Quantitative differences in immunohistochemical biomarkers between the two histological types
All aforementioned IHC biomarkers were retrieved. The quantitative differences in the IHC biomarkers are illustrated in Table 2 and Fig. 1. When using the H-scores of the SCC IHC markers as a reference, increased expression of E-cadherin, calretinin, CAIX, and c-Myc and decreased levels of HIF-1α, VEGF, TNF-α, galectin-9, CCL5, Bax, EGFR, and IGF-1R were found in AC tumors. The mean PD-L1 CPS score was higher in AC than SCC, but no statistical significance existed.
Table 2
Quantitative differences in the expression levels of the immunohistochemical markers between cervical squamous cell carcinoma and adenocarcinoma
Immunohistochemical markers | squamous cell carcinoma (n = 76) | adenocarcinoma ( n = 42) | P value |
HIF-1 α H-score | 377.86 ± 127.41 | 27.07 ± 34.92 | < 0.001 |
CAIX H-score | 35.53 ± 39.74 | 70.56 ± 64.05 | 0.011 |
Glut1 H-score | 125.13 ± 55.67 | 130.12 ± 54.84 | 0.58 |
VEGF H-score | 76.20 ± 58.88 | 1.51 ± 6.60 | < 0.001 |
EGFR H-score | 37.16 ± 45.88 | 13.95 ± 22.64 | 0.001 |
c-Myc H-score | 5.78 ± 10.09 | 28.22 ± 24.12 | < 0.001 |
IGF-1R H-score | 40.00 ± 32.33 | 13.22 ± 21.17 | < 0.001 |
E-cadherin H-score | 162.50 ± 54.78 | 231.90 ± 54.11 | < 0.001 |
Vimentin H-score | 12.01 ± 21.86 | 18.59 ± 40.35 | 0.09 |
Calretinin H-score | 3.62 ± 9.61 | 229.88 ± 39.94 | < 0.001 |
Bax H-score | 47.13 ± 58.61 | 2.9 0 ± 7.20 | < 0.001 |
Mcl-1 H-score | 123.75 ± 49.24 | 135.88 ± 41.57 | 0.57 |
Bcl-2 H-score | 15.91 ± 24.20 | 39.15 ± 59.31 | 0.43 |
TNF-α H-score | 57.24 ± 38.10 | 32.44 ± 38.81 | < 0.001 |
galectin-9 H-score | 25.54 ± 24.61 | 13.71 ± 16.81 | 0.003 |
CCL5 H-score | 48.08 ± 31.58 | 3.83 ± 16.30 | < 0.001 |
PD-L1 combined positive score | 3.27 ± 4.22 | 12.58 ± 16.07 | 0.13 |
Note: The quantitative differences between H-scores of the biomarkers were examined using the Mann–Whitney U test. |
In summary, cervical SCC exhibited strong expression of the inflammatory, hypoxic, ans angiogenesis markers, whereas the levels of cell to cell adhesion markers were higher in cervical AC. In addition, the Bax levels, an apoptotic activator, were higher in the SCC than in the AC specimens.
Predictive ability s for lymph nodes metastases
Based on the baseline PET/CT, 60 and 15 patients were identified as having pelvic and paraaortic lymph node metastases, respectively. As shown in Supplemental Table 1, ROC curve analysis showed that the H-scores of CCL5 [the area under the ROC curve (AUC): 0.63, P = 0.013], Bcl-2 (AUC: 0.38, P = 0.025), Mcl-1 (AUC: 0.63, P = 0.022) were associated with pelvic lymph node metastasis, whereas the H-scores of Mcl-1 (AUC: 0.67, P = 0.033), TNF-α (AUC: 0.68, P = 0.023), and Glut1 (AUC: 0.71, P = 0.008) were associated with paraaortic lymph node metastasis.
Logistic regression analysis found that a higher CCL5 H-score was a predictor for
pelvic lymph node metastasis [P = 0.004, odds ratio (OR) = 1.018, 95% confidence interval (CI) = 1.006–1.030]. The mean CCL5 H-scores of tumors with and without pelvic lymph node metastasis were 41.90 ± 37.56 and 22.98 ± 28.27, respectively. The analysis also indicated that a higher Glut1 H-score predicted paraaortic lymph node metastasis (P = 0.01, OR = 1.018, 95% CI = 1.004–1.031). The mean Glut1 H-scores of tumors with and without paraaortic lymph node metastasis were 163.00 ± 55.19 and 121.57 ± 55.62, respectively.
Predictive ability for treatment outcomes
In total, 76 patients were alive and 42 patients had died of cancer progression after a median follow-up of 50 months (range, 7–122). Seventy-two patients had no evidence of cancer progression. Thirteen of the 46 patients with tumor progression had pelvic recurrence, 19 had distant metastasis, and 14 had both. None of the 27 patients with pelvic recurrence experienced sole relapse in the lymph nodes. In summary, 27 patients had local residual or recurrent tumors at primary sites, whereas 33 patients experienced distant metastasis. The 4-year DFS, PRFS, and DMFS for SCC and AC patients were 60% and 57% (P = 0.97), 80% and 67% (P = 0.06), 67% and 70% (P = 0.92), respectively.
Table 3 lists the biomarkers and the AUC if the predictive value was greater than 0.6 or less than 0.4 for any endpoint. The H-scores of 4 IHC markers were associated with the presence of local residual or recurrent tumors, including c-Myc (AUC: 0.65, P = 0.018), Bax (AUC: 0.30, P = 0.002), E-cadherin (AUC: 0.64, P = 0.027), and calreticulin (AUC: 0.63, P = 0.047).
Table 3
Predictive immunohistochemical and clinical parameters and AUC (reported immunohistochemical markers have AUC values ≧ 0.6 or ≦ 0.4)
Variables | cancer progression AUC / P value | local failure AUC / P value | distant metastasis AUC / P value |
c-Myc H-score | 0.59 ± 0.05/0.095 | 0.65 ± 0.06/0.018* | 0.51 ± 0.06/0.85 |
Bax H-score | 0.43 ± 0.06/0.19 | 0.30 ± 0.05/0.002* | 0.52 ± 0.06/0.80 |
TNF-α H-score Calretinin H-score E-cadherin H-score galectin-9 H-score | 0.42 ± 0.05/0.15 0.58 ± 0.06/0.16 0.60 ± 0.05/0.078 0.43 ± 0.06/0.17 | 0.39 ± 0.06/0.09 0.63 ± 0.06/0.047* 0.64 ± 0.07/0.027* 0.39 ± 0.07/0.075 | 0.51 ± 0.06/0.94 0.51 ± 0.06/0.86 0.55 ± 0.06/0.45 0.41 ± 0.06/0.13 |
PD-L1 combined positive score | 0.54 ± 0.06/0.47 | 0.60 ± 0.06/0.10 | 0.39 ± 0.06/0.076 |
Maximum tumor dimension | 0.63 ± 0.05/0.018* | 0.64 ± 0.06/0.036* | 0.61 ± 0.06/0.07 |
Pretrement hemoglobulin | 0.44 ± 0.06/0.29 | 0.47 ± 0.07/0.59 | 0.51 ± 0.06/0.90 |
SUVmax of primary tumor | 0.49 ± 0.05/0.80 | 0.42 ± 0.06/0.23 | 0.60 ± 0.06/0.10 |
Abbreviation: AUC = area under the receiver operating characteristic curve; *SUVmax = maximum standardized uptake value. |
ROC analysis disclosed that none of the other IHC biomarkers, including the hypoxia, cell adhesion, or immunogenicity biomarkers, appeared to be prognostic for distant metastasis or cancer progression. Logistic regression analysis revealed that the existence of pelvic lymph node disease was the sole factor that predicted distant metastasis (P = 0.016, OR = 2.91, 95% CI = 1.22–6.93), and the maximum tumor diameter was the only paramater that predicted cancer progression (P = 0.037, OR = 1.49, 95% CI = 1.03–2.17).
Prognostic factors for DFS, PRFS, and DMFS
To test the prognostic values of the IHC markers, tumors were dichotomized using the median cut-offs of the 4 predictable IHC markers mentioned above. By combining with the clinical parameters, Cox regression analysis was performed. As summarized in Table 4, the results indicated that an E-cadherin H- score > 50% percentile [P = 0.006, hazard ratio (HR) = 2.35, 95% CI = 1.27–4.35] and stage III disease (P = 0.019, HR = 2.15, CI = 1.13–4.07) were two prognostic factors for an inferior DFS. The 4-year DFS of patients with low and high E-cadherin H- scores was 67% and 50% (P = 0.029).
Table 4
Multivariate analysis with Cox regression model for disease-free survival, pelvic relapse-free survival, and distant metastasis-free survival
| Disease-free survival | Pelvic relapse-free survival | Distant metastasis-free survival |
| Univariate model | Multivariate analysis | Univariate model | Multivariate analysis | Univariate model | Multivariate analysis |
Variables | P | P | HR | 95% CI | P | P | HR | 95% CI | P | P | HR | 95% CI |
Clinical variables | | | | | | | | | | | | |
AC vs. SCC | 0.91 | | | | 0.06 | 0.76 | | | 0.92 | | | |
FIGO stage III vs. IB3-IIB | 0.07 | 0.019* | 2.15 | 1.13–4.07 | 0.94 | | | | 0.012 | 0.50 | | |
Pelvic lymph node | | | | | 0.89 | | | | | | | |
positive vs.negative | 0.15 | | | | | | | | 0.007 | 0.01* | 2.66 | 1.26–5.59 |
Age (continuous) | 0.86 | | | | 0.08 | 0.10 | | | 0.80 | | | |
Maximum tumor dimension (continuous) | 0.63 | | | | 0.14 | | | | 0.33 | | | |
Pretreatment hemoglobulin (continuous) | 0.12 | | | | 0.16 | | | | 0.38 | | | |
IHCl biomarkers | | | | | | | | | | | | |
E-cadherin H-score (> 50% percentile vs ≦ 50% percentile) | 0.029 | 0.006* | 2.35 | 1.27–4.35 | 0.007 | 0.002* | 2.72 | 1.17–6.35 | 0.20 | | | |
c-Myc H-score (> 50% percentile vs ≦ 50% percentile) | 0.95 | | | | 0.032 | 0.07 | 2.50 | 0.93–6.77 | 0.61 | | | |
Bax H-score (> 50% percentile vs ≦ 50% percentile) | 0.50 | | | | 0.003 | 0.001* | 0.31 | 0.12–0.76 | 0.85 | | | |
calretinin H-score (> 50% percentile vs ≦ 50% percentile) | 0.09 | 0.10 | | | 0.37 | | | | 0.41 | | | |
Abbreviations: HR = hazard ratio; CI- confidence interval; FIGO = International Federation of Gynecology and Obstetrics; AC = adenocarcinoma; ACC = squamous cell carcinoma. |
Cox regression analysis disclosed that high E-cadherin and low Bax H-scores were the two predictors of poor PRFS (P = 0.002, HR = 2.72, CI = 1.17–6.35 and P = 0.001, HR = 0.31, CI = 0.12–0.76, respectively). As depicted in Fig. 2, the 4-year PRFS of patients with tumors with high and low E-cadherin values was 53% and 84% (P = 0.007), and the 4-year PRFS of patients with high and low expression of Bax was 88% and 63% (P = 0.003), respectively. In SCC patients, the impact of E-cadherin and Bax remained statistically significant (Supplemental Fig. 1).
None of IHC biomarkers were prognostic for DMFS. The major determinant for a low DMFS was pelvic lymph node disease (P = 0.01, HR = 2.66, CI = 1.26–5.59). In multivariate analysis, histology, age, maximum tumor size, and pretreatment hemoglobulin were not identified as independent prognostic factors for the aforementioned endpoints.