Assessment of Immunohistochemical Staining of Calgranulin A
Only the staining reaction within tumor cells was used for the classification of the immunohistochemical staining patterns. Calgranulin A expression was preferably detected within the cytoplasm of tumor cells (Fig. 1). In a few cases, nuclear staining was observed, however, this revealed no prognostic information for the present cohort of patients.
Correlation of Calgranulin A Expression and Clinicopathological Parameters
Of the 158 NMIBC specimens included in the present investigation, 122 (77.2%) were from male and 36 (22.8%) from female patients. The median age at diagnosis was 69.0 years (range 32–90) and the median follow-up interval was 49.5 months (range 4–139). According to histopathologic evaluation, tumor stages were classified as pTa in 89 (56.3%) and pT1 in 69 (43.7%). Concomitant carcinoma in situ (pTis) was found in 7 (4.4%) patients, one in stage pTa and six in stage pT1. The tumor grades of the cases included in the prognostic evaluation were as follows: G1 in 60 (38.0%) cases, G2 in 86 (54.4%), and G3 in 12 (7.6%) (Table 1). Twenty-four cases with stage ≥ pT2 were compared to the NMIBC specimens. Consequently, 182 specimens were included in the analysis of calgranulin A expression regarding tumor stage and grade, of which 89 (48.9%) were pTa, 69 (37.9%) pT1, and 24 (13.2%) ≥ pT2. The distribution of tumor grades in the 182 cases was G1 in 60 (33.0%), G2 in 93 (51.1%), and G3 in 29 (15.9%) cases.
Calgranulin A expression patterns were tested for various patient and tumor characteristics, such as patient’s age and gender, the detection of pTis in addition to the primary papillary lesion, multifocality, and tumor stage and grade. There was no positive or negative correlation between patient age and calgranulin A expression levels (sample size 158, Spearman’s correlation coefficient rs = 0.13, p = 0.103). Moreover, there were no significant differences in calgranulin A expression levels for gender (sample size 158, p = 0.108, Mann–Whitney U test). Furthermore, patients with or without multifocal tumor growth showed no significant differences in calgranulin A expression levels (sample size 158, p = 0.075, Mann–Whitney U test). Because the results for age, gender, and multifocality were close to significance, further tests of these variables were also included in the Kaplan–Meier survival analysis. Simultaneous detection of pTis in the specimens showed significant differences in calgranulin A expression levels when compared to specimens without pTis (sample size 158, p = 0.030, Mann–Whitney U test).
Higher calgranulin A expression was significantly associated with a higher tumor stage (sample size 182, p = 0.000, Kruskal–Wallis test) and shows significant difference between pTa and pT1 lesions (sample size 158, p = 0.000, Mann–Whitney U test). Calgranulin A expression was also significantly divergent between tumor grades (sample size 182, p = 0.015, Kruskal–Wallis test) (Fig. 2).
Correlation of Calgranulin A Expression and Follow-up Data (Recurrence-Free Survival, Progression-Free Survival, Cancer-Specific Survival)
Information about RFS, PFS, and CSS at the end of the follow-up period was available in 158, 158, and 145 cases, respectively. Recurrence occurred in 67 (42.4%) patients, of which 39 (58.2%) had stage pTa and 28 (41.8%) stage pT1. Progression developed in 24 (15.2%) of the patients. In the group with progression, 11 (45.8%) had stage pTa and 13 (54.2%) stage pT1. The median time to recurrence was 10 months (range 3–72) and the median time to progression was 11 months (range 0–79). Cancer-specific death occurred in 14 (8.9%) patients and the median time to cancer-specific death was 29.0 months (range 6–83).
Cutoff calculation resulted in different cutoff values of calgranulin A staining index: 95 for RFS and 108 for PFS and CSS. Calgranulin A staining indices less than the cutoff values were classified as low calgranulin A expression, and values greater than or equal to the cutoff as high calgranulin A expression.
In the Kaplan–Meier survival analysis we found significant differences between low and high calgranulin A expression in terms of RFS (sample size 158, 5y-RFS 70.4 ± 4.0% vs. 35.9 ± 12.5%, median RFS not reached (NR) vs. 12.0 ± 4.4 month (95% CI 3.3–20.7), p = 0.029, log-rank test), PFS (sample size 158, 5y-PFS 90.3 ± 2.7% vs. 51.5 ± 14.0%, median PFS NR in both groups, p = 0.000, log-rank test), and CSS (sample size 145, 5y-CSS 92.9 ± 2.6% vs. 70.7 ± 12.4%, median CSS NR in both groups, p = 0.005, log-rank test) (Fig. 3). Moreover, Kaplan–Meier survival analysis for RFS dependent on age (< 69 years vs. ≥ 69 years) showed no significant difference (sample size 158, 5y-RFS 59.5 ± 5.9% vs. 50.0 ± 6.1%, median RFS NR vs. 40.0 ± 13.0 month (95% CI 14.4–65.6), p = 0.103, log-rank test). On the other hand, Kaplan–Meier survival analysis for PFS (sample size 158, 5y-RFS 88.2 ± 3.9% vs. 50.0 ± 6.1%, median RFS NR in both groups, p = 0.030, log-rank test) and CSS (sample size 145, 5y-RFS 91.9 ± 4.2% vs. 76.4 ± 7.2%, median RFS NR in both groups, p = 0.036, log-rank test) resulted in significant differences dependent on age. Furthermore, there were no significant differences in Kaplan–Meier survival analysis comparing gender (male vs. female) against RFS (sample size 158, 5y-RFS 49.7 ± 5.0% vs. 62.8 ± 8.9%, median RFS NR in both groups, p = 0.423, log-rank test), PFS (sample size 158, 5y-RFS 83.6 ± 3.9% vs. 75.1 ± 8.6%, median RFS NR in both groups, p = 0.357, log-rank test), and CSS (sample size: 145, 5y-RFS 91.0 ± 3.4% vs. 81.8 ± 7.5%, median RFS NR in both groups, p = 0.233, log-rank test). Surprisingly, Kaplan–Meier survival analysis comparing unifocal tumors vs. multifocal tumors were not significant for RFS (sample size 158, 5y-RFS 57.7 ± 5.4% vs. 44.1 ± 7.2%, median RFS not reached (NR) vs. 26.0 ± 15.6 month (95% CI 0.0-56.6), p = 0.151, log-rank test) and CSS (sample size 145, 5y-RFS 88.4 ± 4.0% vs. 89.9 ± 4.8%, median RFS not NR in both groups, p = 0.892, log-rank test), but resulted in a significant difference concerning PFS (sample size 158, 5y-RFS 86.2 ± 4.0% vs. 74.1 ± 6.7%, median RFS not NR in both groups, p = 0.044, log-rank test).
Cox proportional hazard analysis was performed to evaluate independent variables for RFS, PFS, and CSS (Table 2). As the limited number of events concerning recurrence (n = 67), progression (n = 24), and cancer-specific deaths (n = 14) in our dataset was a delimiter for the number of covariates in multivariate Cox regression, we decided to determine the covariates based on the already existing EORTC risk factors (without prior recurrence rate, because of the fact our cases contained only primary tumors). Covariates in the Cox proportional hazard analysis were calgranulin A expression (low vs. high), tumor stage (pTa vs. pT1), concomitant pTis (no vs. yes), tumor size (< 3 cm vs. ≥ 3 cm), number of tumors, and tumor grade (G1 vs. G2 or G3). In the univariate Cox regression calgranulin A expression was consistently the only significant factor among the chosen covariates regarding RFS (sample size 158, p = 0.036, HR 2.06), PFS (sample size 158, p = 0.000, HR 5.35), and CSS (sample size 145, p = 0.011, HR 4.55). In the multivariate Cox regression calgranulin A expression remained an independent factor for RFS (sample size 158, p = 0.024, HR 2.43) and PFS (sample size 158, p = 0.002, HR 5.92), but failed to be significant in CSS (sample size 145, p = 0.147, HR 3.24).