L1CAM in Early-Stage Type I Endometrial Cancer
L1CAM in Early-Stage Type I Endometrial Cancer
Of the 1021 investigated FIGO stage I endometrioid endometrial carcinomas, 181 (17.7%) were found to be L1CAM-positive. One hundred thirty-seven cancers (75.7%) exhibited focal staining of cell clusters, and 44 (24.3%) showed diffuse immunostaining in more than 50% of cancer cells. No special staining pattern, such as isolated or particularly intensive L1CAM expression at the myoinvasive front, was revealed. Of the included cancers, 96.4% were purely endometrioid carcinomas, whereas the remaining 3.6% showed areas of nonendometrioid differentiation (all comprising <10% of the tumor). Cancers containing a second cell type were more frequently L1CAM-positive than were purely endometrioid carcinomas (P = .003) (Table 1). Stage Ib cancers were more frequently L1CAM-positive than were stage Ia carcinomas (P < .001).
Median age of the entire study population at diagnosis was 64 years (range = 34–96 years). L1CAM status was not related to patient age at diagnosis or to the classical epidemiological risk factors for type I endometrial cancer such as diabetes, obesity, nulliparity, hypertension, and unopposed estrogen exposure (Supplementary Table 1, available online).
Classical risk assessment for the entire cohort is listed in Table 1 together with the corresponding L1CAM-positive rates. In the low-risk group, 13.2% of the cancers were L1CAM-positive, whereas in intermediate- and high-risk cancers the positive rate was 25.8% (P < .001). L1CAM positivity was associated with histopathological grade (P < .001) and increasing depth of myometrium infiltration (P < .001).
During a median follow-up of 5.3 years, 117 patients (11.5%) experienced recurrence. Of these recurrences, 69.2% (n = 81) and 94.9% (n = 111) occurred during the first 2 and 5 years, respectively. With regard to L1CAM status, 51.4% (n = 93) of the L1CAM-positive tumors and 2.9% (n = 24) of the L1CAM-negative tumors recurred. As depicted in Figure 1A, in L1CAM-positive and L1CAM-negative cancers, 69.9% (n = 65) and 66.7% (n = 16) of the observed recurrences, respectively, occurred within 2 years after initial treatment. Time to recurrence, when subdivided into early (≤2 years), intermediate (>2 and ≤5 years) and late (>5 years) relapses, was unrelated to L1CAM status. For the 117 observed recurrences, the crude L1CAM-positive rate was 79.5%, and the crude L1CAM-negative rate was 20.5%. Moreover, regarding distant recurrences the L1CAM-positivity rate was even higher, namely 85.7% (66 of 77 events). In Table 2, observed study events were compared with the estimates generated by the Peto-modified Mantel–Haenzel approach. Whereas for L1CAM-positive cancers this approach estimated that the "observed" frequencies exceed by far the "expected" frequencies of recurrences and deaths, the diametric opposite was assessed for L1CAM-negative tumors (P < .001). This was true for the whole study population and for the separated cancer risk classes.
In univariate survival analyses, disease-free and overall survival were poorer in patients with L1CAM-positive cancers than in patients with tumors lacking relevant L1CAM expression (P < .001) (Figure 1). Median disease-free and overall survival in patients with L1CAM-positive tumors were 4.5 years and 8.9 years, respectively, whereas median disease-free and overall survival were not reached in L1CAM-negative cancers. Furthermore, it should be stated that disease-free and overall survival were poorer for cancers with diffuse L1CAM immunostaining in more than 50% of the tumor cells than for cancers exhibiting focal (10%–50%) L1CAM expression (P < .001) (Supplementary Figure 3, available online).
(Enlarge Image)
Figure 3.
Univariate disease-free survival analyses for Fédération Internationale de Gynécologie et d'Obstétrique (FIGO) stage, risk assessment, and grading according to the L1CAM status. A) FIGO stage: Ia (red), Ib (black). B) Risk assessment: low risk (red), intermediate (Interm)/high Risk (black). C) Histopathological grading: grade I (GI) (red), grade II (GII) (blue), grade III (GIII) (black). Similar results were obtained for overall survival (shown in Supplementary Figure 4, A–C, available online). Differences between clinicopathologic groups were assessed by the log-rank test. The numbers of patients at risk are given below the graphs. G = grade; H = high risk; I = intermediate risk; Ia and Ib = FIGO; L1 = L1CAM; stages; Low = low risk.
Furthermore, differences in disease-free and overall survival between L1CAM-positive and -negative cancers were more prominent in tumors classified as intermediate- and high-risk than in low-risk tumors. A similar adverse impact of L1CAM positivity was seen in FIGO stage Ib (Figure 2). Interestingly, in L1CAM-negative cancers, disease-free and overall survival did not differ statistically between FIGO stages Ia and Ib, between the various grades of differentiation, or between the conventional risk classes. As depicted in Figure 3 and Supplementary Figure 4 (available online), only when tumors were L1CAM-positive did the mentioned variables exhibit statistically significant relevance for patient survival (Supplementary Table 2, available online).
In the multivariable Cox model, L1CAM expression retained independent prognostic significance for disease-free as well overall survival (P < .001). It is noteworthy that L1CAM-positivity exhibited the most impressive hazard ratios (HRs) for recurrence (HR = 16.33; 95% confidence interval [CI] = 10.55 to 25.28) and death (HR = 15.01; 95% CI = 9.28 to 24.26), as compared with the other prognostic variables included in the multivariable calculations (Table 3). In L1CAM-positive cancers, the hazard ratios were 13.37 (95% CI = 6.71 to 26.67) for predicting locoregional recurrence and 34.07 (95% CI = 17.06 to 68.04) for predicting distant recurrence (Table 4). Removal cancers containing small (<10%) serous or clear-cell components from the calculations did not substantially affect the study outcome (Supplementary Table 3, available online).
Additionally, when the CRT decision tree was used as a classifier, only L1CAM was selected for the final model to predict recurrence and death within 5 years, with a sensitivity of 0.74 and a specificity of 0.91 (accuracy = 87.90%) and a sensitivity of 0.77 and a specificity of 0.89 (accuracy = 87.60%), respectively.
Results
Of the 1021 investigated FIGO stage I endometrioid endometrial carcinomas, 181 (17.7%) were found to be L1CAM-positive. One hundred thirty-seven cancers (75.7%) exhibited focal staining of cell clusters, and 44 (24.3%) showed diffuse immunostaining in more than 50% of cancer cells. No special staining pattern, such as isolated or particularly intensive L1CAM expression at the myoinvasive front, was revealed. Of the included cancers, 96.4% were purely endometrioid carcinomas, whereas the remaining 3.6% showed areas of nonendometrioid differentiation (all comprising <10% of the tumor). Cancers containing a second cell type were more frequently L1CAM-positive than were purely endometrioid carcinomas (P = .003) (Table 1). Stage Ib cancers were more frequently L1CAM-positive than were stage Ia carcinomas (P < .001).
Median age of the entire study population at diagnosis was 64 years (range = 34–96 years). L1CAM status was not related to patient age at diagnosis or to the classical epidemiological risk factors for type I endometrial cancer such as diabetes, obesity, nulliparity, hypertension, and unopposed estrogen exposure (Supplementary Table 1, available online).
Classical risk assessment for the entire cohort is listed in Table 1 together with the corresponding L1CAM-positive rates. In the low-risk group, 13.2% of the cancers were L1CAM-positive, whereas in intermediate- and high-risk cancers the positive rate was 25.8% (P < .001). L1CAM positivity was associated with histopathological grade (P < .001) and increasing depth of myometrium infiltration (P < .001).
During a median follow-up of 5.3 years, 117 patients (11.5%) experienced recurrence. Of these recurrences, 69.2% (n = 81) and 94.9% (n = 111) occurred during the first 2 and 5 years, respectively. With regard to L1CAM status, 51.4% (n = 93) of the L1CAM-positive tumors and 2.9% (n = 24) of the L1CAM-negative tumors recurred. As depicted in Figure 1A, in L1CAM-positive and L1CAM-negative cancers, 69.9% (n = 65) and 66.7% (n = 16) of the observed recurrences, respectively, occurred within 2 years after initial treatment. Time to recurrence, when subdivided into early (≤2 years), intermediate (>2 and ≤5 years) and late (>5 years) relapses, was unrelated to L1CAM status. For the 117 observed recurrences, the crude L1CAM-positive rate was 79.5%, and the crude L1CAM-negative rate was 20.5%. Moreover, regarding distant recurrences the L1CAM-positivity rate was even higher, namely 85.7% (66 of 77 events). In Table 2, observed study events were compared with the estimates generated by the Peto-modified Mantel–Haenzel approach. Whereas for L1CAM-positive cancers this approach estimated that the "observed" frequencies exceed by far the "expected" frequencies of recurrences and deaths, the diametric opposite was assessed for L1CAM-negative tumors (P < .001). This was true for the whole study population and for the separated cancer risk classes.
In univariate survival analyses, disease-free and overall survival were poorer in patients with L1CAM-positive cancers than in patients with tumors lacking relevant L1CAM expression (P < .001) (Figure 1). Median disease-free and overall survival in patients with L1CAM-positive tumors were 4.5 years and 8.9 years, respectively, whereas median disease-free and overall survival were not reached in L1CAM-negative cancers. Furthermore, it should be stated that disease-free and overall survival were poorer for cancers with diffuse L1CAM immunostaining in more than 50% of the tumor cells than for cancers exhibiting focal (10%–50%) L1CAM expression (P < .001) (Supplementary Figure 3, available online).
(Enlarge Image)
Figure 3.
Univariate disease-free survival analyses for Fédération Internationale de Gynécologie et d'Obstétrique (FIGO) stage, risk assessment, and grading according to the L1CAM status. A) FIGO stage: Ia (red), Ib (black). B) Risk assessment: low risk (red), intermediate (Interm)/high Risk (black). C) Histopathological grading: grade I (GI) (red), grade II (GII) (blue), grade III (GIII) (black). Similar results were obtained for overall survival (shown in Supplementary Figure 4, A–C, available online). Differences between clinicopathologic groups were assessed by the log-rank test. The numbers of patients at risk are given below the graphs. G = grade; H = high risk; I = intermediate risk; Ia and Ib = FIGO; L1 = L1CAM; stages; Low = low risk.
Furthermore, differences in disease-free and overall survival between L1CAM-positive and -negative cancers were more prominent in tumors classified as intermediate- and high-risk than in low-risk tumors. A similar adverse impact of L1CAM positivity was seen in FIGO stage Ib (Figure 2). Interestingly, in L1CAM-negative cancers, disease-free and overall survival did not differ statistically between FIGO stages Ia and Ib, between the various grades of differentiation, or between the conventional risk classes. As depicted in Figure 3 and Supplementary Figure 4 (available online), only when tumors were L1CAM-positive did the mentioned variables exhibit statistically significant relevance for patient survival (Supplementary Table 2, available online).
In the multivariable Cox model, L1CAM expression retained independent prognostic significance for disease-free as well overall survival (P < .001). It is noteworthy that L1CAM-positivity exhibited the most impressive hazard ratios (HRs) for recurrence (HR = 16.33; 95% confidence interval [CI] = 10.55 to 25.28) and death (HR = 15.01; 95% CI = 9.28 to 24.26), as compared with the other prognostic variables included in the multivariable calculations (Table 3). In L1CAM-positive cancers, the hazard ratios were 13.37 (95% CI = 6.71 to 26.67) for predicting locoregional recurrence and 34.07 (95% CI = 17.06 to 68.04) for predicting distant recurrence (Table 4). Removal cancers containing small (<10%) serous or clear-cell components from the calculations did not substantially affect the study outcome (Supplementary Table 3, available online).
Additionally, when the CRT decision tree was used as a classifier, only L1CAM was selected for the final model to predict recurrence and death within 5 years, with a sensitivity of 0.74 and a specificity of 0.91 (accuracy = 87.90%) and a sensitivity of 0.77 and a specificity of 0.89 (accuracy = 87.60%), respectively.
Source...