In the study group comprising 114 periampullary adenocarcinomas, there were 67 with pancreatobiliary and 47 with intestinal histologic type of differentiation. These consisted of 40 pancreatic, 41 ampullary, 17 common bile duct, and 16 duodenal adenocarcinomas. Interobserver agreement between the senior and routine pathologist was almost perfect (kappa 0.90; 95% CI 0.82–0.99) for determination of histologic type of differentiation, while it was only fair (kappa 0.37; 95% CI 0.25–0.49) in classification of tumour origin. However, while the routine pathologist who reevaluated the microscopic slides was blinded towards clinical and macroscopic data, the senior pathologist not only reevaluated the microscopic slides but also considered information from the operative and macroscopic reports. This type of information may be more important for accurate tumour origin classification than for histologic type classification. Thus, comparing final consensus with the original reports, interobserver agreement was substantial for both tumour origin classification (25 reclassified cases; kappa 0.68; 95% CI 0.57–0.79) and histologic type classification (13 reclassified cases; kappa 0.74; 95% CI 0.61–0.87).
Histopathologic prognostic factors in periampullary adenocarcinoma
As expected patients with pancreatic tumours had the poorest prognosis among all periampullary adenocarcinomas in univariate analysis (p < 0.001, figure ). Pancreatobiliary type of differentiation was an adverse predictor of survival both in the whole cohort of periampullary adenocarcinomas (p < 0.001, figure ) and in the ampullary subgroup (p < 0.022, figure ).
Overall survival after pancreaticoduodenectomy for periampullary adenocarcinoma (n = 114) originating in duodenum (n = 16), ampulla (n = 41), distal bile duct (n = 17), and pancreas (n = 40) (p < 0.001).
Figure 3 Overall survival after pancreaticoduodenectomy: (A) Periampullary adenocarcinoma (n = 114) with intestinal (n = 47) and pancreatobiliary (n = 67) type of histologic differentiation (p < 0.001). (B) Ampullary adenocarcinoma (n = 41) with intestinal (more ...)
Table describes associations between histologic type of differentiation and the other histopathologic factors. Compared to intestinal type adenocarcinomas, pancreatobiliary type adenocarcinomas significantly more often showed presence of histopathologic features associated with a poor prognosis, in particular resection margin involvement, perineural infiltration, areas with poor differentiation, advanced pT stage, and pancreatic tumour origin (p < 0.001 for each).
Associations between histologic type of differentiation and other histopathologic factors in 114 periampullary adenocarcinomas
In multivariable analysis adjusting for tumour origin, pT stage, maximum tumour diameter, degree of differentiation, regional lymph node metastasis, resection margin involvement, vessel involvement, and perineural infiltration, the histologic type of differentiation was found to be an independent predictor of survival (p = 0.032; HR 2.8; 95% CI 1.1–7.1), while tumour origin was only borderline significant (p = 0.054). Stepwise backward variable selection resulted in a final model that included the histologic type of differentiation, which in fact was the strongest predictor of survival (table ). The validity of the final model was tested in the ampullary subgroup, confirming that pancreatobiliary versus intestinal type of differentiation was an independent adverse predictor of survival also among these patients (p < 0.002; HR 4.0; 95% CI 1.6–9.6).
Multivariable Cox regression analysis of histopathologic prognostic factors in periampullary adenocarcinomas (n = 114)
Prognostic factors in pancreatobiliary differentiated periampullary adenocarcinomas
A separate analysis of histopathologic factors among the patients who had pancreatobiliary adenocarcinoma (n = 67) was performed in order to identify prognostic factors in this subgroup, and in particular, to evaluate whether tumour origin could independently predict the prognosis in pancreatobiliary type adenocarcinoma. In univariate survival analysis, pancreatic tumour origin was significantly associated with a poorer prognosis compared to non-pancreatic tumour origin (table , figure ). Even when adjusting for pT stage, the difference in survival between patients who had pancreatic and non-pancreatic tumour origin was statistically significant (p = 0.003; HR = 2.9; 95% CI 1.4–6.0). However, adjusting for tumour diameter instead of pT stage demonstrated that there was in fact no survival difference between patients who had pancreatic and non-pancreatic tumours (p = 0.25). The pT staging for periampullary adenocarcinomas is based on the assumption that clinical outcome depends more on tumour extension beyond organ of origin than of tumour size. An ampullary pT3 tumour slightly invading the pancreas may thus be as small as 1 cm, while pancreatic pT3 tumours are normally much larger. In the present subgroup analysis of pancreatobiliary differentiated tumours, pancreatic pT3 tumours were significantly larger than non-pancreatic pT3 tumours (median diameter 3.5 versus 2.6 cm; p = 0.033).
Survival analysis of histopathologic prognostic factors in pancreatobiliary resections (n = 67)
Figure 4 Pancreaticoduodenectomy for pancreatobiliary type periampullary adenocarcinoma: (A) Univariate survival for patients with pancreatobiliary type periampullary adenocarcinomas (n = 67) originating in ampulla (n = 13), distal bile duct (n = 16), and pancreas (more ...)
Among all 67 pancreatobiliary differentiated adenocarcinomas, pancreatic tumours significantly more often than non-pancreatic tumours had regional lymph node metastasis (29/38 for pancreatic versus 12/29 for non-pancreatic tumours, p = 0.004) and were larger (median diameter 3.1 cm for pancreatic versus 2.0 cm for non-pancreatic tumours, p < 0.001) (figure ). The differences in survival seen among all cases (figure ) are thus related to the differences in tumour diameter (figure ) between non-pancreatic and pancreatic tumours. Selecting tumours of comparable size (range 2.0–3.0 cm, n = 30) demonstrated no difference in survival between pancreatic (n = 18) and non-pancreatic cases (n = 12) (p = 0.851, figure ). These groups were comparable with respect to tumour diameter (median 2.5 and mean 2.4 cm for both groups), and the equal survival was not due to less frequent lymph node metastasis among pancreatic cases (positive lymph nodes in 15/18 pancreatic compared to 6/12 non-pancreatic cases).
Starting with all the histopathologic factors in the base model for multivariable analysis, backward variable selection thus resulted in a final model that did not include tumour origin (table ). Only lymph node status, vessel involvement and tumour diameter independently predicted the prognosis after resection of pancreatobiliary type periampullary adenocarcinoma. The final model obtained from stepwise backward analysis was confirmed by repeating variable selection with forward stepwise analysis. Although perineural infiltration also seemed to be an important prognostic factor in univariate analysis, this factor did not independently predict survival, due to a strong association with lymph node metastasis (p = 0.002), vessel involvement (p < 0.001), and tumour diameter (p = 0.024).
Validation of main conclusions in an independent dataset
We finally validated our main findings by performing a separate analysis of histopathologic prognostic factors in the historical control group consisting of patients operated in our institution before standardization of histopathologic assessment. Among these patients, 73 and 26 were upon reevaluation of the histologic slides found to have pancreatobiliary and intestinal differentiation, respectively. Tumour origin for all 99 cases was classified as ampullary (n = 23), duodenal (n = 14), distal bile duct (n = 10), and pancreatic (n = 52).
In univariate survival analysis, pancreatobiliary type of differentiation (p < 0.001) and pancreatic tumour origin (pancreatic versus ampullary, p = 0.03) both predicted a poor prognosis. Adjusting for maximum tumour diameter, lymph node and resection margin involvement, degree of differentiation, and whether there was presence of perineural infiltration, the histologic type of differentiation remained highly significant (p < 0.001; HR 2.7; 95% CI 1.5–4.9). In contrast, although approaching significance, tumour origin did not significantly predict the prognosis after adjustment for these other factors (pancreatic versus ampullary, p = 0.10; HR 1.6; 95% CI: 0.9–2.9).
In the subgroup analysis including only pancreatobiliary differentiated tumours (n = 73), stepwise variable selection resulted in a final multivariable model in which lymph node status (N1 versus N0, p < 0.001; HR 3.4; 95% CI: 1.9–6.0) was confirmed to independently predict the prognosis, adjusting for grade of differentiation (high/moderate versus low, p = 0.002; HR 2.6; 95% CI 1.4–4.8) and resection margin status (R1/2 versus R0, p = 0.011; HR 1.9; 95% CI 1.2–3.2). Although significant in univariate analysis (p = 0.021), tumour diameter was not confirmed to independently predict survival in this cohort, possibly due to the small size of these tumours (mean diameter 2.1 cm; 95% CI 1.9–2.4) and a significant association between tumour diameter and lymph node status (p = 0.04). Finally, after stepwise variable selection, tumour origin and perineural infiltration did not remain in the final multivariable model evaluating histopathologic factors of pancreatobiliary differentiated periampullary adenocarcinomas, in accordance with the results from the study group analysis.