vs periampullary) was the only significant and independent prognosticator (p = 0.01, Cox regression model). Resection margins involvement and grade remained nearly significant prognostic factors (p=0.07 and p=0.09, respectively). Conclusion. We conclude that p27^kip1 has limited overall prognostic utility in resected carcinoma of the pancreatic head region, but its potential role as a marker of residual disease needs to be further assessed.

Introduction

The incidence of pancreatic carcinoma has increased during recent decades, and it is the fourth most common cause of cancer-related mortality ¹,². Early systemic dissemination and local cancer progression are limitations of curative treatment ³. Overall cumulative 5-year survival is extremely poor (0.4–1%), and after curative resection is only 5–15% ⁴,⁵,⁶,⁷. Pathological stage, histological differentiation, localization of the tumour and involved margins are known prognostic factors ⁸,⁹,¹⁰,¹¹,¹²,¹³.

The serum concentration of CA 242, CA 19.9 and MUC1 markers may have the same prognostic significance only in advanced pancreatic cancer ¹⁴,¹⁵,¹⁶. Additional molecular prognosticators, such as p53, cyclin D1, matrix metalloproteinase 1, vascular endothelial growth factor, transforming growth factor beta and DNA ploidy have also been described ¹⁷,¹⁸,¹⁹,²⁰,²¹,²².

Neoplastic cells progress rapidly through the cell cycle, often ignoring the cell division control mechanisms. Cyclin-dependent kinase (CDK) plays an important role in cell cycle control. p27^kip1 is a putative tumour suppressor gene and p27^kip1 protein is regarded as an inhibitor of the G1-to-S cell cycle progression by suppressing the kinase activity of cyclin/cyclin-dependent kinase inhibitor family (D-CDK4 and E-CDK2). In addition, p27 is a regulator of drug resistance in solid tumours and promoter of apoptosis ²³,²⁴,²⁵. It has been characterized as an independent prognostic factor in breast, colon, and prostate adenocarcinomas ²⁶. There are conflicting data on the prognostic significance of p27 expression in adenocarcinoma of the pancreas ²⁷,²⁸.

The aim of this study was to establish the prognostic value of p27^kip1 expression in adenocarcinoma of the pancreatic head region following surgical treatment.

Patients and methods

Patients and tumour samples

Between October 1998 and December 2003, 59 consecutive patients with histologically verified pancreatic adenocarcinoma and one patient with cystadenocarcinoma underwent resection. Tissue specimens were evaluated only from 45 patients with proven adenocarcinoma of the pancreatic head region: 24 (53%) pancreatic head, 18 (40%) periampullary and three (7%) uncinate process (male/female ratio 2:1; mean age±SD, 59.4 ±11.8, range 38–82 years).

The patients underwent Kausch-Whipple pancreatoduodenectomy (n=39, 87%), Travesrso-Longmire pylorus-preserving pancreato-duodenectomy (n=5, 11.1%), or nearly total pancreatectomy (n=1, 2.2%) with no postoperative mortality. Eight patients (18%) received adjuvant chemotherapy postoperatively. Hospitalization time ranged from 8 to 46 days (mean time ±SD = 19 ±10.4).

The evaluation of the tumour stage was based on the pTNM classification (UICC 1997). Differentiation of adenocarcinoma was described as well/moderate or poor grade. Residual tumour classification (R) was used for assessment of complete tumour removal. Four margins were assessed pathologically following precise marking (coloured pins) in the resection specimen by a surgeon immediately after the operation: 1) common bile duct, 2) stomach or duodenum, 3) pancreatic, 4) retroperitoneal. Resections were classified as microscopically radical (R0), microscopically non-radical (R1), or macroscopically non-radical (R2). Follow-up time ranged from 3 to 62 months (median 19 months).

Immunohistochemistry

Tissue material was routinely fixed for 12–24 hours in 4% buffered formalin of a volume adequate to the size of the surgical specimen. Immunohistochemical staining was performed on processed paraffin-embedded blocks from representative tumour sections using monoclonal antibody p27^Kip1 (clone SX53G8, Dakocytomation), which has been confirmed to recognize p27^kip1 protein exclusively, showing that p27^kip1 staining was absent when the antibody was preabsorbed with blocking p27^kip1 peptide ²⁹,³⁰,³¹. Tumour-specific staining not seen in the surrounding pancreatic or duodenal tissue, with a distinct staining in the cell nucleus and with markedly less or no immunoreactivity in the tumour cell cytoplasm, was considered as a positive cell. Expression of p27^Kip1 in small lymphocytes was used as an internal positive control of immunostaining. Sections from pancreatic adenocarcinomas incorporated in the present study material that underwent the staining procedure, with omission of the specific antibody, were used as negative controls.

Quantitative determination of p27^kip1 expression was based on the proportion of p27^kip1-positive cells (p27^kip1 labelling index). Cases with < 5% positive cells were considered negative.

Statistical analysis

The Pearson's χ² test and Fischer's exact test were used to test the association between the p27 labelling index and clinico-pathological features. The Kaplan-Meier method and the log rank test were used for univariate analysis of the relationship between clinico-pathological features and survival. Multivariate survival analysis was performed with a Cox regression model entering the following covariates: p27^kip1 expression, localization of the tumour (pancreatic head/uncinate process vs periampullary), size of the tumour (< 3 cm vs ≥3 cm), resection margins status (R0 vs R1/R2) and differentiation of the tumour (G1/G2 vs G3).

A p value < 0.05 was considered to represent a statistically significant difference. Statistical calculations were performed using Statistica 6.1 for Windows (StatSoft Inc., Tulsa, OK, USA).

Results

There were 45 patients with adenocarcinoma of the pancreatic head region. In 39 of the patients, the tumour was graded as well/moderately differentiated (G1/G2). In the remaining six patients, adenocarcinomas were poorly differentiated (G3). Distribution of the pathological stages was as follows: stage I, 14 patients; stage II, 15 patients; stage III, 14 patients, and stage IV, two patients. Tumour-positive lymph nodes were detected in 12 patients. The number of investigated lymph nodes and number of tumour-positive lymph nodes were as follows (range, mean±SD): 1–16, 4.2±4.0 and 1–13, 5.6±3.5, respectively. An R0 resection was achieved in 33 of 45 cases. A resection line involvement was microscopically recognized in 4 of 45 cases (R1 resection). In eight patients, the tumour mass was removed incompletely as assessed macroscopically (R2 resection). Tumour size ranged from 0.5 to 6.0 cm (mean±SD = 3.5±1.6). Jaundice before operation was detected in 36 (80%) patients.

Positive p27^kip1 expression was detected in 22 tumours (49%), whereas 23 tumours (51%) were p27^kip1-negative (Figure 1). Patients’ gender, age, tumour size, UICC stage, lymph node metastasis, preoperative jaundice, differentiation and localization of the tumour were not related to p27^kip1 labelling index. Among patients with p27^kip1-negative tumours radical resection (R0) was performed in 20 (86%) patients, whereas non-radical (R1/R2) resection was carried out in 3 (14%) patients. In the group of patients with a p27^kip1 labelling index ≥ 5%, an R0 resection was performed in 13 (60%) patients, whereas non-radical resection was carried out in 9 (40%) patients. A significant relationship was found between p27^kip1-negative tumours and radical resection (p=0.04; Fisher's exact test) (Table I).

Figure 1. .

Sections of adenocarcinoma of the pancreas. (A) p27^Kip1-positive nuclear immunohistochemistry staining. (B) p27^Kip1-negative immunohistochemistry staining. Magnification ×200.

Table I. Association between p27^kip1 expression and pathological features of carcinomas (Fisher's exact test).

Variables	Number of patients with LI <5% p27kip1	Number of patients with LI ≥5% p27kip1	Total (n)	p value
Gender
Female	9	6	15
Male	14	16	30	0.5
Age (years)
≤ 60	15	9	24
> 60	8	13	21	0.13
Tumour size (cm)
≤ 3	10	12	22
> 3	13	9	22	0.54
NA			1
Stage
I	9	5	14
II	6	9	15
III	8	6	14
IV	0	2	2	0.26*
Lymph node metastases
Present	7	5	12
Absent	16	17	33	0.7
Jaundice preoperatively
Present	21	15	36
Absent	2	7	9	0.07
Differentiation of the tumour
Well/moderate (G1/G2)	18	21	39
Poor (G3)	5	1	6	0.18
Resection margins
Negative (R0)	20	13	33
Involved (R1 + R2)	3	9	12	0.04
Localization of the tumour
Periampullary	9	9	18
Pancreatic head/uncinate process	14	13	27	0.9

LI, labelling index; NA, not assessed.

*Pearson's χ² test.

The 1-, 2- and 3-year survival rates were 72%, 57% and 43%, respectively. Median survival was 19 months. All deaths were related to the pancreatic adenocarcinoma. Nineteen patients (42%) died during the follow-up period of the study (median 11 months). Survival of patients with involved margins (R1/R2) was considerably shorter than that of patients after R0 resection. On univariate survival analysis, tumour size (p=0.0002), localization of the tumour (p = 0.0001), differentiation of the tumour (p=0.006) and involved margins (p=0.01) were significant prognostic factors (Table II).

Table II. Univariate analysis between clinico-pathological features and survival (log rank test).

Variables	Patients n (%)	3-year cumulative survival	Median survival time (months)	p value
Gender
Female	15 (33%)	40%	18
Male	30 (67%)	47%	23	0.9
Age (years)
≤ 60	24 (53%)	48%	19
> 60	21 (47%)	38%	18	0.54
Tumour size (cm)
≤ 3	22 (49%)	77%	26
> 3	22 (49%)	0%	13
NA	1 (2%)			0.0002
Stage
I	14 (31%)	50%	27
II	15 (33%)	46%	16
III	14 (31%)	40%	15
IV	2 (5%)	0%	14	0.25*
Localization of the tumour
Periampullary	18 (40%)	77%	29
Pancreatic head/uncinate process	27 (60%)	14%	10	0.0001
Lymph node metastases
Present	12 (27%)	40%	15	0.41
Absent	33 (73%)	44%	20
Jaundice preoperatively
Present	36 (80%)	39%	19
Not present	9 (20%)	54%	11	0.99
Differentiation of the tumour
Well/moderate (G1/G2)	39 (87%)	50%	19
Poor (G3)	6 (13%)	0	5	0.006
Resection margins
Negative (R0)	33 (73%)	51%	25
Involved (R1 + R2)	12 (27%)	0%	10	0.01
p27kip1 labelling index
< 5%	23 (51%)	38%	18
≥ 5%	22 (49%)	47%	19	0.53
Adjuvant chemotherapy
Surgery + chemotherapy	8 (18%)	25%	14
Surgery alone	37 (82%)	46%	19	0.4

NA, not assessed.

*Cox regression test.

Median survival time in patients with p27^kip1-positive tumours was 19 months, whereas in patients with p27^kip1-negative tumours it was 18 months (p=0.53) (Figure 2). Even when analysis was stratified according to localization of the tumour (pancreatic head/uncinate process vs periampullary), tumour size or involved margins the survival comparisons of patients with p27^kip1-positive and -negative tumours did not reveal significant differences.

Figure 2. .

Overall survival of 45 patients with pancreatic adenocarcinoma in relation to tumour p27^kip1 expression (p=0.53; log rank test).

Multivariate survival analysis included five factors: p27^kip1 expression, localization of the tumour, involved margins, differentiation of the tumour and tumour size. Multivariate survival analysis revealed that the localization of the tumour (pancreatic head/uncinate process vs periampullary) was the only significant and independent prognosticator (p=0.01, Cox regression model; Table III). Resection margins involvement and grade remained nearly significant prognostic factors (p=0.07 and p=0.09, respectively; Table III).

Table III. Multivariate analysis for predictive factors in patients with pancreatic cancer (Cox regression model).

Variables	Risk ratio (RR)	95% CI	p value
Tumour size (cm)	2.848	2.214–3.482	0.15
≤ 3
> 3
Localization of the tumour	3.471	2.210–4.732	0.01
Periampullary
Pancreatic head/uncinate process
Differentiation of the tumour	3.232	2.568–3.896	0.09
Well/moderate (G1/G2)
Poor (G3)
Resection margins	1.252	1.068–1.510	0.07
Negative (R0)
Involved (R1 + R2)
p27kip1 labelling index	1.396	1.000–1.792	0.62
< 5%
≥ 5%

CI, confidence intervals.

Discussion

Tumour suppressor gene p27^kip1 is regarded as an inhibitor of the G1-to-S cell cycle progression by suppressing the kinase activity of cyclin/cyclin-dependent kinase inhibitor family ²⁶. Many reports have validated the utility of p27^kip1 as a prognostic and/or diagnostic marker in breast, colon, lung and prostate adenocarcinomas ²⁶,³²,³³,³⁴. Almost all studies report decreased p27^kip1 expression in more aggressive tumours ²⁶. Contradictory results come from studies on p27^kip1 prognostic value in pancreatic carcinoma ²⁷,²⁹,³⁵,³⁶,³⁷,³⁸.

In the present study a p27^kip1 labelling index of <5% was found in 51% of the tumours. In other studies such low p27^kip1 expression was demonstrated in 41–70% of tumours ²⁷,²⁹,³⁶.

The loss of p27^kip1 expression in the tumour tissue was correlated with progressive stage of the pancreatic neoplasm ²⁸,³⁷. Feakins and Ghaffar and Juuti et al. were not able to find correlation between p27^kip1 expression and tumour stage ²⁷,³⁶. In two studies association of low p27^kip1 expression and poor tumour differentiation was reported ²⁷,³⁸, while in the other reports this phenomenon was not confirmed ²⁸,³⁶. In the present study the p27^kip1 expression did not correlate with tumour stage, differentiation of the tumour or tumour size. These varied results suggest that association between p27^kip1 expression and clinico-pathological features is weak.

Radical resection is a well established significant prognosticator ³⁹. In tumours located in the pancreatic head region, we found a significant relationship between negative p27^kip1 expression and microscopically radical resection (R0). Therefore, low p27^kip1 labelling index may identify patients with pancreatic tumours likely to be removed with negative margins. Our results are contradictory to those of Lu et al., who found correlation between low p27^kip1 expression and increasing peripancreatic invasion that reduced the possibility of achieving complete tumour removal ²⁸. Previous studies had shown that low p27^kip1 expression may contribute to tumour development by either increasing the proliferation of cells or decreasing their apoptosis ⁴⁰,⁴¹,⁴².

On the other hand, findings consistent with ours have been reported in a series of rectal cancer patients receiving preoperative chemoradiation ⁴³. On the basis of immunohistochemical assessment of residual tumour in the resected surgical specimen, patients with rectal cancers exhibiting positive p27^kip1 expression had a fourfold higher risk of recurrence. In that study positive expression of p27^kip1 was an independent negative predictor of recurrence-free survival. The sample size of 45 patients in our study may not be adequate to detect a discrete difference in survival of patients with tumours presenting various degrees of p27^kip1 expression that was associated with completeness of tumour removal. This may be a probable explanation as to why the low p27^kip1 expression was found to be of no prognostic value.

Insignificantly lower 3-year survival rate in the group of patients following resection and adjuvant chemotherapy may be explained by the fact that only patients with more advanced tumours were offered postoperative treatment.

Our study confirmed well-known prognostic factors in pancreatic carcinoma such as tumour size, localization of the tumour, differentiation of the tumour and radicality of resection. We found a correlation between p27^kip1 expression and possibility of curative resection. This relationship, if confirmed in larger series, may serve for selection of patients who possibly can benefit from primary radical surgery.

We conclude that p27^kip1 has limited overall prognostic utility in resected carcinoma of the pancreatic head region, but its potential role as a marker of residual disease needs to be further assessed.