If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
School of Medicine University College Dublin, Belfield, Dublin 4, IrelandChildren's Health Ireland at Crumlin, Crumlin, Dublin 12, IrelandKing's College Hospital NHS Foundation Trust, Denmark Hill, London, UK
School of Medicine University College Dublin, Belfield, Dublin 4, IrelandChildren's Health Ireland at Crumlin, Crumlin, Dublin 12, IrelandKing's College Hospital NHS Foundation Trust, Denmark Hill, London, UK
Corresponding author. Marion Rowland, School of Medicine, University College Dublin, Catherine McAuley Research Centre, Nelson Street, Dublin D07 A8NN, Ireland.
Liver Disease (CFLD) is a complication of Cystic Fibrosis (CF).
•
There is uncertainty about the impact of CFLD on mortality in CF.
•
This systematic review demonstrates that CFLD shortens life expectancy in CF.
•
Pulmonary complications are the commonest cause of death for persons with CFLD.
•
Mortality rates for persons with CFLD have changed very little over the last 30 years.
Abstract
Background
There is conflicting evidence on the impact of liver disease (CFLD) on life expectancy in CF. Therefore the aim of this systematic review was to evaluate the impact of liver disease (CFLD) on mortality in CF.
Methods
The protocol was published at (https://hrbopenresearch.org/articles/3-44/v3) using PRISPMA-P guidelines and registered in Prospero 2020 (CRD42020182885). Three databases were searched for publications (1938-2020) where the outcome was all-cause mortality (defined as death and transplantation) or CF-specific mortality in participants with CFLD. Studies with and without a comparator group were included. Studies were divided into 2 groups based on the definition of CFLD: Group 1 used 2 categories of liver disease (i) liver disease with portal hypertension (PH) (ii) non-specific abnormalities which did not meet the criteria for PH, Group 2 studies only included participants with PH.
Results
All 14 eligible studies were observational, with a moderate-high risk of bias, Six of the 14 studies directly compared mortality between those with CFLD and those with no liver disease, and 5/6 demonstrated that those with CFLD had at least 3 time the risk of death compared to those with no liver disease. Pulmonary complications were the primary cause of death.
Conclusion
This SR demonstrates that liver disease shortens life expectancy in CF, and that pulmonary complications are the primary cause of death in those with CFLD. There has been no improvement in survival for persons with CFLD despite significant improvements in life expectancy for persons with CF who have no evidence of liver disease
Liver disease as a complication of CF (CFLD), was first described by Andersen in 1938 when biliary cirrhosis was reported in 3 of 44 (6.8%) post mortem examinations of children with cystic fibrosis (CF).[
Heterogeneous liver on research ultrasound identifies children with cystic fibrosis at high risk of advanced liver disease: interim results of a prospective observational case-controlled study.
] The identification of patients with biochemical or nonspecific ultrasound changes who will develop progressive liver disease with portal hypertension in CF remains a challenge.
There is an ongoing debate as to what clinical criteria are required to make a diagnosis of CFLD. Some suggest that CFLD should only be diagnosed when there is clinical or radiological evidence of portal hypertension.[
] consensus guidelines suggest that there are two categories of liver disease: (i) those with clinically significant liver disease as evidenced by PH (CFLD) and (ii) moderate or non-specific changes on ultrasound or clinical biochemistry which do not meet criteria for portal hypertension (NSCFLD).
Currently there is conflicting evidence as to whether liver disease shortens life expectancy in CF with some studies suggesting that liver disease does not impact on mortality in CF [
]. If there is to be further improvement in life expectancy of people with CF (PWCF), it is important to document the impact of liver disease on survival and understand the risk factors which contribute to reduced life expectancy in CF.
1.1 Objectives
To date there has been no systematic review on the impact of CFLD on life expectancy in CF. In this study we aim to (1) evaluate the impact of liver disease on mortality in cystic fibrosis and (2) provide clarity as to the cause of death for PWCF with liver disease.
1.2 Specific objectives
i
Examine the crude and population based mortality rates for participants with liver disease ((CFLD and/or non-specific liver disease (NSCFLD)) with or without a comparison group of PWCF with no evidence of liver disease (NoLD).
ii
Determine if CFLD/NSCFLD contributes to excess mortality in CF.
iii
Determine if the outcome for CFLD has improved with advances in management of CF.
iv
Document the specific causes of death where reported including liver, pulmonary, other CF, non-CF related causes of death as well as organ transplantation (liver, lung, liver and lung, lung and other).
v
Identify risk factors which may influence the mortality for patients with CFLD/NSCFLD such as age, gender, nutritional status, pulmonary function, cystic fibrosis related diabetes (CFRD), ursodeoxycholic acid (URSO) treatment and meconium ileus (MI).
2. Methods
A protocol for this SR in compliance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-P) [
We included all observational studies examining the outcome for liver disease in CF in adults or children or both in any clinical care setting between January 1st 1938 and December 31st 2020. Inclusion criteria required that publications defined the CF population from which the study sample was drawn, reported the prevalence of CFLD (and if included NSCFLD, see Section 2.4 for definition), or the prevalence could be extrapolated or was provided by the lead investigator, and that mortality was an outcome measure for the study. Studies using CF Registry data were included if they met the inclusion criteria. Studies with or without a comparison group were included.
2.2 Outcome
The outcome metric for this SR was all-cause mortality. All-cause mortality was defined as death from any cause including, hepatic, pulmonary, other CF-related causes, non-CF related causes and organ transplantation. The number of patients undergoing transplantation (liver or lung) was combined with the number of patients who died from CF and non-CF causes because in the absence of transplantation, death due to native organ failure is the expected outcome. We also examined specific mortality rates for hepatic, pulmonary, other causes of death as well as transplantation (liver, lung or multiple organs) rates in those with liver disease.
2.3 Exclusion criteria
We excluded studies using the following criteria:
•
Studies without a clearly articulated definition of CFLD
•
Studies which examined the outcome for multiple causes of liver disease in children or adults (e.g. biliary atresia, alcoholic liver disease)
•
Letters and short case series where the reported number of participants with CFLD was <5.
•
Randomised and non-randomised clinical trials of ursodeoxycholic acid for the management of CFLD
•
Studies which described the outcome for transplanted organs (liver, lung, lung and liver) based on transplant registry data, and which did not describe the population of PWCF or CFLD on which the study was based.
•
Older studies of autopsy data, which do not describe the population of PWCF or CFLD on which the autopsy results are based.
2.4 Classification of CFLD studies
The criteria and the terminology used to define liver disease in CF are not standard, and potentially could lead to classification bias when examining outcome in this SR. To reduce the risk of bias we have divided studies included into two groups as follows:
Group 1
Included in Group 1 are studies which are broadly in line with the definition used by Colombo [
]. Group 1 studies comprise 2 categories of liver disease (i) participants with clinically significant liver disease, defined by clinical or radiological evidence of portal hypertension (PH) for which we use the term CFLD and (ii) participants with non-specific liver changes clinically, ultrasonographically or biochemically which do not meet the criteria for PH and for which we use the term non-specific liver disease (NSCFLD). For those with NSCFLD we did not set cut-offs for radiological or biochemical abnormalities rather NSCFLD is defined when participants cannot be considered to have no evidence of liver disease yet do not meet criteria for PH.
Group 2
Studies assigned to Group 2 define only one category of liver disease. Participants in these studies meet clinical and or radiological criteria for PH with or without evidence of hypersplenism. We use the terminology CFLD also to describe this group, as the participants meet the same criteria for PH as those with CFLD in Group 1 studies. For the remainder of this SR we use CFLD and NSCFLD according to the definitions outlined above.
2.5 Information sources
AS and DS developed the literature search strategy using words related to cystic fibrosis liver disease and outcome, all-cause mortality and liver specific mortality. The following databases were searched PubMed, Embase and Web of Science from 1938 to December 31 2020. Proceedings of the North American Cystic Fibrosis Conference and European Cystic Fibrosis Conference between 2000 and 2019 were searched, and any abstracts published as full manuscripts were included. Bibliographies of published articles and previous reviews were hand-searched. Only papers published in English with full-text accessibility were included.
2.6 Search strategy
The following search strategy terms were used which were adapted with relevant thesaurus terms for each database: (("Cystic Fibrosis" OR "Mucoviscidosis") AND ("Liver Disease" OR "Liver Dysfunction" OR "Hepatic Diseases" OR "Portal vein hypertension" OR "Portal hypertension" OR "Portal congestion" OR "Liver cirrhosis" OR "Hepatic cirrhosis" OR "Liver fibrosis" OR "Hepatic fibrosis" OR "Liver disorder" OR "Hepatic disorder" OR "Liver illness" OR "Liver failure" OR "Hepatic failure" OR "Cystic Fibrosis Liver Disease" OR "Cystic Fibrosis Associated Liver Disease" OR "Cystic Fibrosis-Associated Liver Disease" OR "Cystic fibrosis-related Liver disease" OR "Cystic fibrosis related liver disease" OR "CF-related Liver Disease" OR "CF-associated liver disease" OR "CFLD" OR "CFALD" OR "CFRLD)) AND ("Mortalit*" OR "Death" OR "Death Rate" OR "Survival Rate" OR "Survival Time" OR "Survival Probability" OR "Mean Survival" OR "Cumulative Survival" OR "Fatality" OR "Fatality Rate" OR "Case Fatality Rate" OR "Fatal Outcome" OR "Lethal Outcome").
No study design, was imposed on the search strategy. The search strategy was validated by BB with assistance from DS to ensure that the strategy retrieved a high proportion of eligible studies.
2.7 Study selection
Results from all three electronic databases (n=543) were compiled in Endnote™ where duplicates and reviews were removed (n=394). These publications were exported to Rayyan QCRI, [
] a free web-based application for initial screening of abstracts and titles. Four authors (AS, MR, EF, LC) independently screened the titles and abstracts against the inclusion and exclusion criteria. Any conflicts or uncertainties were reviewed and a consensus reached. This provided a preliminary list of 27 studies for full-text review.
Publications from the same CF centre/s or national CF database which examined a similar population/cohort and used a previously reported definition of CFLD and/or outcome measure received robust scrutiny to prevent the possibility of double counting, (excluded n=2) [
Data extraction from the eligible studies was conducted independently by 3 reviewers (AS, EF, MR) using a summary of evidence table(DOI:10.5281.zenodo.org.4032408). The table was developed with the involvement of all the reviewers and BB, was piloted for ease of completion, clarity, scope and interpretation of the variables included, and was modified following suggestions by reviewers of the protocol. No evaluation of observer variation was conducted.
2.9 Data items
The following data were extracted: bibliographic information, study aims and design, time frame, study population, the classification for CFLD used with the assignment of each publication to Group 1 or Group 2 (see above), the prevalence of CFLD/NSCFLD reported in the study or as supplied by investigators [
], the exclusion criteria used in the original study with an emphasis on how liver, lung or lung and liver transplantation was assigned, risk factor data for mortality including age, gender, genotype, height, weight, body mass index, pulmonary function, meconium ileus (MI), treatment with ursodeoxycholic acid and presence of CF-related diabetes (CFRD), and the criteria used to define CFRD.
The primary outcome for this study was the mortality in persons with CFLD in comparison to PWCF with no evidence of liver disease if a comparator group was included in the study. We examined baseline characteristics and exclusion criteria used in study populations to ensure that participants who received a transplant (liver or lung) prior to baseline were not included in the analysis of mortality rates. Due to the variability in how data are reported in different studies, we collected crude mortality rates (percentages), mortality rates as a function of follow-up time (clinical life table) median survival age and age at death if reported for both CFLD, NSCFLD and for the NoLD comparator group if included. Where possible we extrapolated excess mortality rates as the risk difference between those with CFLD and the comparison group with no liver disease
All-cause mortality was defined as the sum of deaths and transplants (liver, lung, heart). We extracted specific mortality rates for pulmonary, hepatic and other causes of death in those with CFLD. We extracted data on the number of reported transplants (liver, lung, heart, multi organ) but did not examine mortality post transplantation.
We examined risk factors for mortality where reported and tabulated whether the factors were protective or risk factors for mortality without assigning any numerical value to the risk. This was due to the heterogeneity of the studies and the lack of any standardised selection or reporting of risk factors.
We only considered mortality in those who met standard criteria for the diagnosis of CF https://hrbopenresearch.org/articles/3-44/v3. Composite, soft or patient reported outcomes were not considered. The language in the review uses the words mortality and outcome interchangeably.
2.10 Risk of bias in individual studies
The risk of bias for most studies of CFLD is significant for many reasons including small sample size, short duration of follow-up, the variability in classification of CFLD. There are a number of risk of bias assessment tools available, and while we considered a number of potential tools the Appraisal Tool for Cross-Sectional Studies (AXIS) [
] provided the most transparent assessment of Risk of Bias for observational studies included in this SR DOI:10.5281.zenodo.org.4032408.
Three reviewers (AS, EF, MR) scored the studies independently for risk of bias after which we came together to discuss the results. For any discrepancies an independent fourth person, BB was consulted. No study was excluded due to risk of bias. Studies were simply ranked based on their risk of bias as high, medium or low (Tables 1 and 2) with details of specific risks outlined in Supplementary Table 2.
Table 1Details of studies in Group 1 with the prevalence of CFLD and rates of all-cause mortality.
Study Details First Author Year Publication Study Design Study Setting Diagnostic Criteria
Estimates adjusted based on authors communication as only 50% of CFLD patients included in the liver biopsy study. Mortality estimates based on numbers reported in study.
Estimates adjusted based on authors communication as only 50% of CFLD patients included in the liver biopsy study. Mortality estimates based on numbers reported in study.
Estimates adjusted based on authors communication as only 50% of CFLD patients included in the liver biopsy study. Mortality estimates based on numbers reported in study.
Estimates adjusted based on authors communication as only 50% of CFLD patients included in the liver biopsy study. Mortality estimates based on numbers reported in study.
Estimates adjusted based on authors communication as only 50% of CFLD patients included in the liver biopsy study. Mortality estimates based on numbers reported in study.
Estimates adjusted based on authors communication as only 50% of CFLD patients included in the liver biopsy study. Mortality estimates based on numbers reported in study.
Estimates adjusted based on authors communication as only 50% of CFLD patients included in the liver biopsy study. Mortality estimates based on numbers reported in study.
2001 Retrospective France: Adult CF Unit Colombo Criteria
1991 - 2004
3.9 yrs
365 285 (78)
23 (8.1)
67 (23.5)
195 (68.4)
9/23 (39.1)
NR
NR
HR 1.66 (1.1-2.3 P= 0.01
M-H
Abbreviations CFLD Cystic Fibrosis Liver Disease, NSCFLD Non-specific Liver Disease, NoLD No Liver Disease, HR Hazard Ratio, RR Rate Ratio, NR not reported. H High, M Medium, L Low M-H Medium to High, M-L Medium to Low
Explanatory notes
a Diagnostic criteria as reported by the authors.
b Median and range unless otherwise stated.
c CF population from which sample is drawn.
d Risk of Bias: High, Medium and Low using AXIS risk of bias tool
e Estimates adjusted based on authors communication as only 50% of CFLD patients included in the liver biopsy study. Mortality estimates based on numbers reported in study.
f Steatosis and splenomegaly groups included in NSCFLD category
The Dutch CF Registry was used to generate a comparator group of participants over 8 years of age with no liver disease while those with Cystic Fibrosis Cirrhosis (CFC) were attending CF care centre. To estimate all-cause mortality participants who were reported to have had a lung transplant at baseline (CFC n = 3, NoLD n = 44) were excluded. Liver transplant recipients at baseline were excluded by the authors in their calculations.
The Dutch CF Registry was used to generate a comparator group of participants over 8 years of age with no liver disease while those with Cystic Fibrosis Cirrhosis (CFC) were attending CF care centre. To estimate all-cause mortality participants who were reported to have had a lung transplant at baseline (CFC n = 3, NoLD n = 44) were excluded. Liver transplant recipients at baseline were excluded by the authors in their calculations.
The Dutch CF Registry was used to generate a comparator group of participants over 8 years of age with no liver disease while those with Cystic Fibrosis Cirrhosis (CFC) were attending CF care centre. To estimate all-cause mortality participants who were reported to have had a lung transplant at baseline (CFC n = 3, NoLD n = 44) were excluded. Liver transplant recipients at baseline were excluded by the authors in their calculations.
The Dutch CF Registry was used to generate a comparator group of participants over 8 years of age with no liver disease while those with Cystic Fibrosis Cirrhosis (CFC) were attending CF care centre. To estimate all-cause mortality participants who were reported to have had a lung transplant at baseline (CFC n = 3, NoLD n = 44) were excluded. Liver transplant recipients at baseline were excluded by the authors in their calculations.
Cumulative death rate reported of 39.2% 95% CI 30.8-46.6. Deaths post-transplant not included. Number of lung transplants difficult to determine with accuracy (possibly n = 5)
National case control study with follow-up. Prevalence of CFLD is based on national population of CF under 16 years of age, while comparator is an age and gender matched comparator group. National prevalence of NoLD is estimated at 92.7% (personal communication)
National case control study with follow-up. Prevalence of CFLD is based on national population of CF under 16 years of age, while comparator is an age and gender matched comparator group. National prevalence of NoLD is estimated at 92.7% (personal communication)
National case control study with follow-up. Prevalence of CFLD is based on national population of CF under 16 years of age, while comparator is an age and gender matched comparator group. National prevalence of NoLD is estimated at 92.7% (personal communication)
Cumulative death rate reported of 39.2% 95% CI 30.8-46.6. Deaths post-transplant not included. Number of lung transplants difficult to determine with accuracy (possibly n = 5)
f The Dutch CF Registry was used to generate a comparator group of participants over 8 years of age with no liver disease while those with Cystic Fibrosis Cirrhosis (CFC) were attending CF care centre. To estimate all-cause mortality participants who were reported to have had a lung transplant at baseline (CFC n = 3, NoLD n = 44) were excluded. Liver transplant recipients at baseline were excluded by the authors in their calculations.
g Cumulative death rate reported of 39.2% 95% CI 30.8-46.6. Deaths post-transplant not included. Number of lung transplants difficult to determine with accuracy (possibly n = 5)
h National case control study with follow-up. Prevalence of CFLD is based on national population of CF under 16 years of age, while comparator is an age and gender matched comparator group. National prevalence of NoLD is estimated at 92.7% (personal communication)
i Personal communication. An estimate of total CF population is not possible but prevalence of CFLD is estimated at 4.8%
Fig. 1 outlines the selection process for the 14 included studies from the initial 543 studies identified by the search strategy. All studies were published between 1976 and December 2020. No study prior to 1976 met our inclusion criteria.
As described in the Methods (Section 2.4) studies were divided into groups based on their classification of liver disease with 6/14 (42.8%) studies were classified as Group 1 [
] in which liver disease was categorized as (i) liver disease with evidence of portal hypertension (CFLD) or (ii) non-specific changes which do not meet the classification for portal hypertension (NSCFLD). Toledano [
] of the studies were based on patient populations attending specialist referral CF centers. The prevalence rates for CFLD ranged from 3.9-19%, while the prevalence of NSCFLD was 3.2-33.8%
The crude mortality rate ranged from 5.8% to 47.1% (Table 1). Only 2/6 studies in Group 1 provided comparative mortality rates for those with NoLD, and they are conflicting [
] using the UK CF registry data reported that the combined mortality rates for those with cirrhotic and non-cirrhotic liver disease was higher than those with no evidence of liver disease (Table 1). In contrast, Colombo et al concluded that liver disease does not impact on the outcome for patients with CFLD using data from a paediatric tertiary referral center. [
While all studies in Group 1 divided those with liver disease into 2 categories only one study compared mortality between participants with CFLD and NSCFLD. [
] Lewindon et al reported that 2/23 (8.7%) of children with NSCFLD died or received a transplant compared to 8/17 (47.1%) children classified as CFLD. [
] However, this is a highly selected patient population from a tertiary pediatric hepatology center, without a comparison group of PWCF with no evidence of liver disease.
Group 2
Studies in Group 2 examine the outcome for PWCF with CFLD and who have clinical or radiological evidence of liver disease with PH. Criteria for the diagnosis of CFLD in this group are broadly similar in all studies, except the study of Ye et al [
] which relied on a NACCF registry-based definition without the support of clinical, radiological or biochemical features of CFLD. When the data from Ye is excluded [
], the estimate of the overall prevalence of CFLD in Group 2 studies is less than 10%. A number of studies in Group 2 were conducted before the introduction of routine ultrasonography and rely primarily on clinical examination which, although unlikely, may underestimate the prevalence of liver disease [
Five of the eight (62.5%) publications included in Group 2 compare mortality in those with CFLD to those with no liver disease. These studies demonstrate that those with CFLD have more than three times a risk of death compared to those with no liver disease.
A prospective study that followed children diagnosed following newborn screening (1986-2007) provides evidence that PWCF with CFLD continue to have much higher all-cause mortality rate (mortality/transplant) compared to those with no liver disease (excess mortality 18.8% 95%CI 6.9-30.6). [
] Using data based on the Dutch CF Registry which included children over 8 years of age and adults, Pals et al demonstrated that the risk of death was greatest in those under 25 years of age. [
] in the second decade of the 21st century. While there has also been an improvement in mortality rates for those with CFLD since the 1980’s, clinically important differences in excess risk remain. The excess mortality (risk difference) suggests the possibility of an increasing disparity in outcome. In 1991 Scott Jupp [
] reported that the excess risk of mortality was 10.1% (95%CI 1.7-18.4), while in 2020 Cipolli et al reported an excess mortality rate of 18.78% (95%CI 7.9-30.56) [
] (Table 2) for those with clinically significant liver disease. This excess mortality in those with CFLD is also supported by evidence from other studies included in this SR [
In this study, in addition to examining all-cause mortality (combined mortality and transplant) we also examined specific mortality rates for hepatic, pulmonary, other causes of death including the number of transplants (liver, liver and lung, lung and/or other). Apart from in the study of Cipolli et al [
] pulmonary complications accounted for mortality in more than 50% of those with CFLD who died. It is difficult to draw any conclusions from Table 3 as to the role of transplantation (liver, lung) in CFLD management. Change in clinical practice in recent years has increased the rate of liver transplantation as a therapeutic intervention in CFLD.
Table 3Cause of death or organ transplanted in those with CFLD.
Extrapolated from included papers- there may be some small discrepancies in numbers presented due to differences in categorisation of transplant as death (see methods). No deaths following transplant are included.
To estimate all-cause mortality participants who were reported to have had a lung transplant at baseline (CFC n = 3, NoLD n = 44) were excluded. Liver transplant recipients at baseline were excluded by the authors in their calculations.
CFLD Cystic Fibrosis Liver Disease, NR not reported
Explanatory Notes
a Deaths and Transplant (all-cause mortality) in those with liver disease. Number of deaths as a percentage of participants with CFLD
b Extrapolated from included papers- there may be some small discrepancies in numbers presented due to differences in categorisation of transplant as death (see methods). No deaths following transplant are included.
c Extrapolated from data reported in included paper
d Pulmonary Causes
e Other CF related causes and non-Cf causes of death
f Liver and Lung Transplantion
g Lung, Heart or Heart and Lung Transplantion
h Only the specific cause of death is reported for 8 of the 200 participants with CFLD who died during follow-up
i Cause of death reported for 40 study participants.
j Includes deaths in CFLD and NSCFLD
k Combines causes of death for those with CFLD and NSCFLD.
l To estimate all-cause mortality participants who were reported to have had a lung transplant at baseline (CFC n = 3, NoLD n = 44) were excluded. Liver transplant recipients at baseline were excluded by the authors in their calculations.
m Deaths post-transplant not included. Number of lung transplants difficult to determine with accuracy (possibly n = 5)
Most studies have not reported the contribution of other risk factors for mortality in CF such as age, gender, pulmonary function. Those that did examine other independent risk factors for mortality are summarized in Table 4. However, because different studies have reported on a limited number of risk factors, often without taking into account the confounding effects of age and gender, we only analyze those factors that were examined and whether there was an independent association (positive or negative) with mortality while controlling for the impact of CFLD on mortality.
Table 4Independent risk factors for mortality examined in association with CFLD.
Supplementary Table 2 provides specific details of the risk of bias for individual studies while in Table 1 and 2 we have provided a summary measure of the risk of bias for the included studies. Issues including details of participant selection, the definition of CFLD used, and the lack of a comparison group all contribute to a moderate to high risk of bias for most studies. No study was excluded because of risk of bias.
4. Discussion
Research in any rare disease is a challenge and this SR highlights some of the complexities in performing outcome studies in the field of cystic fibrosis liver disease. The heterogeneity of the studies also presents difficulties for a narrative synthesis. Despite these constraints, we demonstrate that CFLD, defined in this SR as liver disease with clinical or radiological evidence of PH, contributes to reduced life expectancy in CF, with estimates suggesting that those with CFLD have more than 3 times the risk of death compared to those with no liver disease (Table 2). Furthermore, we demonstrated that the most frequent cause of death in those with CFLD was end stage pulmonary disease or lung transplantation which may explain why liver disease is not perceived as a life shortening complication of CF. [
] While only 4 studies in this SR examined the independent relationship of CFLD with other established risk factors for mortality in CF including gender pulmonary function and cystic fibrosis related diabetes (CFRD), all concluded that liver disease was an independent risk factor for mortality in CF [
Support for a causal role for liver disease in reduced life expectancy in CF, comes from the consistent findings in studies across a diverse range of settings. Studies in this SR have been conducted in different CF populations, including specialist hepatology centres as well as CF disease specific registries, using different study designs, with a range of different classifications of liver disease. Of the 14 studies included all but one [
] point to a causal role for liver disease in reducing life expectancy in CF. There are a number of factors which may help to explain the lack of association between CFLD and mortality reported by Colombo et al. [
] In particular the age of participants with CFLD at baseline (median 4 months range 0-49) in this study was younger than those who did not develop liver disease (median 13 months range 0 – 277) (p<0.001).[
] In addition the median duration of follow up of 8.5 years for those with CFLD was too short to determine the impact of CFLD on outcome given that many of these children were under 10 years at the end of follow-up.[
] In contrast, when PWCF diagnosed by newborn screening were followed over a longer period, the excess mortality was estimated at 19 (95% CI= 7-30) per 100 participants with CFLD in comparison to those with no liver disease.[
] highlight the importance of including late adolescents and early adulthood in examining outcome for CFLD. They reported that mortality associated with CFLD is highest in those under 25 years of age.[
We assumed this SR would clearly demonstrate that mortality rates associated with CFLD would improve in line with overall improved life expectancy in CF. As this is the first SR on CFLD we did not set exclusion criteria for year of publication as this was open to the risk of bias, but instead organised data in the Tables by year of publication. While early studies are likely to underestimate the prevalence of CFLD due to a lack of screening diagnostics, it is remarkable that the reported prevalence of CFLD, has remained consistently under 10% in most studies since the 1980’s (Table 2). It is noteworthy that our data point to a lack of any discernible improvement in the burden of CFLD despite dramatic advances in overall survival in CF. A prospective Italian-Australian study [
] using newborn screening to diagnosis CF, clearly shows that while mortality for PWCF with no liver disease has greatly improved compared to earlier studies, [
] it is clear that the benefits of newborn screening and improved CF care has not benefited those with CFLD to the same extent. The reason is uncertain but suggests factors other than nutritional interventions or variceal haemorrhage management contribute to reduced life expectancy of those with clinically significant liver disease.
This study highlights the need for standardized terminology in any international consensus of diagnostic criteria for liver disease in CF. To reduce the impact of classification bias on mortality in this SR we adopted the following terminology and allocated included studies to one of two groups (Section 2.4). The term CFLD was used to differentiate participants with evidence of PH (clinical or radiological) and by definition all had clinically significant liver disease. For patients with non-specific abnormalities of clinical biochemistry or ultrasonography who did not meet the criteria for PH, and who could not be considered to have no evidence of liver disease we used the term NSCLFD. We did not set any criteria for the upper limit of normal (ULN) of liver biomarkers or ultrasound abnormalities. Those with CFLD in Group 1 and Group 2 are comparable in that all these participants had clinical or radiological evidence of PH. As previously noted all but one [
] and include participants with NSCFLD as a separate category of liver disease are allocated to Group 1 (Table 1). These studies may or may not have a comparator group of PWCF with no evidence of liver disease. However, as can be seen from Table 1 reporting of incomplete data is frequent, while comparing outcomes between groups is complex and requires greater methodological expertise.
However, because at least 20% of PWCF have persistent non-specific clinical, biochemical, or radiological evidence of liver disease but without evidence of portal hypertension, there must be transparency around how this group of PWCF is handled in any analysis of outcome. All approaches have an impact on the prevalence and mortality rates of liver disease in CF. Comparing those with portal hypertension to those in the full population of PWCF as was done by Cipolli et al [
] is likely to underestimate differences in mortality, while excluding them from the data inflates the prevalence of CFLD. Data reported by Lewindon et al [
] appears to suggest that biopsy proven NSCFLD is a portent of a worse outcome, an observation that would have been strengthened had a comparator group been included in the study.
The complex nature of clinical care over the lifetime of PWCF makes collecting complete outcome data difficult. The majority of PWCF with CFLD die from pulmonary causes or receive a lung transplantation. Death due to hemorrhage or synthetic liver failure is uncommon. [
] as registries are in a position to capture long-term outcomes including all-cause mortality, disease specific mortality and transplantation in CF. However, the difficulties of applying current diagnostic criteria for CFLD to registry data is obvious from the three registry-based studies included in this SR. A standardised approach, as outlined above would facilitate national and international comparisons studies using registry-based data. Furthermore, such an approach would greatly facilitate the use of Registry data to examine the impact of modulators or other therapies in PWCF with both CFLD and NSCFLD. As transplantation (liver and/or lung) becomes standard practice in the management of CFLD registries are best placed to examine the long-term impact of transplantation on survival.
We acknowledge that there are limitations to this review. It would have benefited from an analysis of the raw data, but for many reasons including the heterogeneity of studies in terms of definition of CFLD, and the timeframe over which studies took place made this impractical. We could have applied more stringent inclusion criteria, limited studies to those with an appropriate comparator group, and excluded studies from specialist hepatology centers but consider that because this is the first systematic reviews in the field a broader approach was more informative.
] proposed that liver disease should be included in any prognostic model of survival in CF. This SR highlights that despite advances in the care of PWCF, liver disease reduces life expectancy in CF. Pulmonary complications, rather than liver related events are the primary cause of death. The burden of CFLD has not diminished with improvements in CF care and the impact of liver disease is greatest in late adolescence early adulthood. We need a greater emphasis on providing early interventions for those at risk of CFLD if we are to further improve survival in CF.
Authors' contribution
Conception and Design of the Systematic Review
Ao Sasame, Lucy Connolly, Marion Rowland
Protocol Development and Publication
Ao Sasame, Lucy Connolly, Billy Bourke, Diarmuid Stokes, and Emer Fitzpatrick and Marion Rowland,
Search Strategy and Information Management
Ao Sasame, and Diarmuid Stokes.
Data Extraction and Interpretation
Ao Sasame, Lucy Connolly, Emer Fitzpatrick, Marion Rowland
Drafting the manuscript
Ao Sasame (first draft)
Critical revision of the article
Marion Rowland, Ao Sasame, Emer Fitzpatrick, Billy Bourke
Final approval of the version to be published
Ao Sasame, Diarmuid Stokes, Billy Bourke, Lucy Connolly, Emer Fitzpatrick, Marion Rowland
Funding source
This work was supported by the Health Research Board in Ireland (HRA-2014-PHR-662). The funding body had no role in the design, conduct or analysis of this systematic review.
Declaration of Competing Interest
None of the authors declare any conflict of interest.
Acknowledgements
We would like to thank Mr Edward Moore, Research Office, School of Medicine University College Dublin in preparing the graphical abstract for this manuscript.
Heterogeneous liver on research ultrasound identifies children with cystic fibrosis at high risk of advanced liver disease: interim results of a prospective observational case-controlled study.
01328-X&id=ga1.jpg){kind=link}
01328-X&id=gr1.jpg){kind=link}