Survival comparison of patients with cystic fibrosis in Canada and the United States: a population-based cohort study

Abstract

Background

In 2011, the median age of survival in cystic fibrosis (CF) reported in the United States (US) was 36.8 years compared to 48.5 years in Canada; however direct comparison of survival estimates between national registries is challenging because of inherent differences in methodologies used, data processing techniques, and ascertainment bias.

Objectives

To use a standardized approach to calculate CF survival estimates and to explore differences between Canada and the United States (US).

Design

Population-based study.

Setting

42 Canadian CF clinics and 110 CF care centers in the US.

Patients

Patients followed in the Canadian CF Registry (CCFR) and the US Cystic Fibrosis Foundation (CFF) Patient Registry (CFFPR) between 1990 and 2013.

Measurement

Cox proportional hazards models were used to compare survival between patients followed in the CCFR (n=5,941) and CFFPR (n=45,448). Multivariable models were used to adjust for factors known to be associated with survival.

Results

Median age of survival in patients with CF increased in both countries between 1990 and 2013; however in 1995 and 2005, survival in Canada increased at a faster rate relative to the US (p<0.001). Using contemporary data from 2009-2013, the median age of survival in Canada was ten years greater than the US (50.9 vs. 40.6 years, respectively). The adjusted risk of death was 34% lower in Canada compared to the US (hazard ratio 0.66, 95% CI 0.54-0.81). A greater proportion of patients had transplants in Canada (10.3% vs. 6.5% respectively, SD 13.7). Differences in survival between US and Canadian patients varied according to the US patients' insurance status.

Limitations

Ascertainment bias as a result of missing data or non-random lost to follow-up could impact the results.

Conclusions

Differences in CF survival between Canada and the US persisted after adjusting for risk factors associated with survival. Differential access to transplantation, improved post-transplant survival and differences in the healthcare systems may, in part, explain the Canadian survival advantage.

Introduction

A comprehensive, multidisciplinary approach to medical care, early treatment of pulmonary disease and aggressive management of malnutrition has translated into longer lives for individuals with cystic fibrosis (CF).(1-3) Life expectancy depends on a patient's characteristics, as well as their access to medical care and medications, which may vary both within and between countries.(4;5) Comparisons of national CF registry data have led to important discoveries in CF disease progression and increased our understanding of epidemiologic trends in CF mortality and clinical care.(6-12) Recently, Goss et al. showed that pulmonary function among patients ages 6-25 years was higher in CF patients in the United States (US) compared to the United Kingdom (UK), which was associated with differential prescribing of inhaled mucolytic therapies between the two countries.(7)

In 2011, the median age of survival for CF in the US Annual Registry report was 36.8 years whereas it was 48.5 years in the Canadian CF Registry report.(13;14) Direct comparison of median age of survival estimates between national registry reports is problematic due to the inherent differences in methodologies used, data processing techniques, and ascertainment bias.(15;16) We recently showed that survival estimates may be over- or underestimated depending on how data are captured and processed, and the proportion of missing data.(15) Using a standardized approach to data processing and survival calculations would provide greater confidence in international comparisons and in the identification of factors that may contribute to the observed differences.

The objectives of this study were (1) to determine whether using a standardized approach to data processing and survival calculations confirmed the observed difference between Canada and the US, and (2) if a survival gap persisted, to explore potential contributing factors. We hypothesized that differences in survival between the countries could be explained by variations in clinical characteristics of the patient populations, differential rates of transplantation and variations between the health care systems.

Methods

This population-based cohort study used prospectively collected Canadian CF Registry (CCFR) and US CF Foundation (CFF) Patient Registry (US CFFPR) data from 1990 to 2013 inclusive. Each CF center that submits data to the registry obtains patient consent for data to be collected. This study was approved by the Research Ethics Board at St. Michael's Hospital, Toronto, Ontario (Research Ethics Board # 14-148) and at Seattle Children's Hospital (Institutional Research Board # 15294).

A detailed description of both registries as well as information on how individuals are accrued and monitored within each registry is outlined in the online supplement. The CCFR and the US CFFPR contain detailed demographic and clinical information on patients with a confirmed diagnosis of CF(17) receiving clinical care at accredited CF centers across Canada and the US. Data including age, sex, age of diagnosis, race, symptoms at presentation, whether or not the patient was diagnosed through newborn screening, date of transplantation, date of death and cause of death, as well as anthropometric measurements, lung function, bacteriology, CF-related complications, and pancreatic status are collected.

Each variable within the Canadian and US registries was evaluated by the research team and comparisons were made regarding data definitions and how data are collected so that a unified approach to each variable was established. Because the CCFR records the first stable clinical measurement per person per year, we selected the first stable measurement of the year for each subject within the US CFFPR in order to mirror the method of data collection. A stable measurement was defined as one that is taken from a routine outpatient clinic visit when the patient was not being treated for a pulmonary exacerbation. Missing date of diagnosis was imputed using date of birth plus 30 days whereas age at diagnosis was categorized as < 2 or ≥ 2 years of age.(18) Genotype was classified as homozygous delta F508, heterozygous delta F508, other, or missing. Race was categorized as Caucasian, other, or missing. Forced expiratory volume in one second (FEV₁) was expressed as a percentage of the predicted values for healthy age and gender matched controls using GLI reference equations.(19) Body mass index percentiles (BMI%ile) were calculated for children between two and 19 years of age using the Centers for Disease Control and Prevention growth charts.(20) Children were then classified as: underweight (BMI%ile ≤ 12%), adequate weight (BMI%ile 13-84% or overweight (BMI%ile ≥ 85%).(21) For individuals 19 years of age and older, subjects were classified into BMI categories based on World Health Organization guidelines(22) as underweight (<18.5 kg/m²), adequate weight (18.5-24.9 kg/m²), or overweight (≥ 25.0 kg/m²). Based on the literature, respiratory bacteria can have a differential effect on survival. For those individuals who grew multiple organisms from respiratory tract specimens in a given year, we categorized patients based on the organism that had the worst prognosis in terms of survival. This categorization allowed us to create mutually exclusive groups to avoid combining less virulent organisms with ones associated with a worse prognosis. Using this hierarchical classification, patients with multiple types of bacteria cultured in a given year were categorized as follows: Burkholderia cepacia complex took precedence over any other type of infection, methicillin resistant Staphylococcus aureus (MRSA) took precedence over Pseudomonas aeruginosa. P. aeruginosa took precedence over all remaining organisms. CF-related diabetes (CFRD) was defined by the clinic based on published consensus guidelines.(23) Pancreatic status was determined by pancreatic enzyme replacement therapy usage. Pancreatic status was used as a surrogate for disease severity in the multivariable analysis since CFTR genotype and pancreatic status are highly correlated(24) and fewer patients had missing data for pancreatic enzyme replacement therapy. Previous literature suggests that the majority of patients who were reported as lost to follow-up at some point during the follow-up period will return to the CF center within 2 years of non-attendance.(5) Thus, subjects were considered lost to follow-up if they had no death date recorded and their last reporting year was more than two years from the end of the study. The single-payer universal health insurance program in Canada provides comprehensive medical care to all Canadian citizens whereas the US employs a multi-payer approach to healthcare. Patients in the US were grouped into mutually exclusive categories according to their health insurance coverage as follows: chronic Medicaid or Medicare (receiving Medicaid and/or Medicare in all five years), intermittent Medicaid and/or Medicare (receiving Medicaid or Medicare in at least one but not all five years), other (any other insurance outside of Medicaid or Medicare, which is predominantly employee-provided private insurance), none/unknown insurance (none or unknown insurance unknown or in all five years) for the US patient population and universal healthcare for the Canadian population.

Statistical Analyses

Demographic and clinical variables were summarized by country, with categorical variables expressed as frequency and proportion, and continuous variables summarized as median and range. Differences between countries were compared using the Mann-Whitney test for continuous variables and the Chi-square test for categorical variables. P-values are heavily influenced by sample size, and thus the large sample sizes herein will render very small differences statistically significant. As such, the standardized difference (SD) was utilized to determine statistical significance. A SD of greater than ten was used to determine significant differences between the countries.(25)

We used period survival analysis to calculate median age of survival estimates over time. Median age of survival for each year was calculated for each five-year window beginning with 1986-1990, and ending with the 2009-2013 time period. Differences in median age of survival between the two countries over time were evaluated from 1990 to 2013 using segmented regression analysis to determine whether there were significant change points when the differences in median age of survival diverged.(26)

We further explored differences in survival by adjusting for patient and clinical characteristics using a multivariable Cox proportional hazards (PH) model using data from the most recent five years (2009-2013), which reflects a contemporary cohort of CF patients. The multivariable Cox PH analysis was first conducted on data from all patients, including those who were transplanted, and adjusted for time-independent characteristics (sex, race, age at diagnosis, pancreatic status) and transplant status. Subsequently, we conducted additional multivariable analyses which included time-varying clinical characteristics known to be prognostic for patient survival, such as FEV₁, BMI, CFRD, and microbiology. We excluded individuals under the age of 6 years because lung function measurements are not reliably measured until subjects are 6 years old. Since clinical outcomes post-transplant, particularly FEV₁, do not represent CF lung disease, this analysis censored patients at the time of transplant. Country, sex, age at diagnosis, race, genotype, pancreatic status, and insurance status were modeled as time-independent covariates. Missing lung function and BMI measurements were imputed using multiple imputation(27) (see online supplement for additional details on survival calculations and imputation methodology).

Patients with milder clinical phenotypes have better survival therefore, a differential proportion of patients with milder CF disease in the two countries would bias the survival estimates. In order to address this potential ascertainment bias, we conducted several sub-group analyses. The multivariable Cox PH models were repeated limiting the sample to patients with a severe clinical phenotype as defined by 1) pancreatic status (i.e., pancreatic insufficient subjects); 2) homozygous delta F508 subjects; and 3) subjects diagnosed before the age of two years.(17;28) To address the differences in healthcare systems between the two countries we further classified the US subjects by insurance status and compared the risk of death in these subgroups to Canada as a whole, which has universal healthcare. The insurance status classification has been used in previous studies as Medicaid and Medicare have been associated with worse outcomes in CF.(29;30) Because the hierarchical approach to microbiology used in our study has not been commonly used in registry studies, we conducted a sensitivity analysis whereby we classified pertinent subjects as having co-infections with multiple bacteria.

Role of Funding Source

This study was funded by a US Cystic Fibrosis Foundation Grant (# STEPHE14A0). The funding source provided access to the US CFF registry data and contributed to the design, analysis, and interpretation of the findings as well as critical review of the manuscript.

Results

Longitudinal Trends in Survival over Time (1990-2013)

There were 45,448 and 5,941 patients followed in the US and Canadian registries respectively between 1990 and 2013, with 9,654 (21.2%) and 1,288 (21.7%) deaths respectively. Demographic and clinical characteristics of the CF patients followed between 1990 and 2013 are included in the online supplement (Table S1). Median age of survival in both countries increased over the study period (Figure 1a). However, the rate of improvement in median age of survival over time between the two countries diverged in 1995 and 2005 (Davies Test p<0.001) (Figure 1b); most notably in 2005 when the improvements in the Canadian population were greater. The rate of change in survival between males and females differed by country (p<0.001). Canadian male survival grew at a significantly faster rate compared to US males (p<0.001) from 2005 onwards whereas there was no statistically significant difference in survival improvement over time between the two countries for females (p=0.19). Changes in the median age of death over time and the difference in median age of death between the two countries are summarized in the online supplement in Figure S1.

Figure 1. (A) Median Survival Age using a 5-year rolling window, Canada vs. USA, 1990 to 2013; (B) Difference in Median Age of Survival between Canada and the USA, 1990-2013.

Contemporary CF Population (2009-2013)

The demographic and clinical characteristics of the contemporary CF cohort are summarized in Table 1. A higher proportion of subjects were asymptomatic at the time of diagnosis in the US compared to Canada (19.9% vs. 10.9%, SD 22.4), which corresponds to a higher prevalence of newborn-screened subjects in the US (17.3% vs. 6.2%, SD 34.4). A greater proportion of patients had transplants (any organ) in Canada (10.3% vs. 6.5% respectively, SD 13.7). Seventy-eight percent of all patient deaths in the US occurred among individuals never transplanted compared to 61% of deaths in the Canada (SD 38.0). Overall, fewer Canadian patients were lost to follow-up in the registry (2.5% vs. 5.5%) and a higher proportion of the lost to follow-up patients in the US were post-transplant compared to Canada (12.2% vs. 8.0% respectively, SD 13.9). The characteristics of the patients lost to follow-up in both countries is shown in the online supplement (Table S8).

Table 1. Characteristics of contemporary CF patients in Canada and United States followed between 2009-2013.

Variable	Canada	%	United States	%	p-valuea	Std. Differenceb
N	4,662		32,699
Vital status
Censored at follow-up	4,421	94.8	30,525	93.4
Dead	241	5.2	2,174	6.6	<0.001	6.3
Death without transplant	147	61.0	1,699	78.2	<0.001	38.0
Age at death
Median (range)	31.9	(0.3-79.3)	26.9	(0.3-76.9)	<0.001	29.5
Cause of death
Cardio-respiratory	164	68.0	1,462	67.2	0.18	10.5
Other	51	21.2	578	26.6
Unknown	26	10.8	134	6.2
Transplantc
No	4,182	89.7	30,572	93.5
Yes	480	10.3	2,127	6.5	<0.001	13.7
Lost to follow-upd
Overall	112	2.5	1,679	5.5	<0.001	14.4
Males	70	62.5	956	56.9
Females	42	37.5	723	43.1
Post-Transplant	9	8.0	205	12.2	0.24	13.7
Age at diagnosis
Median (range)	0.523	(0-71.7)	0.386	(0-81.7)	<0.001	4.7
<2 years	3,108	66.7	23,191	70.9	<0.001	9.2
≥2 years	1,554	33.3	9,508	29.1
Sex
Female	2,167	46.5	15,738	48.1	0.037	3.3
Male	2,495	53.5	16,961	51.9
Race
Caucasian	4,305	92.3	30,184	92.3	<0.001	0.1
Other	213	4.6	2,515	7.7
Missing	144	3.1	0	0
Genotype
Homozygous ΔF508	2,193	47.0	14,272	43.6	<0.001	6.8
Heterozygous ΔF508	1,831	39.3	12,517	38.3	0.20	2.0
Other	531	11.4	4,427	13.5	<0.001	6.5
Missing	107	2.3	1,483	4.5	<0.001	12.4
Use of pancreatic enzymes
PS (did not use enzymes in most recent yr)	688	14.8	4,585	14.0	0.22	1.9
PI (used enzymes most recent yr)	3,974	85.2	27,977	85.6
Unknown	0	0	137	0.4
Neonatal bowel obstruction
No	3,826	82.1	26,712	81.7	<0.001	9.7
Yes	660	14.2	5,987	18.3
Unknown	176	3.8	0	0
Symptoms at diagnosis
Asymptomatic	507	10.9	6,507	19.9	<0.001	22.4
Symptomatic	3,671	78.7	25,489	78.0
Unknown	484	10.4	703	2.1
Newborn screening
No	4,291	92.0	27,044	82.7	<0.001	34.4
Yes	291	6.2	5,655	17.3
Unknown	80	1.7	0	0
Insurance status
Chronic Medicaid/Medicare	0	0	7,971	24.4	N/A	N/A
Intermittent Medicaid/Medicare Use	0	0	10,814	33.1
Other Insurance	0	0	13,131	40.2
Unspecified Insurance	0	0	783	2.4
Canadian Healthcare	4,662	100	0	0
CF-related diabetes
Yes	1,174	25.2	8,806	26.9	<0.001	4.0
No	3,488	74.8	23,893	73.1
FEV1 % predictede
Median (range)	74.03	(11.9-136.7)	74·99	(6.8-148.8)	0.35	1.4
Missing	864	18.5	6,870	21.0	<0.001	6.2
<40%	610	13.1	4,405	13.5	0.13	2.7
40-69%	1,108	23.8	7,048	21.6	0.02	4.2
≥70%	2,080	44.6	14,376	44.0	0.31	1.8
BMIf
Underweight	449	9.6	2,862	8.8	0.25	2.1
Normal Weight	2,527	54.2	16,855	51.5	0.36	1.7
Overweight	629	13.5	4,597	14.1	0.04	3.8
Missing	1,057	22.7	8,385	25.6	0.55	1.0
Microbiology
S. aureus	3,165	70.2	22,267	72.6	<0.001	5.3
MRSA	416	9.2	11,729	38.2	<0.001	72.5
P. aeruginosa	2,945	65.3	21,006	68.5	<0.001	6.7
B. cepacia complex	323	7.2	1,387	4.5	<0.001	11.3
S. maltophilia	1,316	29.2	8,996	29.3	0.87	0.3
Aspergillus	1,705	37.8	8,720	28.4	<0.001	20.1
Atypical mycobacterium	213	4.7	3,025	9.9	<0.001	19.9
Treatmentg
Mucolyticsh
Hypertonic Saline	1,141	32.3	15,540	65.1	<0.001	69.5
Dornase alfa	1,604	45.4	20,219	84.8	<0.001	90.5
Other mucolytics	31	0.9	786	3.3	<0.001	17.0
Inhaled antibioticsi
Tobi	837	38.3	10,603	68.9	<0.001	64.4
Tobi Podhaler	731	33.4	142	0.9	<0.001	95.5
Tobramycin	621	28.4	735	4.8	<0.001	67.0
Colistin	436	19.9	2,234	14.5	<0.001	14.4
Cayston	263	12.0	6,565	42.6	<0.001	73.1
Other
Azithromycin	1,153	52.7	10,877	70.6	<0.001	37.4
Number of clinic visitsj
Median (range)	3.8	0 – 33	4.2	0 – 26	<0.001	24.9
Missing	79	1.7%	629	1.9%

Abbreviations: B. cepacia complex, Burkholderia cepacia complex; BMI, body mass index; CF, cystic fibrosis; CFRD, cystic fibrosis-related diabetes; FEV₁ % predicted, forced expiratory volume in 1 second percent predicted; MRSA, methicillin-resistant Staphylococcus aureus; P. aeruginosa, Pseudomonas aeruginosa; S. aureus, Staphylococcus aureus; S. maltophilia, Stenotrophomonas maltophilia; Std Difference, standardized difference

All values are summarized based on the last recorded measurement in the time period.

^aDifferences between countries were compared using the Mann-Whitney test for continuous variables and the chi-square test for categorical variables. P-values are heavily influenced by sample size, and thus the large sample sizes herein will render very small differences statistically significant. As such, the standardized difference was utilized to determine statistical significance.

^bThe standardized difference is the mean difference as a percentage of the average standard deviations. A standardized difference greater than 10 is generally used to determine those variables that remain sufficiently different between the two countries.

^cThe patient's first transplant of any type is considered only

^dLost to follow-up is defined as patients who are alive but whose last available year of data occurs more than 2 years before the cohort end year. That is, a patient would be considered lost to follow-up if their last available reporting year is 2011 or earlier.

^ePercent predicted FEV₁ was calculated using the GLI reference values using the subject's FEV₁ value from the most recent year of follow-up.

^fBMI categories are defined using the WHO classification. The patient's BMI in the most recent year of follow-up was used.

^gAs treatment data are only available in the Canadian registry beginning in 2011, these estimates are calculated for the 2011-2013 window.

^hOut of a total of 3,530 Canadians and 23,855 Americans over the age of 6 and with no prior transplant.

ⁱOut of a total of 2,186 Canadians and 15,398 Americans over the age of 6, with no prior transplant and infected with Pseudomonas aeruginosa at some point in the 2011-2013 window.

^jThe average number of clinic visits per patient during the interval 2009 to 2013.

The median age of survival in Canada was ten years higher than the US (50.9 years (95% CI: 50.5-52.2) vs. 40.6 (95% CI: 39.1-41.8) years respectively). The median age at death was 26.9 years in the US compared to 31.9 years in Canada (SD 29.5). The risk of death between the countries was consistently lower in Canada compared to the US in multiple sub-groups of the population (Figure 3). In order to account for the differential lost to follow-up post-transplant between the two countries, we re-calculated median survival age censoring patients on the date of transplant, which effectively excluded post-transplant survival time as well as deaths after transplant in both countries. This analysis caused the survival gap between the two countries to widen further (57.1 years vs. 44.0 years in Canada and US respectively).

Figure 3. Unadjusted univariate sub-group analysis comparing the risk of death in the Canada vs. US (2009-2013).

Details regarding the creation of the study cohort used for the multivariable analysis are presented in Figure 2 and a summary of the demographic and clinical characteristics of the included patients can be found in the online supplement (Table S2).

Figure 2. Creation of Study Cohort for Multivariable Analysis (2009-2013).

In the multivariable model which adjusted for patient characteristics only, the risk of death was 41% lower in Canada compared to the US (HR 0.59, 95% CI: 0.52-0.68, p<0.001, n=37,080). (Table 2). In the model which adjusted for both patient characteristics and clinical factors, the risk of death remained significantly lower in Canada compared to the US (HR 0.66, 95% CI 0.54-0.81; p=0.002).(Table 3) There was no evidence of a statistically significant interaction between country and any of the clinical factors. When US patients were categorized according to their insurance status, Canadians had a 44% lower risk of death compared to those in the US receiving continuous Medicaid/Medicare (n=6,230, 24.3%) (HR 0.56, 95% CI: 0.45-0.71, p<0.001), a 36% lower risk compared to US patients on intermittent Medicaid/Medicare (n=8,429, 32.9%) (HR 0.64, 95% CI: 0.51-0.80, p=0.002) and a 77% lower risk of death compared to US patients with unknown or no health insurance (n=205, 0.8%) (HR 0.23, 95% CI: 0.14-0.37, p<0.001). There was no difference in risk of death between Canadians and individuals in the US with “other” insurance (n=10,754, 42.0%), including private insurance (HR 0.85, 95% CI 0.67-1.07, p=0.15). (see online supplement, Table S3) To address potential ascertainment bias, the analyses were limited to severe patient phenotypes. Canadian CF patients consistently had a statistically significant lower risk of death in all sub-groups (see online supplement, Tables S4-S6). Further, results were unchanged regardless of the method used to classify patients into microbiologic categories (See online supplement, Table S7).

Table 2. Multivariable model adjusting for patient characteristics 2009-2013 (N=37,080)^*.

Variable	HR	95% CI	p-value
Country
Canada	0.59	0.52-0.68	<0·001
United States	1·00	--	--
Race
Caucasian	1·00	--	--
Other	1·25	1·05-1·49	0·011
Sex
Males	1·00	--	--
Females	1·29	1·19-1·39	<0·001
Age at diagnosis
<2 years	1·25	1·14-1·37	<0·001
≥2 years	1·00	--	--
Pancreatic Status
Insufficient	2·12	1.79-2.50	<0.001
Sufficient	1·00	--	--
Transplant
Yes	3·21	2·89-3·56	<0·001
No	1·00	--	--

Abbreviations: HR, hazard ratio; CI, confidence interval

^*281 patients were missing either race (n=144) or pancreatic status (n=137)

Table 3. Multivariable analysis adjusted for patient and clinical characteristics (2009-2013) (N=29,209).

	Multivariable
Variable	Hazard Ratio	95% CI	p-value
Country
Canada	0.66	(0.54-0.81)	0.002
United States	1.00	--	--
Sex
Males	1.00	--	--
Females	1.40	(1.25-1.58)	<0.001
Age at diagnosis
<2 yrs	1.02	(0.90-1.17)	0.70
≥2 yrs	1.00	--	--
Pancreatic Status
Insufficient	1.18	(0.93-1.49)	0.15
Sufficient	1.00	--	--
BMI
Underweight	1.48	(1.29-1.69)	<0.001
Normal	1.00	--	--
Overweight	0.92	(0.74-1.15)	0.43
CF-related diabetes
Yes	1.67	(1.47-1.88)	<0.001
No	1.00	--	--
FEV1 % predicted Microbiology	0.939	(0.935-0.943)	<0.001
B. cepacia complex	1.64	(1.25-2.14)	0.002
MRSA	1.35	(1.09-1.66)	0.011
P. aeruginosa	1.17	(0.96-1.42)	0.11
None of the Above	1.00	--	--

Abbreviations: B. cepacia complex, Burkholderia cepacia complex; BMI, body mass index; CI, confidence interval; FEV₁ % predicted, forced expiratory volume in 1 second percent predicted; HR, hazard ratio; MRSA, methicillin-resistant Staphylococcus aureus; P. aeruginosa, Pseudomonas aeruginosa

Data were limited to patients ≥ 6 years of age with a valid FEV₁ measurement. Reference values for percent predicted FEV₁ were calculated using the GLI reference values. Patients were censored at transplant.

Discussion

Our study confirms a significant survival gap between Canada and the US in individuals living with CF. In the most recent time period (2009-2013), a ten-year survival advantage was identified for patients living in Canada which cannot be attributed to methodology or data processing techniques. The risk of death was consistently lower in Canada after accounting for multiple patient demographic and clinical factors, and persisted in patients with severe clinical phenotypes. The systematic approach used to compare data from these two countries demonstrates that national disease registries can be meaningfully compared and can elucidate factors associated with differences in survival. While we cannot draw definitive causal inferences from these analyses, the observed differences raise the question of differences in access to therapeutic approaches or health care delivery, such that hypotheses can be generated for further investigation.

Although our study was not designed to specifically evaluate the impact of transplant on survival, few therapeutic interventions have the ability to impact survival in a short time period, and several of our findings supported the hypothesis that transplantation may be contributing to the observed survival gap. Firstly, we found that more Canadians with CF received a transplant in the contemporary study period compared to US CF patients despite having similar proportion of patients with severe lung disease based on FEV₁. Secondly, a higher proportion of deaths in the US occurred in patients who had not received a transplant compared to the Canada. Thirdly, when we censored patients at transplant, the median age of survival increased from 10 years to 13 years suggesting transplant contributes to the survival gap. And lastly, we observed a significant widening of the survival gap between the countries since 2005, a year that corresponds to a marked change in how US CF patients were prioritized for lung transplant with the implementation of the lung allocation score (LAS) in May 2005, which is not used in Canada.(31)

Existing literature further supports the possible role that transplant plays in the Canada-US survival gap. For instance, median survival post-lung transplant is longer in Canadian CF patients compared to published studies of US CF patients.(32) (33) Although the LAS has been associated with shorter waiting times and improved transplant survival(35), and the number of patients with CF listed for lung transplantation has increased since its implementation, the proportion who have received a transplant has decreased.(36) In addition, Merlo et al showed that CF patients transplanted since the LAS was implemented had a higher risk of death compared to CF patients transplanted in the pre-LAS period.(35) These studies suggest that differential access to lung transplant and survival post-transplant between Canada and the US may contribute to the survival gap although this hypothesis requires further study.

Although transplant may play a role, there are likely other factors that are contributing to the survival gap seen. One such factor may be the substantial difference in healthcare delivery between Canada and the US. In Canada, universal healthcare provides all individuals with the opportunity to access medical services across the country; in the US, differing health insurance coverage impacts access to medical care and costs of care and is related to socioeconomic status (SES).(30;33) Indeed we observed health insurance status in the US had a differential effect on the risk of death for individuals living with CF compared to Canada. Interestingly, there was no statistically significant difference in risk of death among Canadians and those in the US with “other” insurance (which was comprised primarily of individuals with private coverage). We were unable to draw firm conclusions on the impact of SES on the survival gap since the Canadian registry did not capture SES data. While all Canadians have access to healthcare regardless of SES, this does not imply that SES does not impact survival in Canada. There are hidden financial factors that may limit the ability of patients to seek medical attention in Canada, for example, the costs associated with taking time off work, driving long distances to be seen at a CF center, or having to relocate to a CF transplant center. The fact that the risk of death in the US varied by the type of health insurance raises the possibility that differences in the health care systems may explain part of the survival gap documented; however, further study is needed to formally evaluate this question.

Previous comparisons between Canada and the US in the 1980's suggested that implementation of aggressive nutritional support in the early 1970s in Toronto led to improved survival compared with Boston.(6) As a result, the high-fat diet was adopted worldwide for individuals with CF. Those born in Canada in the 1970s would have been exposed to aggressive nutritional supplementation from birth and the impact of early nutrition would have reduced their risk of death in their later teen and early adult years. As such, one would expect that US survival would catch up to Canada since CF centers in the US also implemented the high calorie diet, albeit 5-10 years later.

The strengths of our study include the large sample size, the longitudinal data within both CF registries, the consistency of our results across multiple sub-groups and the unified approach to the analysis. There is a high participation rate in both registries and CF clinics receive financial incentives for data submission which results in a comprehensive national picture of the CF population. However, missing vital statistics could result in biased survival estimates, particularly if there was a disproportionate number of missing events in one country compared with the other.(15) Prior literature suggests death dates are missing in the US CFFPR, particularly in those over the age of 45 years, when compared to national vital statistics;(37) whereas the number of deaths in the CCFR is similar to Canadian vital statistics.(5) If the US data were preferentially missing deaths, this would result in an overestimation of the true US CF survival. Our data also show that there were more US patients lost to follow-up compared to Canada, which would have the opposite effect on survival estimates (typically having more patients lost to follow-up results in an underestimation of median survival).(15) It is also worth noting that more of the US lost to follow-up patients had characteristics associated with a worse prognosis for example, female sex, non-Caucasian race, pancreatic insufficiency, CFRD etc. (Table S8) compared to Canada. By preferentially excluding US patients with a worse prognosis from the survival analysis, the survival gap may actually be wider than reported in our study. Differential inclusion of patients with a mild clinical phenotype in the registry may result in a survival gap simply because one country captures patients with mild disease compared to another country. The fact that our findings were consistent across multiple subgroups, including those that have a severe clinical phenotype, suggests that higher survival in Canada relative to the US is not due to following milder patients in the Canadian cohort. Differences in medication use may play a role in the survival gap; interestingly, between 2011 and 2013 medication use was higher in the US, however we were unable to examine medication for the full study period due to lack of medication data prior to 2011 in Canada. While environmental factors such as air pollution or climate may impact lung health in CF, these data were not equally available within the registries (air pollution data have only been linked to the US CFFPR). Newborn screening (NBS) programs have been in place longer in the US compared to Canada however this is unlikely to significantly affect the current survival estimates as there are relatively few pediatric deaths in both countries. Furthermore, having more extensive NBS programs in the US would bias the results towards the null as these individuals would presumably be healthier. Regional or center-specific death rates between the countries were not conducted due to relatively few deaths within any given center which would produce unreliable survival estimates.

In conclusion, we show a difference in survival between Canada and the US even after adjusting for patient characteristics and clinical factors known to be associated with survival. The survival differences appear to diverge at two distinct time points, most notably in 2005, when the gap between the two countries widens. Differential access to transplantation related to the referral or donor lung allocation process, variable post-transplant survival and differences in the healthcare systems may, in part, explain the Canadian survival advantage. Further studies are necessary in order to specifically test these hypotheses.