Standardised Ki‐67 proliferation index assessment in early‐stage laryngeal squamous cell carcinoma in relation to local control and survival after primary radiotherapy

Abstract Objectives Ambiguous results have been reported on the predictive value of the Ki‐67 proliferation index (Ki‐67 PI) regarding local control (LC) and survival after primary radiotherapy (RT) in early‐stage laryngeal squamous cell cancer (LSCC). Small study size, heterogenic inclusion, variations in immunostaining and cut‐off values are attributing factors. Our aim was to elucidate the predictive value of the Ki‐67 PI for LC and disease‐specific survival (DSS) using a well‐defined series of T1‐T2 LSCC, standardised automatic immunostaining and digital image analysis (DIA). Methods A consecutive and well‐defined cohort of 208 patients with T1‐T2 LSCC treated with primary RT was selected. The Ki‐67 PI was determined using DIA. Mann‐Whitney U‐tests, logistic and Cox regression analyses were performed to assess associations between Ki‐67 PI, clinicopathological variables, LC and DSS. Results In multivariate Cox regression analysis, poor tumour differentiation (HR 2.20; 95% CI 1.06‐4.59, P = .04) and alcohol use (HR 2.84, 95% CI 1.20‐6.71; P = .02) were independent predictors for LC. Lymph node positivity was an independent predictor for DSS (HR 3.16, 95% CI 1.16‐8.64; P = .03). Ki‐67 PI was not associated with LC (HR 1.59; 95% CI 0.89‐2.81; P = .11) or DSS (HR 0.98; 95% CI 0.57‐1.66; P = .97). In addition, continuous Ki‐67 PI was not associated with LC (HR 2.03; 95% CI 0.37‐11.14, P = .42) or DSS (HR 0.62; 95% CI 0.05‐8.28; P = .72). Conclusion The Ki‐67 PI was not found to be a predictor for LC or DSS and therefore should not be incorporated in treatment‐related decision‐making for LSCC.

as the Ki-67 proliferation index (Ki-67 PI). However, the results of earlier studies investigating the relationship between the Ki-67 PI, local control (LC) and survival after primary RT in laryngeal squamous cell cancer (LSCC) are not unambiguous, as shown in Table 1. [4][5][6][7][8][9][10][11][12][13][14] Possible explanations for these differences are variations in patient group factors, immunostaining and scoring-related factors. [15][16][17] The aim of this study was to assess the value of Ki-67 PI in predicting LC and disease-specific survival (DSS) after primary RT in a well-defined consecutive series of patients with early-stage (T1-T2) LSCC. By using standardised and automated immunohistochemistry along with digital image analysis (DIA) to assess the Ki-67 PI, we reduced staining and scoring variability.

| Patients
Patients treated for LSCC at our institution are included in a database by the Netherlands Cancer Registry (NCR) by using the results of the nationwide network and registry of histo-and cytopathology in the Netherlands (PALGA). Retrospectively, data from the hospital con-

| Ethical considerations
According to the Central Committee on Research involving Human Subjects (CCMO), this type of study did not require approval from an ethics committee in the Netherlands. This study was approved by the Privacy Review Board of the NCR by following "The Code of Conduct for the Use of Data in Health Research" of the CCMO. 23

| Treatment
All patients were treated by a multidisciplinary head & neck team.
Patients received primary RT with curative intent using 6MV linear accelerator equipment as previously described. [20][21][22] In short, T1 tumours received 2 Gy fractions five times weekly with a total dose of 66 Gy. T2 tumours were treated with six fractions weekly to a total dose of 70 Gy.
In case of lymph node metastasis, a total dose of 46 Gy was electively delivered to the primary planning target volume together with an additional boost of 70 Gy to the primary tumour and pathologic lymph nodes. From the year 2000 onwards, planning of field arrangements was performed by using contrast-enhanced computed tomography (CT). Before 2000, this was calculated by direct simulation (Figure 1).

| Follow-up
All patients had standardised follow-up after completing RT in accordance with the Dutch Working Party on Head and Neck Tumours (NWHHT) guidelines. 24 For the first 2 years, the otorhinolaryngology and radiotherapy department alternately performed physical examination with laryngoscopy every 3 months. After 2 years, this was alternately performed every 6 months up till 5 years after completing radiation treatment. Patients were discharged from followup after 5 years if no evidence of disease was found.

| Immunohistochemistry
All tumour material was formalin-fixed and paraffin-embedded.  • Small study size, heterogeneous inclusion, variations in immunostaining and cut-off values are factors attributing to these contradictory results.
• We used a well-defined series of T1-T2 laryngeal tumours treated with radiotherapy, standardised automatic immunostaining and automatic digital scoring.
• Standardised and automated staining minimises variable staining intensity and improves reproducibility.
Automated digital scoring eliminates interobserver variability.
• The Ki-67 proliferation index was not a predictor for local control or disease-specific survival and therefore should not be incorporated in treatment-related decision-making for early-stage laryngeal squamous cell cancer.

| Evaluation of immunohistochemical staining
A whole tumour slide was analysed in order to reduce sampling error. All glass slides digitized using the Hamamatsu Nanozoomer

| Definitions
LC was defined as local tumour recurrence at the primary tumour site within 2 years after RT and was calculated from the date of diagnosis until the date of recurrence. After this period of 2 years, any local recurrence was defined as a second primary tumour. DSS was defined as the date of diagnosis until the date of death by disease or last date of follow-up within 5 years.
In the analyses, Ki-67 PI was considered both as a continuous and a dichotomous variable. For dichotimisation, the cut-off value for high vs low Ki-67 PI was set to 50%, which was defined by the median Ki-67 expression in our cohort. In addition, we also tried to compare our data with previously published studies, which used cutoff values of 10% and 20%. [7][8][9]12,13 Alcohol use was defined as drinking one or more units per day either in the past or at date of diagnosis. The same was applied for tobacco use with smoking one or more cigarettes or sigars per day.

| Statistical analysis
Patients were dichotomised based on their Ki-67 PI, and for correlations between patient and tumour characteristics, univariate logistic TA B L E 1 Patients and disease characteristics related to local recurrence after radiotherapy

| RE SULTS
Patient and tumour characteristics are described in Table 2. The majority of tumours were of glottic origin, had a T2 status, did not have lymph node metastasis and were moderately differentiated. Most patients were male and median age was 64.4 years.  we also used a 50% cut-off between high and low Ki-67 PI for comparability. Using a 10% and 20% cut-off, a high Ki-67 PI was found in 207 cases (99.5%) and 199 cases (95.7%), respectively. As the low Ki-67 PI group was too small for both cut-offs (one and nine cases respectively), no further statistical analyses were performed using these cut-offs.
In the univariate regression analysis using a 50% cut-off for the Ki-67 PI, no significant associations between clinicopathological variables and Ki-67 PI were found (Table 3). When treated as a continuous variable, no significant associations between Ki-67 PI and the evaluated variables were found (Table 3). were found ( Figure S1A, Table 4).
In univariate and stepwise multivariate Cox regression analysis, a significant negative association was found between lymph node positivity and DSS (HR 3. 16 Figure S1B, Table 5). From the eleven previously conducted studies, 15 (sub)analyses were reported or could be calculated using the data and cut-off values provided in the papers (Table 1). Of those, nine did not find a significant association between Ki-67 PI and LC after RT. [6][7][8][9][10][11]13,14 Two subgroup analyses in one study showed a negative association between high Ki-67 and LC in both a cohort treated with accelerated RT (ART) and in a combined cohort treated with either ART or conventional RT (HR 2.66; 95% CI 1.17-6.08 and HR 5.11;

| D ISCUSS I ON
95% CI 1.53-17.06 respectively). 5 Nichols et al found a worse local, regional or distant control in patients with high Ki-67 tumours. 12 Three studies showed a significant positive association between high Ki-67 and LC after RT using continuous values, and one study showed a positive association using a 50% cut-off (no HR or 95% CI was given or could be calculated). 4,9,10  From five subanalyses of the four studies that assessed the association between Ki-67 and survival, none found a difference in OS, 5,8 DSS 5,13 or survival (not otherwise specified). 10 In one paper, worse regional control and metastasis-free survival were reported. 13 26 A moderate reproducibility across the laboratories was found when they used their own scoring methodology on sections stained in a central laboratory. This reproducibility declined even further when both staining and scoring were done locally.
By using a standardised and automated staining platform with a pre-diluted antibody by the supplier, we minimised this problem and improved reproducibility, enabling future interlaboratory comparison.