IJC International Journal of Cancer
Seasonal variations do not affect the superiority of fecal immunochemical tests over guaiac tests for colorectal cancer screening bastien Chausserie1,2,3,4, Romuald Levillain5, Josette Puvinel6, Olivier Ferrand7, Angela Ruiz5, Thibaut Raginel1,2,3,4, Se Olivier Lantieri5, Guy Launoy1,2,3,4 and Lydia Guittet1,2,3,4 1
Centre Hospitalier Universitaire (CHU) de Caen, Evaluations et recherches en epid emiologie, Caen Normandie University, Caen 3 Universite de Caen Basse-Normandie (UCBN), Cancers and Preventions, Caen 4 INSERM U1086, Cancers and preventions, Caen 5 Institut InterRegional pour la Sante (IRSA) de Tours, Centre de lecture des tests de recherche de sang dans les selles, La Riche, France 6 Association Bourbonnaise Interd epartementale de Depistage des Cancers (ABIDEC), Moulins 7 Action de Depistage Organise des Cancers 18 (ADOC18), Saint-Doulchard
The aim of this study was to compare the seasonal variation in performance of a faecal immunochemical test for haemoglobin (FIT) and a guaiac test (gFOBT) for colorectal cancer screening. From June 2009 to May 2011, 18,290 screening participants (50–74 years old) performed OC-SENSOR quantitative FIT (1 sample) and Hemoccult II gFOBT (3 stool samples with 2 spots/sample). Referral for colonoscopy required a minimum of one positive spot (gFOBT), or a positive FIT [cut-off 150 ng haemoglobin/mL buffer (i.e. 30 lg haemoglobin/g feces)]. The performance of tests for detection of advanced neoplasia was compared according to seasons using Receiver Operating Characteristics (ROC) curves, at various FIT cut-off values. The positivity rate of FIT was significantly lower in the summer compared with other seasons (2.3% versus 3.0%, p 5 0.03), whilst the positivity rate of gFOBT increased in the autumn (1.8% versus 1.5%, p 5 0.11). FIT was clinically more effective than gFOBT over the four season-specific ROC curves. At the cut-off concentration used in the study, the season-specific FIT/gFOBT ratios for true positive rates were: 2.8 (Autumn), 2.5 (Winter), 3.0 (Spring), 3.7 (Summer), and for false positive rates: 1.2 (Autumn), 1.5 (Winter), 1.8 (Spring), 0.9 (Summer). Therefore, in this study with this cut-off concentration and in spite of lower positivity rate in summer, the seasonal variations of performance of OC-SENSOR FIT led to improved gain in specificity in the summer, without a decrease in gain in sensitivity compared with gFOBT.
Introduction Colorectal cancer is one of the most frequent cancers in the world. Two-step screening, with colonoscopic follow-up of positive guaiac faecal occult blood tests (gFOBT), is effective in reducing colorectal cancer mortality and incidence.1 Several studies have demonstrated that automated quantitative faecal immunochemical tests for haemoglobin (FIT) perform better than gFOBT for detection of colorectal advanced neoplasia.2–7 At the 150 ng haemoglobin/mL buffer (i.e., 30 mg Key words: occult blood, colorectal neoplasms, early detection of cancer, sensitivity and specificity, immunochemistry/methods, comparative study Grant sponsors: French National Institute for Cancer (INCa, Institut National du Cancer), The French National League Against Cancer (LNCC, Ligue Nationale Contre le Cancer). DOI: 10.1002/ijc.29187 History: Received 26 July 2013; Accepted 12 Aug 2014; Online 5 Sep 2014 Correspondence to: Dr. Lydia Guittet, Inserm U1086, Cancers & Preventions, CHU Caen Centre Franc¸ois Baclesse, Avenue du General Harris, Caen 14000, Tel.: 133.231.45.86.24, E-mail: [email protected].
C 2014 UICC Int. J. Cancer: 136, 1827–1834 (2015) V
haemoglobin/g feces) cut-off, a single sample of OC-SENSOR FIT (Eiken Chemical Co, Tokyo, Japan) markedly increased the number of true positives (subjects with colorectal advanced neoplasia) compared with a 3 stools, 2 spots/stool, gFOBT (positivity defined as at least one positive slot over six), with only a moderate increase in false positives.6,7 Increasing the single sample OC-SENSOR cut-off concentration until 300 ng haemoglobin/mL buffer (i.e., 60 mg haemoglobin/g feces), decreased the positivity rate until the gFOBT’s one, but still increased the number of true positives, hence allowing health care systems to select the colonoscopy workload.6,7 Finally, FITs measurement can be automated and will reduce variability from subjective assessment of the gFOBT by human operators. However, a decrease in the positivity rate of FITs has been observed in the summer period for both MagStream FIT (Fujirebio, Tokyo, Japan),8 and to a lesser extent, for OC-SENSOR FIT (Eiken Chemical Co, Tokyo, Japan), in Italy and The Netherlands.9,10 The impact of decreased positivity rate of FIT in the summer period on the effectiveness of screening programmes depends on whether it affects the risk of being false positive, the risk of being false negative, or
Early Detection and Diagnosis
2
1828
Seasonal variation in FIT performance
Early Detection and Diagnosis
What’s new? Fecal immunochemical tests (FITs) for hemoglobin are expected to replace traditional guaiac based fecal occult blood tests (gFOBTs) as a screening tool for colorectal cancer. But while more sensitive than gFOBTs, some studies have suggested that FIT performance is influenced by seasonal variations in temperature, particularly during the summer season. Here, summer temperatures were found to have no effect on the gain in sensitivity of an automated quantitative FIT (OC-SENSOR, cut-off 30lg hemoglobin/g feces), despite decreased positivity rates in summer. Across the four seasons, FIT outperformed gFOBT in clinical effectiveness.
both.11 In addition, seasonal variations can result from differences in the characteristics of the population participating during the summer, or to lack of stability of the FIT sample. In vitro analytical studies demonstrated the poor stability of FITs to temperature, and showed strong variations between buffers used in FITs,12,13 the OC-SENSOR Auto-sampling bottle 3 being less sensitive to high temperature than the MagStream New Hemtube. Recent laboratory analyses demonstrated improved stability with new generation buffers.14 This lack of robustness of FIT to elevated temperature could offset their clinical superiority compared with gFOBT. Nevertheless, the performances of gFOBT has also been described to be subject to variation in positivity rate according to season,15,16 and to increased positivity rates in humid weather. The aim of this study was to compare directly the seasonal variation in screening performances, measured as true-positive and false-positive rates, of OC-SENSOR FIT and Hemoccult II gFOBT in colorectal cancer screening.
Material and Methods Study design
From June 2009 to May 2011, a trial screening programme for the detection of colorectal cancer was provided for residents of two French administrative regions (Cher and Allier) involving several FOBTs, using the guaiac Hemoccult II (SDK, Villepinte, France) and OC-SENSOR FIT (Eiken Chemical Co, Tokyo, Japan). This study was undertaken using centralized postal invitations sent to the current screening programme population. Each patient met specific inclusion criteria: 50 to 74 years old, no recent digestive symptoms requiring colonoscopy evaluation, no personal or first degree relatives with prior colorectal cancer or advanced adenoma and no colonoscopy investigation within the past five years. These data were obtained by general practitioners who provided the tests after patients’ acceptance and their written consent. Each patient was asked to perform the sample collections at home, record the date of each sample collection and return the samples together to the centralized IRSA laboratory (Institut interRegional pour la Sante, Tours, France) by post in a pre-stamped envelope as soon as possible after the last collection. Three consecutive stools were required for sampling. Sampling was applied on a collecting card for Hemoccult II (2 spots/sampled stool), and into OC Auto
sampling bottle3 collecting tube for OC-SENSOR (Product code V-PZ25 sampling tube delivered in April 2009 and March 2010). The gFOBT was performed on the three stools, but the FIT was only performed on the two last stools (one FIT device each). In this analysis, to fit with a one sample FIT, only the last FIT sampling was considered. No specific storage conditions were specified before test return. Guaiac tests were processed independently according to manufacturers’ recommendations and readers of the centralised analysis centre (Institut inter-Regional pour la Sante de Tours reading centre, one of the 32 laboratories approved for gFOBT reading in the French national colorectal cancer screening programme) were unaware of the FIT results. The laboratory workforce was stable throughout the study period, with no particular change during the summer. The gFOBTs performed in this study were read together with other gFOBTs coming from the routine national gFOBT ongoing screening programme and supported by the same national quality insurance programme. Upon arrival at the laboratory, the FITs were kept at the room temperature and analysed on the same day, using an OC-SENSOR DIANA instrument (Automat number N00385). The concentrations of haemoglobin measured in the buffer (ng/mL) are numerically five times the concentration when expressed in the feces (mg/g), which is the unit used throughout the study.12,13,17 Screened persons were considered positive if at least one Hemoccult II sample was positive out of six or if the FIT sample was found to contain more than 30 lg/g of haemoglobin concentration in the feces, a cut-off just over the manufacturer’s one, hence allowing comparison with other studies but limiting the overall positivity rate. The screening result was shared with both the patients and the general practitioners without details on each FOBT result. Positive screened persons were referred to colonoscopy. All colonoscopies were performed by gastroenterologists and were defined as positive if advanced neoplasia was found. The subjects were considered as having advanced neoplasia, if their worst lesion detected at colonoscopy was an advanced adenoma (adenomas measuring 10 mm or more and adenomas with high-grade dysplasia, the villous component being not considered in this definition), an in situ or intra-mucosal carcinoma, or invasive cancer (invasion of malignant cells beyond the muscularis mucosae). Hyperplasic polyps were not included as neoplasia. Finally, patients that underwent an C 2014 UICC Int. J. Cancer: 136, 1827–1834 (2015) V
1829
Chausserie et al.
Statistical analysis
Subjects were excluded from this analysis if Hemoccult II or OC-SENSOR could not be analysed (no definitive screening result), or date of sampling was missing or aberrant (negative or greater than 10 days). The sample return time was defined as the interval in days between the date of last faecal sample at home and tests laboratory delivery (FITs analysed on the day of receipt). This return time was categorized as short return time (
Seasonal variations do not affect the superiority of fecal immunochemical tests over guaiac tests for colorectal cancer screening bastien Chausserie1,2,3,4, Romuald Levillain5, Josette Puvinel6, Olivier Ferrand7, Angela Ruiz5, Thibaut Raginel1,2,3,4, Se Olivier Lantieri5, Guy Launoy1,2,3,4 and Lydia Guittet1,2,3,4 1
Centre Hospitalier Universitaire (CHU) de Caen, Evaluations et recherches en epid emiologie, Caen Normandie University, Caen 3 Universite de Caen Basse-Normandie (UCBN), Cancers and Preventions, Caen 4 INSERM U1086, Cancers and preventions, Caen 5 Institut InterRegional pour la Sante (IRSA) de Tours, Centre de lecture des tests de recherche de sang dans les selles, La Riche, France 6 Association Bourbonnaise Interd epartementale de Depistage des Cancers (ABIDEC), Moulins 7 Action de Depistage Organise des Cancers 18 (ADOC18), Saint-Doulchard
The aim of this study was to compare the seasonal variation in performance of a faecal immunochemical test for haemoglobin (FIT) and a guaiac test (gFOBT) for colorectal cancer screening. From June 2009 to May 2011, 18,290 screening participants (50–74 years old) performed OC-SENSOR quantitative FIT (1 sample) and Hemoccult II gFOBT (3 stool samples with 2 spots/sample). Referral for colonoscopy required a minimum of one positive spot (gFOBT), or a positive FIT [cut-off 150 ng haemoglobin/mL buffer (i.e. 30 lg haemoglobin/g feces)]. The performance of tests for detection of advanced neoplasia was compared according to seasons using Receiver Operating Characteristics (ROC) curves, at various FIT cut-off values. The positivity rate of FIT was significantly lower in the summer compared with other seasons (2.3% versus 3.0%, p 5 0.03), whilst the positivity rate of gFOBT increased in the autumn (1.8% versus 1.5%, p 5 0.11). FIT was clinically more effective than gFOBT over the four season-specific ROC curves. At the cut-off concentration used in the study, the season-specific FIT/gFOBT ratios for true positive rates were: 2.8 (Autumn), 2.5 (Winter), 3.0 (Spring), 3.7 (Summer), and for false positive rates: 1.2 (Autumn), 1.5 (Winter), 1.8 (Spring), 0.9 (Summer). Therefore, in this study with this cut-off concentration and in spite of lower positivity rate in summer, the seasonal variations of performance of OC-SENSOR FIT led to improved gain in specificity in the summer, without a decrease in gain in sensitivity compared with gFOBT.
Introduction Colorectal cancer is one of the most frequent cancers in the world. Two-step screening, with colonoscopic follow-up of positive guaiac faecal occult blood tests (gFOBT), is effective in reducing colorectal cancer mortality and incidence.1 Several studies have demonstrated that automated quantitative faecal immunochemical tests for haemoglobin (FIT) perform better than gFOBT for detection of colorectal advanced neoplasia.2–7 At the 150 ng haemoglobin/mL buffer (i.e., 30 mg Key words: occult blood, colorectal neoplasms, early detection of cancer, sensitivity and specificity, immunochemistry/methods, comparative study Grant sponsors: French National Institute for Cancer (INCa, Institut National du Cancer), The French National League Against Cancer (LNCC, Ligue Nationale Contre le Cancer). DOI: 10.1002/ijc.29187 History: Received 26 July 2013; Accepted 12 Aug 2014; Online 5 Sep 2014 Correspondence to: Dr. Lydia Guittet, Inserm U1086, Cancers & Preventions, CHU Caen Centre Franc¸ois Baclesse, Avenue du General Harris, Caen 14000, Tel.: 133.231.45.86.24, E-mail: [email protected].
C 2014 UICC Int. J. Cancer: 136, 1827–1834 (2015) V
haemoglobin/g feces) cut-off, a single sample of OC-SENSOR FIT (Eiken Chemical Co, Tokyo, Japan) markedly increased the number of true positives (subjects with colorectal advanced neoplasia) compared with a 3 stools, 2 spots/stool, gFOBT (positivity defined as at least one positive slot over six), with only a moderate increase in false positives.6,7 Increasing the single sample OC-SENSOR cut-off concentration until 300 ng haemoglobin/mL buffer (i.e., 60 mg haemoglobin/g feces), decreased the positivity rate until the gFOBT’s one, but still increased the number of true positives, hence allowing health care systems to select the colonoscopy workload.6,7 Finally, FITs measurement can be automated and will reduce variability from subjective assessment of the gFOBT by human operators. However, a decrease in the positivity rate of FITs has been observed in the summer period for both MagStream FIT (Fujirebio, Tokyo, Japan),8 and to a lesser extent, for OC-SENSOR FIT (Eiken Chemical Co, Tokyo, Japan), in Italy and The Netherlands.9,10 The impact of decreased positivity rate of FIT in the summer period on the effectiveness of screening programmes depends on whether it affects the risk of being false positive, the risk of being false negative, or
Early Detection and Diagnosis
2
1828
Seasonal variation in FIT performance
Early Detection and Diagnosis
What’s new? Fecal immunochemical tests (FITs) for hemoglobin are expected to replace traditional guaiac based fecal occult blood tests (gFOBTs) as a screening tool for colorectal cancer. But while more sensitive than gFOBTs, some studies have suggested that FIT performance is influenced by seasonal variations in temperature, particularly during the summer season. Here, summer temperatures were found to have no effect on the gain in sensitivity of an automated quantitative FIT (OC-SENSOR, cut-off 30lg hemoglobin/g feces), despite decreased positivity rates in summer. Across the four seasons, FIT outperformed gFOBT in clinical effectiveness.
both.11 In addition, seasonal variations can result from differences in the characteristics of the population participating during the summer, or to lack of stability of the FIT sample. In vitro analytical studies demonstrated the poor stability of FITs to temperature, and showed strong variations between buffers used in FITs,12,13 the OC-SENSOR Auto-sampling bottle 3 being less sensitive to high temperature than the MagStream New Hemtube. Recent laboratory analyses demonstrated improved stability with new generation buffers.14 This lack of robustness of FIT to elevated temperature could offset their clinical superiority compared with gFOBT. Nevertheless, the performances of gFOBT has also been described to be subject to variation in positivity rate according to season,15,16 and to increased positivity rates in humid weather. The aim of this study was to compare directly the seasonal variation in screening performances, measured as true-positive and false-positive rates, of OC-SENSOR FIT and Hemoccult II gFOBT in colorectal cancer screening.
Material and Methods Study design
From June 2009 to May 2011, a trial screening programme for the detection of colorectal cancer was provided for residents of two French administrative regions (Cher and Allier) involving several FOBTs, using the guaiac Hemoccult II (SDK, Villepinte, France) and OC-SENSOR FIT (Eiken Chemical Co, Tokyo, Japan). This study was undertaken using centralized postal invitations sent to the current screening programme population. Each patient met specific inclusion criteria: 50 to 74 years old, no recent digestive symptoms requiring colonoscopy evaluation, no personal or first degree relatives with prior colorectal cancer or advanced adenoma and no colonoscopy investigation within the past five years. These data were obtained by general practitioners who provided the tests after patients’ acceptance and their written consent. Each patient was asked to perform the sample collections at home, record the date of each sample collection and return the samples together to the centralized IRSA laboratory (Institut interRegional pour la Sante, Tours, France) by post in a pre-stamped envelope as soon as possible after the last collection. Three consecutive stools were required for sampling. Sampling was applied on a collecting card for Hemoccult II (2 spots/sampled stool), and into OC Auto
sampling bottle3 collecting tube for OC-SENSOR (Product code V-PZ25 sampling tube delivered in April 2009 and March 2010). The gFOBT was performed on the three stools, but the FIT was only performed on the two last stools (one FIT device each). In this analysis, to fit with a one sample FIT, only the last FIT sampling was considered. No specific storage conditions were specified before test return. Guaiac tests were processed independently according to manufacturers’ recommendations and readers of the centralised analysis centre (Institut inter-Regional pour la Sante de Tours reading centre, one of the 32 laboratories approved for gFOBT reading in the French national colorectal cancer screening programme) were unaware of the FIT results. The laboratory workforce was stable throughout the study period, with no particular change during the summer. The gFOBTs performed in this study were read together with other gFOBTs coming from the routine national gFOBT ongoing screening programme and supported by the same national quality insurance programme. Upon arrival at the laboratory, the FITs were kept at the room temperature and analysed on the same day, using an OC-SENSOR DIANA instrument (Automat number N00385). The concentrations of haemoglobin measured in the buffer (ng/mL) are numerically five times the concentration when expressed in the feces (mg/g), which is the unit used throughout the study.12,13,17 Screened persons were considered positive if at least one Hemoccult II sample was positive out of six or if the FIT sample was found to contain more than 30 lg/g of haemoglobin concentration in the feces, a cut-off just over the manufacturer’s one, hence allowing comparison with other studies but limiting the overall positivity rate. The screening result was shared with both the patients and the general practitioners without details on each FOBT result. Positive screened persons were referred to colonoscopy. All colonoscopies were performed by gastroenterologists and were defined as positive if advanced neoplasia was found. The subjects were considered as having advanced neoplasia, if their worst lesion detected at colonoscopy was an advanced adenoma (adenomas measuring 10 mm or more and adenomas with high-grade dysplasia, the villous component being not considered in this definition), an in situ or intra-mucosal carcinoma, or invasive cancer (invasion of malignant cells beyond the muscularis mucosae). Hyperplasic polyps were not included as neoplasia. Finally, patients that underwent an C 2014 UICC Int. J. Cancer: 136, 1827–1834 (2015) V
1829
Chausserie et al.
Statistical analysis
Subjects were excluded from this analysis if Hemoccult II or OC-SENSOR could not be analysed (no definitive screening result), or date of sampling was missing or aberrant (negative or greater than 10 days). The sample return time was defined as the interval in days between the date of last faecal sample at home and tests laboratory delivery (FITs analysed on the day of receipt). This return time was categorized as short return time (
