Scandinavian Journal of Gastroenterology. 2014; 49: 84–91
ORIGINAL ARTICLE
Scand J Gastroenterol Downloaded from informahealthcare.com by Nyu Medical Center on 06/02/15 For personal use only.
Assessment of the topoisomerase I gene copy number as a predictive biomarker of objective response to irinotecan in metastatic colorectal cancer
SUNE BORIS NYGÅRD1, IB JARLE CHRISTENSEN2, SIGNE LYKKE NIELSEN1, HANS JØRGEN NIELSEN3,4, NILS BRÜNNER1 & KAREN-LISE GARM SPINDLER5* 1
Section for Molecular Disease Biology, Department of Veterinary Disease Biology, Faculty of Health and Medical Sciences, University of Copenhagen, 49 Strandboulevarden, DK-2100 Copenhagen Ø, Denmark, 2Finsen Laboratory, Rigshospitalet and Biotech Research and Innovation Centre (BRIC), University of Copenhagen, Copenhagen Biocenter, 5 Ole Maaløes Vej, building 3, 3rd floor, DK-2200 Copenhagen N, Denmark, 3Department of Surgical Gastroenterology 360, Hvidovre Hospital, Hvidovre, 30 Kettegård Allé, DK-2650 Hvidovre, Denmark, 4Institute of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, 3B Blegdamsvej, DK-2200 Copenhagen N, Denmark, and 5Department of Oncology, Vejle Sygehus, 25 Kabbeltoft, DK-7100 Vejle, Denmark
Abstract Objective. DNA topoisomerase I is a putative biomarker of irinotecan efficacy with clinical associations previously demonstrated at the protein level. The purpose of the present study was to perform the first clinical investigation of the association between the DNA topoisomerase I gene (TOP1) copy number and objective response following irinotecan treatment in patients with metastatic colorectal cancer. Materials and methods. Formalin-fixed, paraffin-embedded tumor samples from 78 patients, who received irinotecan monotherapy in second line, were included. TOP1 was assessed by fluorescence in situ hybridization using a technically validated dual-probe combination that hybridizes to TOP1, located at 20q12-q13.1, and to the centromere region of chromosome 20 (CEN-20). In univariate logistic regression models, the TOP1 signal count per cell and the TOP1/CEN-20 ratio were associated with objective response, which was evaluated according to RECIST v.1.1. Results. Gain of TOP1 was identified in 52.6% and 37.2% using the following cutoff values: TOP1 signal count per cell ‡3.6 and TOP1/CEN-20 ‡1.5, respectively. A borderline significant association (Odds ratio (OR): 1.62; p = 0.07) between a stepwise increase in the TOP1 signal count and objective response was demonstrated. In relation to the applied cutoff values, nonsignificant associations with objective response were identified for the TOP1 signal count (OR: 2.41; p = 0.23) and for the TOP1/CEN-20 ratio (OR: 2.05; p = 0.30). Conclusions. Despite limitations of the study the positive associations between TOP1 and objective response suggest that further analysis in larger tumor material, preferably in a randomized setting, is highly warranted.
Key Words: biomarker, colorectal cancer, DNA topoisomerase I, irinotecan
Introduction The age-standardized incidence rate of colorectal cancer (CRC) has increased in the past decades [1] and despite improved surgical procedures and the use of chemotherapy in the adjuvant and metastatic
setting, more than 200,000 persons die of this disease each year in Europe alone. This makes CRC the second leading cause of death from cancer in Europe [2] and the fourth leading cause worldwide [3]. Although some patients with metastatic CRC (mCRC) are treated with curative intent, the majority
Correspondence: Sune Boris Nygård, MD, Section for Molecular Disease Biology, Department of Veterinary Disease Biology, Faculty of Health and Medical Sciences, University of Copenhagen, 49 Strandboulevarden, DK-2100 Copenhagen Ø, Denmark. Tel: +45 35336766. E-mail: [email protected] * Department of Oncology, Aarhus University Hospital, 44 Nørrebrogade, DK-8000 Aarhus C, Denmark.
(Received 17 September 2013; accepted 13 October 2013) ISSN 0036-5521 print/ISSN 1502-7708 online 2014 Informa Healthcare DOI: 10.3109/00365521.2013.856464
Scand J Gastroenterol Downloaded from informahealthcare.com by Nyu Medical Center on 06/02/15 For personal use only.
TOP1 and objective response to irinotecan of patients are treated with the intention to prolong progression-free survival (PFS) and improve quality of life. In population-based studies the median relative survival of patients with non-resectable mCRC has increased following the introduction of chemotherapy; however, it remains below 10 months [4]. Fluoropyramidines (5-fluorouracil (5FU) and related pro-drugs) have formed the backbone of standard therapy and results have improved by combinations with the cytotoxic agents, oxaliplatin or irinotecan. These combinations have been demonstrated to be equally effective in mCRC [5,6] and are both recommended as first-line treatment [7,8]. Not all patients respond to the chosen first-line treatment combination, but all patients are subject to potential drugrelated adverse effects. This stresses the importance of selecting the most effective option for the individual patient, and predictive biomarkers are therefore highly needed. Potential markers either predictive of 5FU, oxaliplatin, or irinotecan have been suggested [9]; however, none have reached a level of evidence supporting clinical recommendation. The active form of irinotecan exerts its cytotoxicity by forming stable complexes with DNA-bound topoisomerase I (Top1), which ultimately leads to DNA damage and cell death [10,11]. Top1 remains the most promising predictive biomarker of irinotecan efficacy, and positive associations between Top1 status and efficacy of irinotecan have been demonstrated at the gene level [12,13] and protein level [14–19]. However, conflicting results have been published [20,21]. We have previously developed and technically validated a Top1 gene (TOP1) fluorescence in situ hybridization (FISH) probe to be used on formalinfixed paraffin-embedded (FFPE) tissue and demonstrated that TOP1 copy number alterations are present in high numbers in CRC [13,22]. The aim of the present study was to investigate the predictive value of TOP1 assessed by FISH analysis in tumor material from patients with mCRC, who were treated with irinotecan monotherapy in second line, and we hypothesized that gain of TOP1 was associated with improved treatment response. Materials and methods Patients and tumor material Ninety consecutive patients from a single-center clinical trial, which was initiated to investigate different potential predictive biomarkers of irinotecan monotherapy for mCRC, were assessed for eligibility. The study was approved by the local Research Ethics Committee of Southern Denmark (S-20090114). Original inclusion criteria were histologically verified
85
mCRC, indication for treatment with irinotecan monotherapy according to local guidelines, measurable disease according to RECIST v. 1.1 [23], informed consent, and a minimum age of 18 years. Patients with synchronous cancers, except for basal cell carcinoma of the skin, experimental treatment within 30 days of inclusion, or planned radiotherapy for target lesions, were excluded. Between February 2010 and March 2012 patients were treated with intravenous irinotecan monotherapy 350 mg/m2 q3w and supportive care according to local guidelines. Response evaluations with computerized tomography scans of the chest and abdomen were performed every 9 weeks according to RECIST v. 1.1. Follow-up was ended by September 2012. The primary endpoint of the study was objective response and the secondary endpoints were PFS and overall survival (OS). PFS was defined as time in months from the date of inclusion until the occurrence of disease progression or death from any cause, and OS was defined as time in months from the date of inclusion until death from any cause. An additional inclusion criterion for the present TOP1 study was the availability of histological material. Archival diagnostic FFPE tumor material from either the primary tumor or from a metastatic lesion was collected. The tumor material was originally obtained by core needle biopsies or from the surgical resection specimens. The tissue samples were fixed in neutral buffered formalin and processed and paraffinembedded according to standard protocols of the local department of pathology. The FFPE tumor blocks were cut in 3-mm sections and were stored at 5 C until hybridization. FISH procedures A TOP1/CEN-20 probe mix developed and validated by Dako Denmark (Glostrup, Denmark) and the University of Copenhagen (Copenhagen, Denmark) was used for detection of the TOP1 gene (chromosome 20q12-q13.1) and the reference centromere region on chromosome 20 (CEN-20), and the FISH procedure was performed using the Dako Histology FISH Accessory Kit (Dako Denmark) as previously reported [13]. A fluorescence microscope (Zeiss AX10) with a Texas Red/FITC double filter was used for visualization of the probe signals. TOP1 and CEN-20 signals were enumerated in 60 nonoverlapping cancer nuclei with well-defined nuclear morphology and distinct fluorescent signals of high quality. Overlapping signals of the same color were counted as one signal. The TOP1 and CEN-20 signal count per cell as well as the TOP1/CEN-20 ratio were calculated for each tumor sample. New sections were
86
S. B. Nygård et al.
Scand J Gastroenterol Downloaded from informahealthcare.com by Nyu Medical Center on 06/02/15 For personal use only.
cut and FISH analysis repeated if the fluorescent intensity of the probe signals was weak or if tumor tissue was not present in the original sections. Cases were excluded if signal counting failed following the second hybridization attempt. All procedures were performed blinded to all patient data, and the clinical database was only made accessible for statistical analyses when FISH data were obtained. The samples were evaluated by a single experienced technician (SLN), who was well trained in the FISH technique and the pathological appearance of CRC. Statistics The chi-square test and the Wilcoxon rank sum test were used to test for associations between TOP1 status and baseline characteristics. Logistic regression models were used to test TOP1 and TOP1/ CEN-20 data in relation to best objective response, which was categorized in response (complete and partial response) and non-response (stable disease and disease progression). Odds ratio (OR) estimates and Wald chi-square p-values with estimated 95% confidence intervals (CI) were reported. Univariate Cox proportional hazards regression models were used to test for associations between the variables and the survival endpoints, PFS and OS. The considered variables were the TOP1 signal count, the TOP1/CEN-20 ratio, gender, age, primary tumor site, surgical removal of the primary tumor, and the presence of liver and lung metastases. Hazard ratio (HR) estimates and Wald chi-square p-values with estimated 95% CI were reported for all variables. The HR for age was reported for 10-year differences. Multivariate survival analysis was only considered if statistical significance was demonstrated for the TOP1 signal count or the TOP1/CEN-20 ratio in univariate analysis. Results with a two-sided p-value 4 TOP1 copies per cell when correcting for nuclear truncation due to tissue sectioning [22]. In brief, reference ranges for the TOP1 copy number per cell were calculated based on TOP1 signal counts enumerated in enterocytes of the lamina epithelialis in non-neoplastic colon and rectal mucosa. The normal diploid range was defined as (2n minus 1/2n – 2n plus 1/2n) where 2n equals the mean diploid TOP1 count per cell in normal mucosa. To define ranges for gain a multiplex of n was used. A TOP1/CEN-20 ratio ‡1.5 was also used to separate tumors with TOP1 gain from tumors with normal
TOP1 status. As a result of the explorative nature of the study the TOP1 signal count per cell and the TOP1/CEN-20 ratio were also considered as continuous variables in the logistic regression models and only as such in the Cox proportional hazards regression models. The present study was primarily aimed at investigating a possible trend between TOP1 copy number alterations and objective response why a sample size of 90 patients, most of whom had experienced disease progression was considered large enough for analysis. No imputations of missing values were considered in the statistical models and only tumors with a record of the variables in question entered analysis. The Cox proportional hazards assumptions were assessed by Schoenfeld and martingale residuals, and proportional hazards and linearity were applicable. For the logistic regression models goodness-of-fit was assessed by the Hosmer and Lemeshow test. All statistical analyses were performed using SAS (v9.2, SAS Institute, Cary, NC, USA). Results Tumor samples were available from 81 (90.0%) of the 90 patients assessed for study eligibility. In 78 (96.3 %) samples the signals could be quantified following FISH procedures. After a second hybridization attempt the remaining three (3.7%) samples were still unsuitable for counting due to weak fluorescent signals in one sample and no tumor tissue present in two samples. The majority of samples (n = 74; 94.9%) originated from primary tumors, whereas four (5.1%) were from metastatic lesions. A study profile (adapted from the updated REMARK guidelines [24]) including information on the patient flow through the study, the primary markers and other variables considered, and a statistical plan with number of events and distribution of outcomes is presented in Table I. The median TOP1 and CEN-20 signal count per cell were 3.6 (range: 1.8–6.8) and 2.5 (range: 1.5– 3.9), respectively, and the median TOP1/ CEN-20 ratio was 1.4 (range: 1.0–2.8). Based on the ranges for TOP1 copy number changes, only 5 (6.4%) tumors had a diploid TOP1 copy number, whereas the remaining 73 (93.6%) had copy numbers suggesting some level of gene gain (Table II). No cases with TOP1 gene loss were identified. When using the predetermined cutoff values for the TOP1 signal count per cell and the TOP1/CEN-20 ratio, the frequency of tumors with gain of TOP1 was 41 (52.6%) and 29 (37.2%), respectively. In relation to baseline characteristics the TOP1 signal count per cell significantly associated with gender (p = 0.046) with TOP1 gain more frequently observed in tumors from males. No
TOP1 and objective response to irinotecan
87
Table I. Study profile. No. of patients Patients Assessed for eligibility 90
Scand J Gastroenterol Downloaded from informahealthcare.com by Nyu Medical Center on 06/02/15 For personal use only.
Excluded 9 Failed FISH analysis 3 FISH results obtained 78 Available outcome data
Remarks
Patients with metastatic CRC included between February 2010 and March 2012; irinotecan monotherapy; Department of Oncology, Vejle Sygehus, Denmark Histological material not available No tumor tissue in FFPE sections, n = 2; weak fluorescent signals, n = 1 Objective response data, n = 71 PFS data, n = 77; OS data, n = 77
Markers and variables Markers M1 = TOP1 signal count per cell; M2 = TOP1/CEN-20 ratio Additional variables v1 = gender, v2 = age; v3 = primary tumor site; v4 = primary tumor surgery; v6 = lung metastases Statistical analysis of objective response Variables considered Handling of Response Non-response variables CR PR SD PD M1-M2 Categorical; 0 10 26 35 continuous Statistical analysis of survival outcome Variables considered Handling of PFS OS variables Patients Events Patients Events M1-M2; v2 Continuous 77 76 77 59 v1; v3-v6 Categorical 77 76 77 59
v5 = liver metastases;
Remarks Logistic regression analysis
Remarks Univariate Cox proportional hazards regression analysis
Abbreviations: CR = complete response; CRC = colorectal cancer; OS = overall survival; PD = disease progression; PFS = progression-free survival; PR = partial response; SD = stable disease.
other significant associations with baseline characteristics were demonstrated (Table III). The OR estimates for a stepwise increase of the TOP1 signal count per cell and the TOP1/ CEN-20 ratio in relation to best objective response were 1.62 (0.96–2.75; p = 0.07) and 2.68 (0.43– 16.65; p = 0.29), respectively. Based on the cutoff values for the TOP1 signal count and the TOP1/ CEN-20 ratio, the OR estimates for TOP1 gain versus TOP1 normal were 2.41 (0.57–10.20; p = 0.23) and 2.05 (0.53–7.91; p = 0.30), respectively. No significant associations of the TOP1 signal count or the TOP1/CEN-20 ratio with the survival endpoints (PFS and OS) were demonstrated. In univariate Cox proportional hazards models the HR estimates for continuously increasing values of the TOP1 count in relation to PFS and OS were 1.00 (0.83–1.12; p = 1.00) and 1.01 (0.81–1.26; p = 0.95), respectively.
Likewise, the HR estimates for continuously increasing values of the TOP1/CEN-20 ratio in relation to PFS and OS were 1.08 (0.55–2.12; p = 0.83) and 1.01 (0.45–2.28; p = 0.97), respectively. No other tested covariates were significantly associated with survival (Table IV). Discussion The majority of the 78 evaluable tumors harbored alterations involving the TOP1 locus suggesting some level of gene gain. A signal count compatible with at least two extra TOP1 gene copies per cell was identified in 52.6%, which was more frequent than expected from previous work where 35.7% stage III primary CRC tumors showed the same alteration status using the same methodology [22]. TOP1 is located at chromosome 20q12– 13.1 and gain of the 20q arm or parts of it has previously
Table II. Sample distribution according to TOP1 gene status. Signals per truncated nuclei TOP1 gene status per cell Loss of gene copy Normal/diploid 1 extra gene copy 2 extra gene copies >2 extra gene copies
Copies per whole nucleus
Average
1 2 3 4 >4
0.8 1.6 2.4 3.2
Frequency
Range
No.
%
ORIGINAL ARTICLE
Scand J Gastroenterol Downloaded from informahealthcare.com by Nyu Medical Center on 06/02/15 For personal use only.
Assessment of the topoisomerase I gene copy number as a predictive biomarker of objective response to irinotecan in metastatic colorectal cancer
SUNE BORIS NYGÅRD1, IB JARLE CHRISTENSEN2, SIGNE LYKKE NIELSEN1, HANS JØRGEN NIELSEN3,4, NILS BRÜNNER1 & KAREN-LISE GARM SPINDLER5* 1
Section for Molecular Disease Biology, Department of Veterinary Disease Biology, Faculty of Health and Medical Sciences, University of Copenhagen, 49 Strandboulevarden, DK-2100 Copenhagen Ø, Denmark, 2Finsen Laboratory, Rigshospitalet and Biotech Research and Innovation Centre (BRIC), University of Copenhagen, Copenhagen Biocenter, 5 Ole Maaløes Vej, building 3, 3rd floor, DK-2200 Copenhagen N, Denmark, 3Department of Surgical Gastroenterology 360, Hvidovre Hospital, Hvidovre, 30 Kettegård Allé, DK-2650 Hvidovre, Denmark, 4Institute of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, 3B Blegdamsvej, DK-2200 Copenhagen N, Denmark, and 5Department of Oncology, Vejle Sygehus, 25 Kabbeltoft, DK-7100 Vejle, Denmark
Abstract Objective. DNA topoisomerase I is a putative biomarker of irinotecan efficacy with clinical associations previously demonstrated at the protein level. The purpose of the present study was to perform the first clinical investigation of the association between the DNA topoisomerase I gene (TOP1) copy number and objective response following irinotecan treatment in patients with metastatic colorectal cancer. Materials and methods. Formalin-fixed, paraffin-embedded tumor samples from 78 patients, who received irinotecan monotherapy in second line, were included. TOP1 was assessed by fluorescence in situ hybridization using a technically validated dual-probe combination that hybridizes to TOP1, located at 20q12-q13.1, and to the centromere region of chromosome 20 (CEN-20). In univariate logistic regression models, the TOP1 signal count per cell and the TOP1/CEN-20 ratio were associated with objective response, which was evaluated according to RECIST v.1.1. Results. Gain of TOP1 was identified in 52.6% and 37.2% using the following cutoff values: TOP1 signal count per cell ‡3.6 and TOP1/CEN-20 ‡1.5, respectively. A borderline significant association (Odds ratio (OR): 1.62; p = 0.07) between a stepwise increase in the TOP1 signal count and objective response was demonstrated. In relation to the applied cutoff values, nonsignificant associations with objective response were identified for the TOP1 signal count (OR: 2.41; p = 0.23) and for the TOP1/CEN-20 ratio (OR: 2.05; p = 0.30). Conclusions. Despite limitations of the study the positive associations between TOP1 and objective response suggest that further analysis in larger tumor material, preferably in a randomized setting, is highly warranted.
Key Words: biomarker, colorectal cancer, DNA topoisomerase I, irinotecan
Introduction The age-standardized incidence rate of colorectal cancer (CRC) has increased in the past decades [1] and despite improved surgical procedures and the use of chemotherapy in the adjuvant and metastatic
setting, more than 200,000 persons die of this disease each year in Europe alone. This makes CRC the second leading cause of death from cancer in Europe [2] and the fourth leading cause worldwide [3]. Although some patients with metastatic CRC (mCRC) are treated with curative intent, the majority
Correspondence: Sune Boris Nygård, MD, Section for Molecular Disease Biology, Department of Veterinary Disease Biology, Faculty of Health and Medical Sciences, University of Copenhagen, 49 Strandboulevarden, DK-2100 Copenhagen Ø, Denmark. Tel: +45 35336766. E-mail: [email protected] * Department of Oncology, Aarhus University Hospital, 44 Nørrebrogade, DK-8000 Aarhus C, Denmark.
(Received 17 September 2013; accepted 13 October 2013) ISSN 0036-5521 print/ISSN 1502-7708 online 2014 Informa Healthcare DOI: 10.3109/00365521.2013.856464
Scand J Gastroenterol Downloaded from informahealthcare.com by Nyu Medical Center on 06/02/15 For personal use only.
TOP1 and objective response to irinotecan of patients are treated with the intention to prolong progression-free survival (PFS) and improve quality of life. In population-based studies the median relative survival of patients with non-resectable mCRC has increased following the introduction of chemotherapy; however, it remains below 10 months [4]. Fluoropyramidines (5-fluorouracil (5FU) and related pro-drugs) have formed the backbone of standard therapy and results have improved by combinations with the cytotoxic agents, oxaliplatin or irinotecan. These combinations have been demonstrated to be equally effective in mCRC [5,6] and are both recommended as first-line treatment [7,8]. Not all patients respond to the chosen first-line treatment combination, but all patients are subject to potential drugrelated adverse effects. This stresses the importance of selecting the most effective option for the individual patient, and predictive biomarkers are therefore highly needed. Potential markers either predictive of 5FU, oxaliplatin, or irinotecan have been suggested [9]; however, none have reached a level of evidence supporting clinical recommendation. The active form of irinotecan exerts its cytotoxicity by forming stable complexes with DNA-bound topoisomerase I (Top1), which ultimately leads to DNA damage and cell death [10,11]. Top1 remains the most promising predictive biomarker of irinotecan efficacy, and positive associations between Top1 status and efficacy of irinotecan have been demonstrated at the gene level [12,13] and protein level [14–19]. However, conflicting results have been published [20,21]. We have previously developed and technically validated a Top1 gene (TOP1) fluorescence in situ hybridization (FISH) probe to be used on formalinfixed paraffin-embedded (FFPE) tissue and demonstrated that TOP1 copy number alterations are present in high numbers in CRC [13,22]. The aim of the present study was to investigate the predictive value of TOP1 assessed by FISH analysis in tumor material from patients with mCRC, who were treated with irinotecan monotherapy in second line, and we hypothesized that gain of TOP1 was associated with improved treatment response. Materials and methods Patients and tumor material Ninety consecutive patients from a single-center clinical trial, which was initiated to investigate different potential predictive biomarkers of irinotecan monotherapy for mCRC, were assessed for eligibility. The study was approved by the local Research Ethics Committee of Southern Denmark (S-20090114). Original inclusion criteria were histologically verified
85
mCRC, indication for treatment with irinotecan monotherapy according to local guidelines, measurable disease according to RECIST v. 1.1 [23], informed consent, and a minimum age of 18 years. Patients with synchronous cancers, except for basal cell carcinoma of the skin, experimental treatment within 30 days of inclusion, or planned radiotherapy for target lesions, were excluded. Between February 2010 and March 2012 patients were treated with intravenous irinotecan monotherapy 350 mg/m2 q3w and supportive care according to local guidelines. Response evaluations with computerized tomography scans of the chest and abdomen were performed every 9 weeks according to RECIST v. 1.1. Follow-up was ended by September 2012. The primary endpoint of the study was objective response and the secondary endpoints were PFS and overall survival (OS). PFS was defined as time in months from the date of inclusion until the occurrence of disease progression or death from any cause, and OS was defined as time in months from the date of inclusion until death from any cause. An additional inclusion criterion for the present TOP1 study was the availability of histological material. Archival diagnostic FFPE tumor material from either the primary tumor or from a metastatic lesion was collected. The tumor material was originally obtained by core needle biopsies or from the surgical resection specimens. The tissue samples were fixed in neutral buffered formalin and processed and paraffinembedded according to standard protocols of the local department of pathology. The FFPE tumor blocks were cut in 3-mm sections and were stored at 5 C until hybridization. FISH procedures A TOP1/CEN-20 probe mix developed and validated by Dako Denmark (Glostrup, Denmark) and the University of Copenhagen (Copenhagen, Denmark) was used for detection of the TOP1 gene (chromosome 20q12-q13.1) and the reference centromere region on chromosome 20 (CEN-20), and the FISH procedure was performed using the Dako Histology FISH Accessory Kit (Dako Denmark) as previously reported [13]. A fluorescence microscope (Zeiss AX10) with a Texas Red/FITC double filter was used for visualization of the probe signals. TOP1 and CEN-20 signals were enumerated in 60 nonoverlapping cancer nuclei with well-defined nuclear morphology and distinct fluorescent signals of high quality. Overlapping signals of the same color were counted as one signal. The TOP1 and CEN-20 signal count per cell as well as the TOP1/CEN-20 ratio were calculated for each tumor sample. New sections were
86
S. B. Nygård et al.
Scand J Gastroenterol Downloaded from informahealthcare.com by Nyu Medical Center on 06/02/15 For personal use only.
cut and FISH analysis repeated if the fluorescent intensity of the probe signals was weak or if tumor tissue was not present in the original sections. Cases were excluded if signal counting failed following the second hybridization attempt. All procedures were performed blinded to all patient data, and the clinical database was only made accessible for statistical analyses when FISH data were obtained. The samples were evaluated by a single experienced technician (SLN), who was well trained in the FISH technique and the pathological appearance of CRC. Statistics The chi-square test and the Wilcoxon rank sum test were used to test for associations between TOP1 status and baseline characteristics. Logistic regression models were used to test TOP1 and TOP1/ CEN-20 data in relation to best objective response, which was categorized in response (complete and partial response) and non-response (stable disease and disease progression). Odds ratio (OR) estimates and Wald chi-square p-values with estimated 95% confidence intervals (CI) were reported. Univariate Cox proportional hazards regression models were used to test for associations between the variables and the survival endpoints, PFS and OS. The considered variables were the TOP1 signal count, the TOP1/CEN-20 ratio, gender, age, primary tumor site, surgical removal of the primary tumor, and the presence of liver and lung metastases. Hazard ratio (HR) estimates and Wald chi-square p-values with estimated 95% CI were reported for all variables. The HR for age was reported for 10-year differences. Multivariate survival analysis was only considered if statistical significance was demonstrated for the TOP1 signal count or the TOP1/CEN-20 ratio in univariate analysis. Results with a two-sided p-value 4 TOP1 copies per cell when correcting for nuclear truncation due to tissue sectioning [22]. In brief, reference ranges for the TOP1 copy number per cell were calculated based on TOP1 signal counts enumerated in enterocytes of the lamina epithelialis in non-neoplastic colon and rectal mucosa. The normal diploid range was defined as (2n minus 1/2n – 2n plus 1/2n) where 2n equals the mean diploid TOP1 count per cell in normal mucosa. To define ranges for gain a multiplex of n was used. A TOP1/CEN-20 ratio ‡1.5 was also used to separate tumors with TOP1 gain from tumors with normal
TOP1 status. As a result of the explorative nature of the study the TOP1 signal count per cell and the TOP1/CEN-20 ratio were also considered as continuous variables in the logistic regression models and only as such in the Cox proportional hazards regression models. The present study was primarily aimed at investigating a possible trend between TOP1 copy number alterations and objective response why a sample size of 90 patients, most of whom had experienced disease progression was considered large enough for analysis. No imputations of missing values were considered in the statistical models and only tumors with a record of the variables in question entered analysis. The Cox proportional hazards assumptions were assessed by Schoenfeld and martingale residuals, and proportional hazards and linearity were applicable. For the logistic regression models goodness-of-fit was assessed by the Hosmer and Lemeshow test. All statistical analyses were performed using SAS (v9.2, SAS Institute, Cary, NC, USA). Results Tumor samples were available from 81 (90.0%) of the 90 patients assessed for study eligibility. In 78 (96.3 %) samples the signals could be quantified following FISH procedures. After a second hybridization attempt the remaining three (3.7%) samples were still unsuitable for counting due to weak fluorescent signals in one sample and no tumor tissue present in two samples. The majority of samples (n = 74; 94.9%) originated from primary tumors, whereas four (5.1%) were from metastatic lesions. A study profile (adapted from the updated REMARK guidelines [24]) including information on the patient flow through the study, the primary markers and other variables considered, and a statistical plan with number of events and distribution of outcomes is presented in Table I. The median TOP1 and CEN-20 signal count per cell were 3.6 (range: 1.8–6.8) and 2.5 (range: 1.5– 3.9), respectively, and the median TOP1/ CEN-20 ratio was 1.4 (range: 1.0–2.8). Based on the ranges for TOP1 copy number changes, only 5 (6.4%) tumors had a diploid TOP1 copy number, whereas the remaining 73 (93.6%) had copy numbers suggesting some level of gene gain (Table II). No cases with TOP1 gene loss were identified. When using the predetermined cutoff values for the TOP1 signal count per cell and the TOP1/CEN-20 ratio, the frequency of tumors with gain of TOP1 was 41 (52.6%) and 29 (37.2%), respectively. In relation to baseline characteristics the TOP1 signal count per cell significantly associated with gender (p = 0.046) with TOP1 gain more frequently observed in tumors from males. No
TOP1 and objective response to irinotecan
87
Table I. Study profile. No. of patients Patients Assessed for eligibility 90
Scand J Gastroenterol Downloaded from informahealthcare.com by Nyu Medical Center on 06/02/15 For personal use only.
Excluded 9 Failed FISH analysis 3 FISH results obtained 78 Available outcome data
Remarks
Patients with metastatic CRC included between February 2010 and March 2012; irinotecan monotherapy; Department of Oncology, Vejle Sygehus, Denmark Histological material not available No tumor tissue in FFPE sections, n = 2; weak fluorescent signals, n = 1 Objective response data, n = 71 PFS data, n = 77; OS data, n = 77
Markers and variables Markers M1 = TOP1 signal count per cell; M2 = TOP1/CEN-20 ratio Additional variables v1 = gender, v2 = age; v3 = primary tumor site; v4 = primary tumor surgery; v6 = lung metastases Statistical analysis of objective response Variables considered Handling of Response Non-response variables CR PR SD PD M1-M2 Categorical; 0 10 26 35 continuous Statistical analysis of survival outcome Variables considered Handling of PFS OS variables Patients Events Patients Events M1-M2; v2 Continuous 77 76 77 59 v1; v3-v6 Categorical 77 76 77 59
v5 = liver metastases;
Remarks Logistic regression analysis
Remarks Univariate Cox proportional hazards regression analysis
Abbreviations: CR = complete response; CRC = colorectal cancer; OS = overall survival; PD = disease progression; PFS = progression-free survival; PR = partial response; SD = stable disease.
other significant associations with baseline characteristics were demonstrated (Table III). The OR estimates for a stepwise increase of the TOP1 signal count per cell and the TOP1/ CEN-20 ratio in relation to best objective response were 1.62 (0.96–2.75; p = 0.07) and 2.68 (0.43– 16.65; p = 0.29), respectively. Based on the cutoff values for the TOP1 signal count and the TOP1/ CEN-20 ratio, the OR estimates for TOP1 gain versus TOP1 normal were 2.41 (0.57–10.20; p = 0.23) and 2.05 (0.53–7.91; p = 0.30), respectively. No significant associations of the TOP1 signal count or the TOP1/CEN-20 ratio with the survival endpoints (PFS and OS) were demonstrated. In univariate Cox proportional hazards models the HR estimates for continuously increasing values of the TOP1 count in relation to PFS and OS were 1.00 (0.83–1.12; p = 1.00) and 1.01 (0.81–1.26; p = 0.95), respectively.
Likewise, the HR estimates for continuously increasing values of the TOP1/CEN-20 ratio in relation to PFS and OS were 1.08 (0.55–2.12; p = 0.83) and 1.01 (0.45–2.28; p = 0.97), respectively. No other tested covariates were significantly associated with survival (Table IV). Discussion The majority of the 78 evaluable tumors harbored alterations involving the TOP1 locus suggesting some level of gene gain. A signal count compatible with at least two extra TOP1 gene copies per cell was identified in 52.6%, which was more frequent than expected from previous work where 35.7% stage III primary CRC tumors showed the same alteration status using the same methodology [22]. TOP1 is located at chromosome 20q12– 13.1 and gain of the 20q arm or parts of it has previously
Table II. Sample distribution according to TOP1 gene status. Signals per truncated nuclei TOP1 gene status per cell Loss of gene copy Normal/diploid 1 extra gene copy 2 extra gene copies >2 extra gene copies
Copies per whole nucleus
Average
1 2 3 4 >4
0.8 1.6 2.4 3.2
Frequency
Range
No.
%
