Predicting Response to Neoadjuvant Chemotherapy in Women with Locally Advanced Breast Cancer in Kenya: Utility of Ki67

Background: Ki67 levels have been shown to have good predictive value in breast cancer treatment. There is paucity of data on Ki67 levels in predicting response to neoadjuvant chemotherapy (NACT) in Kenya. This study evaluated the utility of Ki67 in predicting response to NACT. 
Methods: This was a prospective observational study carried out at Kenyatta National Hospital between December 2017 and January 2019 onpatients with locally advanced breast cancer. We recruited 61 women through consecutive sampling technique. Data collected included patient demographics, pre-treatment tumor size, Ki67 levels and tumor biology. After 3 cycles of first-line chemotherapy, ultrasonography was used to determine response. Data were analyzed by SPSS for proportion of change in tumor size. The response was correlated with tumor biology and pretreatment levels of Ki67 using chisquare at a 95% confidence interval. A p-value <0.05 was considered statistically significant. 
Results: The response rate after 3 cycles of NACT was 39.4%, sensitivity and specificity of Ki67 levels were 70.8% and 43.2% respectively with a cut-off value of 32.5%. 
Conclusions: Ki67 was found to predict response in our context at a rate of 39.4% at 20% cutoff after 3 cycles. 
Keywords: Ki67, Breast cancer, Neoadjuvant chemotherapy


Introduction
Breast cancer contributes a significant proportion of morbidity and mortality globally. Data from GLOBACAN show it is the 2nd in incidence and 5th cause of mortality globally (1). Data from Nairobi Cancer Registry reveal a high burden of breast cancer: it is the leading malignancy with a prevalence of 51.7 per 100,000 of all cancers in women in Kenya (2). Locally advanced breast cancer (LABC) is the commonest stage accounting for between 50.7% and 60% of breast cancer cases at first presentation (3). Currently in Kenyatta National Hospital, approximately 3-4 patients receive neoadjuvant chemotherapy weekly, or an average of 12-16 patients per month and 144-192 patients per year (unpublished data). Treatment of LABC involves a multidisciplinary approach with neoadjuvant chemotherapy (NACT), surgery, i.e. breast conserving surgery or modified radical mastectomy for operable patients, plus radiation therapy, depending on patient selection (4). January 2021 | Volume 18 | Issue 1 KI-67 IN LOCALLY ADVANCED BREAST CANCER LABC poses difficulties in achieving tumor-free resection margins and low chance of breast-conserving surgeries, results in higher recurrence rate and poses challenges in wound closure (5). NACT therefore enables the surgeon to achieve the acceptable resection margins, increases operability and increases the chances of wound closure (6). DNA microarrays have shown that some breast cancers are resistant to chemotherapy: this could result in some patients receiving unnecessary chemotherapy. DNA microarrays have been used to predict response to chemotherapy, but they are expensive and their utility is not universally adopted (7). The methods that have been used to monitor treatment are physical examination, imaging studies, pathologic response, NACT and biomarkers (8). Clinical methods have been shown to be inadequate on their own in predicting response to therapy (9). Ki67 is a nuclear antigen expressed in actively dividing cells. It is a marker of cellular proliferation (10). It is produced in large amounts in all cancerous tissues and therefore can be used to predict response to treatment (11). It is recommended to measure levels of Ki67 before initiating neoadjuvant therapy in breast cancer (12). Values of Ki67 levels of more than 25% have been shown to have a favorable response to NACT (13). There is however paucity of data on its utility in locally advanced breast cancer in Kenya, and the data available lack uniformity of Ki67 cutoff levels. Ki67 applied with good accuracy has been shown to have good prediction values (13). This study therefore aims at determining the utility of Ki67 in predicting response to neoadjuvant chemotherapy in LABC.

Methods
This was a prospective observational study carried out at Kenyatta National Hospital surgical wards, surgical outpatient clinic, oncology clinics, radiology department and histopathology laboratory. The study was conducted between December 2017 and January 2019. Inclusion criteria were women with locally advanced breast cancer, who had no systemic metastasis, T3 disease on ultrasonography and chemotherapy naïve patients. Staging was done by clinical examination and radiological assessment using CT scan chest or plain chest x-ray, and abdominal ultrasound. Only patients who were negative for systemic metastases and fully staged were included in the study. Ultrasonography was performed by qualified radiologist as per hospital policy. Patients were recruited through convenient sampling procedure after informed consent was administered. Data were collected using the data collection sheet. Data collected included patients' demographics, pretreatment Ki67 levels, pretreatment tumor size, tumor biology (tumor grade, lymphovascular invasion and immunohistochemistry: ER, PR and HER2). Tumor size was determined by ultrasound prior to therapy and at the end of 3 cycles. Patients were assessed using Ultrasound Machine Aplio 400 with a high frequency linear probe of 12 MHz to measure the longest tumour diameter. Standard neoadjuvant first line chemotherapy was given for 3 cycles. Every third week the regime given was Adriamycin 60 mg/m 2 and cyclophosphamide 600 mg/m 2 . Sample size was calculated using Daniel's formula with finite population correction for standard distribution with 95% confidence interval and a desired precision of 0.05. A repeat breast ultrasound was performed after the 3 cycles of chemotherapy, and sizes recorded. The evaluation of response to NACT was performed using the response evaluation criteria in solid tumours (RECIST). Data were entered into MS Excel sheet, cleaned and transferred to SPSS (version 21.0, Chicago-Illinois) for analysis. Analysis was performed for mean and range of age, mean tumor size change from pretreatment to after 3 cycles of NACT, proportions of patient with high or low Ki67 according to St Galen classification, which used 20% as the cutoff, and proportion of patients who had response as per RECIST. Chi-square was used to analyze the association between Ki67 level to response and other aspects of tumor biology: molecular subtype, tumor grade, and perineural and lymphovascular invasion. Receiver operator curve (ROC) was drawn using the Ki67 levels against response to NACT. This was then used to determine the sensitivity, specificity, cutoffs, and area under the curve of Ki67 for our population. p values and 95% confidence intervals (CIs) were calculated as applicable, and p value <0.05 was considered statistically significant. This study commenced after approval from the Department of Surgery and the UoN-KNH ERC, under approval number P247/05/2017.

Results
Sixty-one patients were recruited for this study. Their mean age was 45.9 (SD=10.4) years while the minimum age was 28 years and maximum 73 years; 27 (44.3%) patients were 41-50 years of age. All patients in this study had invasive ductal carcinoma and most (63.9%) had tumor grade II and 34.4% Luminal A molecular subtype (Table 1).   Table 2 show the correlation between Ki67 levels and response to NACT based on sensitivity and specificity, cutoff, and area under the curve for Ki67. Point marked pink on the graph is the cutoff point, giving sensitivity and specificity for prediction of response. The paired sample mean change in tumor size on ultrasonography from pretreatment to posttreatment was 1.24 (SD=2.19) with confidence interval of 0.68-1.80, t(60)=4.420, p<0.001. A dependent samples t-test was performed to test the hypothesis that the size of the breast mass (cm) by ultrasonography at week 0 (M=5.95, SD=4.34) and the size of the breast mass (cm) greatest dimensions following 3 cycles of NACT (M=4.71, SD=4.07) were equal. The correlation between the two conditions was estimated at r=0.866, p<0.001, suggesting that the dependent samples t-test is appropriate in this case. The null hypothesis of equal breast mass means was rejected t(60)=4.42, p<0.001. Thus, the mean of size of breast mass (cm) greatest dimensions at the end of 3 cycles of NACT was statistically significantly lower than the mean of size of breast mass (cm) by ultrasonography at week 0.

Correlation of Ki67 levels with tumor biological characteristics
The pathological response after 3 cycles of chemotherapy was seen in 24 (39.4%) patients, most 31 (50.8%) remained stable while only 6 (9.8%) had progression of their disease while on chemotherapy. When considered against the cutoff agreed at Galen for Ki67, there was no significant difference between those above or below 20%. A chi-square test for proportion was conducted between the Ki67 levels and response. There were no statistically significant differences in the proportions between Ki67 levels and responders and non-responders, χ2(1)=0.884, p=.347 (Table 3).

Response rates in relation to receptor status
HER2+ had a higher proportion of responders at 37.0%, followed by ER+ at 31.4% and PR+ at 20.7%. Results are shown in Table 4.

Relationship between Ki67 and perineural and lymphovascular invasion
Perineural invasion after 3 cycles of chemotherapy was seen in 24 patients (39.4%), most 31 (50.8%) remained stable while only 6 (9.8%) had progression of disease while on chemotherapy (Table 5). When considered against the cutoff agreed at Galen for Ki67, difference between those above or below 20% was not significant. A chi-square test for association was conducted between the lymphovascular invasion and perineural invasion with the Ki67 levels. There was no statistically significant association between lymphovascular invasion and Ki67 levels, χ2(1)=2.198, p=.138, but association was statistically significant between perineural invasion and higher Ki67 levels, χ2(1)=10.509, p =0.005. The study also elucidated that high Ki67 >30% shows a lower HER2-neu expression and lymph node metastasis, hence may have worse prognosis (24). Inter-observer variability during ultrasonography may introduce limitations in the study, but this was delimited by use of standard machine and use of two radiologists in reading the ultrasonographic report as part of quality assurance.

Conclusion
Ki67 was found to predict response in our context with a response rate of 39.4% using the St. Galen's cutoff of 20%. Using ROC, sensitivity was found to be 70.8% and specificity 43.2% at cutoff of 32.5%.