Downloaded from http://gut.bmj.com/ on August 8, 2015 - Published by group.bmj.com
Inflammatory bowel disease
ORIGINAL ARTICLE
Predictors and risks for death in a population-based study of persons with IBD in Manitoba Charles N Bernstein,1,2 Zoann Nugent,2,3,4 Laura E Targownik,1,2 Harminder Singh,1,2,3,4 Lisa M Lix2,3 ▸ Additional material is published online only. To view please visit the journal online (http://dx.doi.org/10.1136/ gutjnl-2014-307983). 1
Department of Internal Medicine, University of Manitoba, Winnipeg, Manitoba, Canada 2 University of Manitoba IBD Clinical and Research Centre, Winnipeg, Manitoba, Canada 3 Community Health Sciences, Winnipeg, Manitoba, Canada 4 CancerCare Manitoba, Winnipeg, Manitoba, Canada Correspondence to Dr Charles N Bernstein, University of Manitoba, 804F715 McDermot Avenue, Winnipeg, Manitoba, Canada R3E3P4; charles.bernstein@ med.umanitoba.ca Received 5 July 2014 Revised 20 August 2014 Accepted 25 August 2014 Published Online First 16 September 2014
ABSTRACT Background and aims We aimed to determine the predictors and risk for death among persons with either Crohn’s disease (CD) or UC compared with the general population. Methods We used the population-based University of Manitoba IBD Epidemiology Database to calculate HRs and their 95% CIs for cases relative to controls using stratified multivariable Cox proportional hazards regression models, controlling for socioeconomic status and comorbidities. Results There were 10 788 prevalent cases of CD and UC and 101 860 matched controls. The HR for all-cause mortality in prevalent CD cases was 1.26 (95% CI 1.16 to 1.38) and in prevalent UC cases was 1.04 (95% CI 0.96 to 1.12). Compared with matched controls, CD cases were more likely to die of colorectal cancer, nonHodgkin’s lymphoma, digestive diseases, pulmonary embolism and sepsis and UC cases were more likely to die from colorectal cancer, digestive diseases and respiratory diseases. For incident cases, there were significant effects on mortality by socioeconomic status, comorbidity score and surgery. The greatest risk for death in both CD and UC was within the first 30 days following GI surgery. The first year from diagnosis was associated with increased risk of death in both CD and UC, but persisted after the 1st year only in CD. Conclusions There is a significantly increased risk of mortality in CD compared with controls while in UC an increased risk for death was only evident in the first year from diagnosis. Surgery poses an increased risk for death in both CD and UC lasting up to 1 year.
INTRODUCTION
To cite: Bernstein CN, Nugent Z, Targownik LE, et al. Gut 2015;64: 1403–1411.
As the prevalence of IBD increases worldwide,1 it is important to determine if outcomes for individuals with IBD are comparable from different jurisdictions. This might provide some insight if the disease in different jurisdictions is the same disease and/or whether the diversity of management in different jurisdictions differentially impacts on important clinical outcomes. A key outcome is mortality and defining predictors of mortality may point to aspects of care that can potentially be enhanced or may identify associated comorbidity that warrants attention. While several other studies have reported on mortality rates in IBD, few have examined predictors of mortality. Further, as IBD is often diagnosed in young persons, even small increases in mortality rates may be impactful on productive years of life lost.
Significance of this study What is already known on this subject?
▸ Compared with the general population, there is an increased mortality rate in Crohn’s disease (CD) with conflicting data as to the relative mortality rates in UC. There is little information regarding predictors of mortality in IBD.
What are the new findings?
▸ The first year from diagnosis was associated with an increased mortality rate in both CD and UC but beyond the first year from diagnosis the increased mortality only persisted in CD. ▸ The top causes of death in descending order for both CD and UC were cancer, circulatory disease and digestive disease. ▸ In terms of cause-specific mortality compared with population controls, prevalent CD cases were more likely to die of colorectal cancer, non-Hodgkin’s lymphoma, digestive diseases, pulmonary embolism and sepsis than matched controls and prevalent UC cases were more likely to die from colorectal cancer, digestive diseases and respiratory diseases. ▸ GI surgery had a strong impact on mortality rates. While the impact was greatest in the first 30 days from surgery, the increased mortality rate persisted in both CD and UC beyond 1 year from surgery.
How might it impact on clinical practice in the foreseeable future?
▸ That surgery, especially among the elderly, had a significant impact on mortality should give clinicians pause both with regard to their reluctance to using immunomodulatory medications in both CD and UC and in considering surgery too late when the disease may be more advanced and the surgical risks higher. ▸ Venous thromboembolism and sepsis in CD and colorectal cancer in both CD and UC are associated with mortality more than in the general population underscoring the importance of thromboembolism prophylaxis when indicated and the need for ongoing colorectal cancer surveillance in IBD.
Bernstein CN, et al. Gut 2015;64:1403–1411. doi:10.1136/gutjnl-2014-307983
1403
Downloaded from http://gut.bmj.com/ on August 8, 2015 - Published by group.bmj.com
Inflammatory bowel disease There has been a considerable literature written to date on mortality in IBD including four meta-analyses.2–5 In these studies, having Crohn’s disease (CD) has been associated with an increased risk of mortality2–5 while results as to whether UC is associated with increased mortality rate are equivocal.4 5 Moreover, there are limited data on IBD associated mortality from North America and the results for CD have been conflicting.6 7 Furthermore, it is unclear whether the mortality risk among persons with IBD is influenced primarily by the underlying disease process, its surgical and medical treatments, or concomitant medical conditions which are associated with IBD. The aim of our study was to compare the all-cause mortality in CD and UC with that in a matched group of persons without IBD in Manitoba, Canada, using our population-based University of Manitoba IBD Epidemiology Database (UMIBDED). Further, we aimed to estimate the mortality rate for cases relative to a matched population due to specific conditions known to be more prevalent in IBD (such as colorectal cancer and venous thromboembolic disease as well as GI surgery) as well as other causes that are not specifically related to IBD (e.g., non-GI cancers, cardiovascular disease and suicide). We also aimed to evaluate the effects of socioeconomic status, medical comorbidity, time from diagnosis and time from GI surgery on overall mortality in IBD.
METHODS Data sources Cases and controls were identified from the population-based UMIBDED, which was established in 1995 and is described in detail elsewhere.8 Manitoba is a central Canadian province with a population of 1.25 million in 2010; its population has experienced relatively little in- or out-migration over the past several decades. Briefly, the UMIBDED contains administrative data for IBD cases, ascertained using a validated case definition and matched controls. It includes individuals who had health coverage between 1 April 1984 and 31 March 2010 as identified from the administrative health databases, which include the registry of all enrolled residents in the province, hospital discharge records and physician billing claims of Manitoba Health (MH). MH is the single provincial health insurance provider for all Manitoba residents. Residents of Manitoba who resided in the province for at least 2 years were identified as IBD cases if they had at least five unique physician contacts (either hospitalisation and/or outpatient physician billing claims) associated with International Classification of Disease (ICD)-9-CM codes 555 (CD) or 556 (UC) recorded as a diagnosis, or ICD-10-CA codes K50 and K51. ICD-10-CA codes have been used to report diagnoses in hospitalisation records since 2004. Persons residing in Manitoba for less than 2 years were included in the UMIBDED as IBD cases if they had at least three separate diagnoses with these ICD codes in the administrative health databases. The sensitivity and specificity of this case definition are both over 90% when compared with self-report data and chart-reviewed diagnosis.8 Specificity is closer to 100% in the entire population. We defined each IBD case as being prevalent or incident. Prevalent IBD cases included all individuals identified with IBD between 1 April 1984 and 31 March 2010. Incident IBD cases were those individuals who had at least 3 years of registration with MH prior to their initial IBD related encounter. The diagnosis date for incident cases was taken to be the date of the initial IBD related healthcare contact. Up to 10 controls were randomly matched to each IBD case, based on sex, region of residence (based on first three letters or numbers of the case’s 6-digit 1404
postal code) and year of birth. A matched control was assigned the same index date as the respective IBD case (date of first IBD related encounter). Deaths (and cause of deaths) were identified from the Vital Statistics Registry for the Province of Manitoba, which was linked to the UMIBDED. While date of death was available up to 31 March 2010, cause of death was available only to 31 December 2009. For prevalent cases through 31 December 2009, we identified all-cause and cause-specific deaths; for incident cases, only all-cause deaths were identified. Within the UMIBDED, 76% of cases and 71% of controls were alive and in the study at the end date. Overall, 11% of cases and 10% of controls died during the study, and 13% of cases and 19% of controls lost MH coverage (likely meaning they moved from the province). These last subjects were censored at the time they lost MH coverage.
Predictors of mortality and statistical analysis Socioeconomic status was assessed by the Socioeconomic Factor Index (SEFI) which is a score derived from the application of exploratory factor analysis techniques to Statistics Canada Census 2006 data and includes several neighbourhood level social determinants of wealth, including age dependency ratio, rate of single parent households, rate of female single parent households, female labour force participation rate, unemployment rate and high school education rate. The SEFI has been validated in Manitoba and is described in detail elsewhere.9 To calculate the Charlson Comorbidity Index (CCI) score,10 comorbidities were identified from all diagnoses recorded in the administrative health data for the time period between 6 and 18 months prior to an individual’s end of follow-up (i.e., date of death or emigration or 31 March 2010, whichever occurred earlier). If persons had an end date prior to 1 October 1985 (i.e., they did not have 18 months of data to assess for comorbidities) or did not have MH coverage for the period of time 18 months prior to the end date, CCI was assigned a missing value. The CCI was categorised as 0, 1, 2 or ≥3 with the reference category being 0.10 The data were descriptively analysed using means, SDs, frequencies and percentages. HRs and 95% CIs were estimated for IBD cases relative to controls using stratified multivariable Cox proportional hazards regression models; the covariates were the SEFI and CCI scores. Separate models were defined for CD and UC cases and their respective controls. For incident IBD cases only, separate models were fit to the data for deaths within 1 year of diagnosis and to the data for all deaths following IBD diagnosis and also by history of GI surgery (defined as small or large bowel resection, repair, stricture surgery or stoma creation). GI surgery was a time-dependent variable: no/presurgery, 0–29 days postsurgery, 30–365 days postsurgery and 366+ days postsurgery. Analyses were conducted for all ages (at the index/ IBD diagnosis date), as well as for the following current age groups:
Inflammatory bowel disease
ORIGINAL ARTICLE
Predictors and risks for death in a population-based study of persons with IBD in Manitoba Charles N Bernstein,1,2 Zoann Nugent,2,3,4 Laura E Targownik,1,2 Harminder Singh,1,2,3,4 Lisa M Lix2,3 ▸ Additional material is published online only. To view please visit the journal online (http://dx.doi.org/10.1136/ gutjnl-2014-307983). 1
Department of Internal Medicine, University of Manitoba, Winnipeg, Manitoba, Canada 2 University of Manitoba IBD Clinical and Research Centre, Winnipeg, Manitoba, Canada 3 Community Health Sciences, Winnipeg, Manitoba, Canada 4 CancerCare Manitoba, Winnipeg, Manitoba, Canada Correspondence to Dr Charles N Bernstein, University of Manitoba, 804F715 McDermot Avenue, Winnipeg, Manitoba, Canada R3E3P4; charles.bernstein@ med.umanitoba.ca Received 5 July 2014 Revised 20 August 2014 Accepted 25 August 2014 Published Online First 16 September 2014
ABSTRACT Background and aims We aimed to determine the predictors and risk for death among persons with either Crohn’s disease (CD) or UC compared with the general population. Methods We used the population-based University of Manitoba IBD Epidemiology Database to calculate HRs and their 95% CIs for cases relative to controls using stratified multivariable Cox proportional hazards regression models, controlling for socioeconomic status and comorbidities. Results There were 10 788 prevalent cases of CD and UC and 101 860 matched controls. The HR for all-cause mortality in prevalent CD cases was 1.26 (95% CI 1.16 to 1.38) and in prevalent UC cases was 1.04 (95% CI 0.96 to 1.12). Compared with matched controls, CD cases were more likely to die of colorectal cancer, nonHodgkin’s lymphoma, digestive diseases, pulmonary embolism and sepsis and UC cases were more likely to die from colorectal cancer, digestive diseases and respiratory diseases. For incident cases, there were significant effects on mortality by socioeconomic status, comorbidity score and surgery. The greatest risk for death in both CD and UC was within the first 30 days following GI surgery. The first year from diagnosis was associated with increased risk of death in both CD and UC, but persisted after the 1st year only in CD. Conclusions There is a significantly increased risk of mortality in CD compared with controls while in UC an increased risk for death was only evident in the first year from diagnosis. Surgery poses an increased risk for death in both CD and UC lasting up to 1 year.
INTRODUCTION
To cite: Bernstein CN, Nugent Z, Targownik LE, et al. Gut 2015;64: 1403–1411.
As the prevalence of IBD increases worldwide,1 it is important to determine if outcomes for individuals with IBD are comparable from different jurisdictions. This might provide some insight if the disease in different jurisdictions is the same disease and/or whether the diversity of management in different jurisdictions differentially impacts on important clinical outcomes. A key outcome is mortality and defining predictors of mortality may point to aspects of care that can potentially be enhanced or may identify associated comorbidity that warrants attention. While several other studies have reported on mortality rates in IBD, few have examined predictors of mortality. Further, as IBD is often diagnosed in young persons, even small increases in mortality rates may be impactful on productive years of life lost.
Significance of this study What is already known on this subject?
▸ Compared with the general population, there is an increased mortality rate in Crohn’s disease (CD) with conflicting data as to the relative mortality rates in UC. There is little information regarding predictors of mortality in IBD.
What are the new findings?
▸ The first year from diagnosis was associated with an increased mortality rate in both CD and UC but beyond the first year from diagnosis the increased mortality only persisted in CD. ▸ The top causes of death in descending order for both CD and UC were cancer, circulatory disease and digestive disease. ▸ In terms of cause-specific mortality compared with population controls, prevalent CD cases were more likely to die of colorectal cancer, non-Hodgkin’s lymphoma, digestive diseases, pulmonary embolism and sepsis than matched controls and prevalent UC cases were more likely to die from colorectal cancer, digestive diseases and respiratory diseases. ▸ GI surgery had a strong impact on mortality rates. While the impact was greatest in the first 30 days from surgery, the increased mortality rate persisted in both CD and UC beyond 1 year from surgery.
How might it impact on clinical practice in the foreseeable future?
▸ That surgery, especially among the elderly, had a significant impact on mortality should give clinicians pause both with regard to their reluctance to using immunomodulatory medications in both CD and UC and in considering surgery too late when the disease may be more advanced and the surgical risks higher. ▸ Venous thromboembolism and sepsis in CD and colorectal cancer in both CD and UC are associated with mortality more than in the general population underscoring the importance of thromboembolism prophylaxis when indicated and the need for ongoing colorectal cancer surveillance in IBD.
Bernstein CN, et al. Gut 2015;64:1403–1411. doi:10.1136/gutjnl-2014-307983
1403
Downloaded from http://gut.bmj.com/ on August 8, 2015 - Published by group.bmj.com
Inflammatory bowel disease There has been a considerable literature written to date on mortality in IBD including four meta-analyses.2–5 In these studies, having Crohn’s disease (CD) has been associated with an increased risk of mortality2–5 while results as to whether UC is associated with increased mortality rate are equivocal.4 5 Moreover, there are limited data on IBD associated mortality from North America and the results for CD have been conflicting.6 7 Furthermore, it is unclear whether the mortality risk among persons with IBD is influenced primarily by the underlying disease process, its surgical and medical treatments, or concomitant medical conditions which are associated with IBD. The aim of our study was to compare the all-cause mortality in CD and UC with that in a matched group of persons without IBD in Manitoba, Canada, using our population-based University of Manitoba IBD Epidemiology Database (UMIBDED). Further, we aimed to estimate the mortality rate for cases relative to a matched population due to specific conditions known to be more prevalent in IBD (such as colorectal cancer and venous thromboembolic disease as well as GI surgery) as well as other causes that are not specifically related to IBD (e.g., non-GI cancers, cardiovascular disease and suicide). We also aimed to evaluate the effects of socioeconomic status, medical comorbidity, time from diagnosis and time from GI surgery on overall mortality in IBD.
METHODS Data sources Cases and controls were identified from the population-based UMIBDED, which was established in 1995 and is described in detail elsewhere.8 Manitoba is a central Canadian province with a population of 1.25 million in 2010; its population has experienced relatively little in- or out-migration over the past several decades. Briefly, the UMIBDED contains administrative data for IBD cases, ascertained using a validated case definition and matched controls. It includes individuals who had health coverage between 1 April 1984 and 31 March 2010 as identified from the administrative health databases, which include the registry of all enrolled residents in the province, hospital discharge records and physician billing claims of Manitoba Health (MH). MH is the single provincial health insurance provider for all Manitoba residents. Residents of Manitoba who resided in the province for at least 2 years were identified as IBD cases if they had at least five unique physician contacts (either hospitalisation and/or outpatient physician billing claims) associated with International Classification of Disease (ICD)-9-CM codes 555 (CD) or 556 (UC) recorded as a diagnosis, or ICD-10-CA codes K50 and K51. ICD-10-CA codes have been used to report diagnoses in hospitalisation records since 2004. Persons residing in Manitoba for less than 2 years were included in the UMIBDED as IBD cases if they had at least three separate diagnoses with these ICD codes in the administrative health databases. The sensitivity and specificity of this case definition are both over 90% when compared with self-report data and chart-reviewed diagnosis.8 Specificity is closer to 100% in the entire population. We defined each IBD case as being prevalent or incident. Prevalent IBD cases included all individuals identified with IBD between 1 April 1984 and 31 March 2010. Incident IBD cases were those individuals who had at least 3 years of registration with MH prior to their initial IBD related encounter. The diagnosis date for incident cases was taken to be the date of the initial IBD related healthcare contact. Up to 10 controls were randomly matched to each IBD case, based on sex, region of residence (based on first three letters or numbers of the case’s 6-digit 1404
postal code) and year of birth. A matched control was assigned the same index date as the respective IBD case (date of first IBD related encounter). Deaths (and cause of deaths) were identified from the Vital Statistics Registry for the Province of Manitoba, which was linked to the UMIBDED. While date of death was available up to 31 March 2010, cause of death was available only to 31 December 2009. For prevalent cases through 31 December 2009, we identified all-cause and cause-specific deaths; for incident cases, only all-cause deaths were identified. Within the UMIBDED, 76% of cases and 71% of controls were alive and in the study at the end date. Overall, 11% of cases and 10% of controls died during the study, and 13% of cases and 19% of controls lost MH coverage (likely meaning they moved from the province). These last subjects were censored at the time they lost MH coverage.
Predictors of mortality and statistical analysis Socioeconomic status was assessed by the Socioeconomic Factor Index (SEFI) which is a score derived from the application of exploratory factor analysis techniques to Statistics Canada Census 2006 data and includes several neighbourhood level social determinants of wealth, including age dependency ratio, rate of single parent households, rate of female single parent households, female labour force participation rate, unemployment rate and high school education rate. The SEFI has been validated in Manitoba and is described in detail elsewhere.9 To calculate the Charlson Comorbidity Index (CCI) score,10 comorbidities were identified from all diagnoses recorded in the administrative health data for the time period between 6 and 18 months prior to an individual’s end of follow-up (i.e., date of death or emigration or 31 March 2010, whichever occurred earlier). If persons had an end date prior to 1 October 1985 (i.e., they did not have 18 months of data to assess for comorbidities) or did not have MH coverage for the period of time 18 months prior to the end date, CCI was assigned a missing value. The CCI was categorised as 0, 1, 2 or ≥3 with the reference category being 0.10 The data were descriptively analysed using means, SDs, frequencies and percentages. HRs and 95% CIs were estimated for IBD cases relative to controls using stratified multivariable Cox proportional hazards regression models; the covariates were the SEFI and CCI scores. Separate models were defined for CD and UC cases and their respective controls. For incident IBD cases only, separate models were fit to the data for deaths within 1 year of diagnosis and to the data for all deaths following IBD diagnosis and also by history of GI surgery (defined as small or large bowel resection, repair, stricture surgery or stoma creation). GI surgery was a time-dependent variable: no/presurgery, 0–29 days postsurgery, 30–365 days postsurgery and 366+ days postsurgery. Analyses were conducted for all ages (at the index/ IBD diagnosis date), as well as for the following current age groups:
