Risk and adverse outcomes of fractures in patients with liver cirrhosis: two nationwide retrospective cohort studies

Objective The aim of this study is to evaluate fracture risk and post-fracture outcomes in patients with and without liver cirrhosis (LC). Design Retrospective cohort study and nested fracture cohort study. Setting This study was based on Taiwan’s National Health Insurance Research Database that included information on: (1) 3941 patients aged 20 years and older newly diagnosed with LC between 2000 and 2003; (2) 688290 hospitalised fracture patients aged 20 years and older between 2006 and 2013. Primary and secondary outcome measures Followed-up events of fracture from 2000 to 2008 were noted from medical claims to evaluate adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) of fracture associated with LC. Adjusted odds ratios (ORs) and 95% CIs of adverse events after fracture were compared among patients with and without LC Results The incidences of fracture for people with and without LC were 29.1 and 17.2 per 1000 person-years, respectively. Compared with controls, the adjusted HR of fracture was 1.83 (95% CI 1.67 to 2.01) for patients with LC. Previous LC was associated with risks of septicaemia (OR 1.77, 95% CI 1.60 to 1.96), acute renal failure (OR 1.63, 95% CI 1.33 to 1.99), and 30-day in-hospital mortality (OR 1.61, 95 %CI 1.37 to 1.89) after fracture. Conclusion LC was associated with higher risk of fracture; patients with LC in particular had more complications and 30-day in-hospital mortality after fracture. Fracture prevention and attention to post-fracture adverse events are needed for these susceptible populations.


IntrODuCtIOn
Liver cirrhosis (LC) is the fourth most common cause of death in Europe, and causes more than one million deaths every year worldwide. 1 24][5] Although the epidemiology, pathogenesis, prevention and treatment of LC have been studied, 6 complications of LC are not well understood.
With the use of reimbursement claims from Taiwan's National Health Insurance programme, we conducted two nationwide cohort studies.The retrospective cohort study seeks to validate the risk of fracture in patients with LC.Whether LC was associated with adverse outcomes after fracture was reported in the nested fracture cohort study.
strengths and limitations of this study

study design
This investigation included two studies.In study I (the retrospective cohort study), our purpose was to evaluate the risk of fracture for people with and without LC.From the representative sample of 1 000 000 insurance enrollees, we required at least two visits for medical care and a physician's primary diagnosis of cirrhosis of the liver to identify a cohort of 3941 newly diagnosed adults aged ≥20 years in 2000-2003.Those with only one medical visit and a physician's diagnosis of LC were not considered as cases of cirrhosis in this study.The frequency-matching procedure (by age and sex) was used to select the cohort with no previous medical records of LC.Both LC and non-LC cohorts had no history of fracture between the index date (date of LC diagnosis) and 1 January 1996 (the starting date of the Taiwan's National Health Insurance Programme).That is to say, there was no recorded previous fracture from onset of the database (1996) until the date of enrollment in the study (2000-2003).The outcome of this retrospective cohort study was an incidence of fracture that was identified during the follow-up period from the index date until the end of 2008 for LC and non-LC cohorts.
In study II (the nested fracture cohort study), our purpose was to evaluate the outcomes after fracture in patients with fracture with and without a history of LC.Study II included 688290 hospitalised patients with fracture in 2004-2013; we identified 7854 patients with a history of LC (defined as at least two visits for medical care and a physician's primary diagnosis of LC) within 24 months pre-fracture.Thirty-day in-hospital mortality, septicaemia, and acute renal failure after fracture were considered post-fracture outcomes and were compared in patients with fracture with and without LC in the nested fracture cohort study.

statistical analysis
In study I, the categorical data for cohorts with and without LC were analysed by χ 2 tests.The adjusted hazard ratios (HRs) and confidence intervals (CIs) of fracture risk associated with LC were calculated using multiple Cox proportional hazard models, controlling for age, sex, low income, mental disorders, hypertension, chronic obstructive pulmonary disease, diabetes, ischaemic heart disease, stroke, hyperlipidaemia, congestive heart failure, renal dialysis, Parkinson's disease, anxiolytics, antipsychotics, antiepileptics, antidepressants, and oral steroids, as were associations between LC and fracture risk in men, women, and every age group.

Open Access
In study II, we used χ 2 tests to examine other sociodemographic factors and medical conditions in hospitalised patients with fracture with and without a history of LC.Multiple logistic regressions were used to calculate adjusted odds ratios (ORs) and 95% CIs of 30-day in-hospital mortality, sepsis and acute renal failure after fracture associated with a history of LC, controlling for age, sex, low income, mental disorders, hypertension, diabetes, chronic obstructive pulmonary disease, ischaemic heart disease, stroke, congestive heart failure, Parkinson's disease, renal dialysis, hyperlipidaemia, and types of fracture.
During the 5-8 years of follow-up (study I), the incidence of fracture for cohorts with and without LC was 29.1 and 17.2 per 1000 person-years, respectively (table 2).The increased risk of fracture was found in the LC cohort  3).Patients with fracture and LC included lower proportions of young adults (p<0.0001)but higher proportions of men (p<0.0001) and those with low-income status (p<0.0001).More patients with a history of mental disorders, diabetes, chronic obstructive pulmonary disease, ischaemic heart disease, stroke, congestive heart failure, Parkinson's disease, and renal dialysis were found in the LC cohort than in the control group (p<0.05 for all).In patients with fracture and previous LC, there were higher proportions of neck or trunk fracture (p<0.0001) and lower limb fracture (p<0.0001).

DIsCussIOn
In study I (the retrospective cohort study), we observed a significant association between LC and fracture, with an 83% increased risk of fracture in patients with LC during the 5-8 years of follow-up.In study II (the nested fracture cohort study), we observed that patients with LC had significantly higher post-fracture complications and 30-day in-hospital mortality.Clinical indicators of the severity of LC, such as alcohol dependence syndrome, jaundice, ascites, gastrointestinal haemorrhage and hepatic coma, were all associated with more post-fracture adverse events.
In general, the prevalence of LC was higher in men than in women. 3 4Older age and low socioeconomic status were factors associated with higher risk of LC. 3 23 24 Older age, male sex and low income were also risk factors for fracture, and the multivariate Cox proportional models was used to control these potential confounding effects when analysing the association between LC and Open Access fracture risk. 25Furthermore, we found that the association between LC and fracture risk remained significant in every age group and in both sexes.The significant impact of LC on post-fracture adverse events was noted in men, women, in various age groups, and in people with various types of fracture.This phenomenon revealed the possible causal inference that LC was associated with fracture risk and post-fracture adverse events from the viewpoint of epidemiology.These findings are crucial because several previous studies were limited by focusing on specific populations and failed to investigate the association in a subgroup analysis. 14-17 19 20ntal disorders, hypertension, chronic obstructive pulmonary disease, diabetes, ischaemic heart disease, stroke, hyperlipidaemia, congestive heart failure, renal dialysis and Parkinson's disease were considered as coexisting medical conditions that were also fracture risk factors. 21 22 26-303][14][15][16][17][18] Therefore, we used multiple Cox proportional hazard and multiple logistic regression models to control the confounding effects of medical conditions Open Access when investigating the risks and outcomes of fracture in patients with LC in studies I and II.
3][14][15][16][17][18][19][20] Patients with fracture and a history of LC had longer hospital stay and increased medical expenditure than people without LC in the nested fracture cohort study.Patients with LC had circulatory dysfunction and poor immune systems that compromised systemic inflammatory response and made them prone to renal failure and septicaemia, 23 31 particularly those patients with cirrhotic indicators such as alcohol dependence syndrome, jaundice, ascites, gastrointestinal haemorrhage, and hepatic coma.Therefore, higher mortality and increased use of medical resources might be encountered in the LC population during fracture admission.
There are several possible explanations for associations between LC and fracture risk.First, many studies found that patients with LC had increased risk of osteoporosis, [9][10][11] a condition that is an important determinant for fracture. 8Fracture due to bone loss and the pathogenesis of osteoporosis among patients with LC is complex and multifactorial, and the exact mechanism remains uncertain.A previous study showed patients with cirrhosis and osteoporosis had lower levels of insulin-like growth factor 1 than patients with cirrhosis without osteoporosis. 32Insulin-like growth factor 1 plays a major role in bone remodelling and maintenance of bone mass, and was found to be reduced in advanced cirrhosis. 33In patients with cirrhosis, hyperbilirubinemia has also been shown to impair osteoblast proliferation, resulting in decreased bone formation and possibly accounting for the increased risk of fracture. 34Second, corticosteroids  Open Access are frequently used in patients with autoimmune hepatitis and other inflammatory disorders.Even budesonide, a corticosteroid with minimal systemic availability, might lead to accelerated bone loss in patients with cirrhosis and postmenopausal women. 35We postulated that medications used in the treatment of LC could also have an adverse effect on bone and calcium mobilisation and subsequent osteoporosis.Third, hepatic coma, poor cognitive function and psychiatric illness may play roles in the association between LC and risk of fracture. 25 36lthough hepatic encephalopathy does not commonly occur in patients with LC, its contribution to falls should not be ignored. 37ome study limitations need to be considered when interpreting the results.First, this study used retrospective reimbursement claims, which lack data on severity of LC, lifestyle factors, personal characteristics and biochemical data.Compared with the previous study, 19 theunavailable information is an important source of bias.Second, since the patients were selected based on diagnoses from hospital inpatient care registers, patients with minor LC but no symptoms might not consult medical services, leading to underestimation of fracture risk in patients with LC because some cases with minor LC may have been in the non-LC group.Third, because our results are based on the data from Taiwan's National Health Insurance, the findings of this study could not be directly generalised to other populations.

Conclusion
Our two cohort studies provide population-based evidence that LC is an important risk factor for fracture.We also note that patients with fracture and various clinical indicators of LC severity face increased risks of post-fracture adverse events.We presented risk factor analysis and a variety of clinical suggestions, including prevention, risk assessment and outcome-related information in patients with fracture and LC.Strategies Open Access to prevent fracture and meticulous care to reduce post-fracture adverse events should be routinely considered for this population.

►
Our study is a longitudinal retrospective cohort with large sample size.► We used frequency matching and multivariate Cox proportional hazard models to control the confounding bias.► This is the first study with comprehensive assessment of the impact of liver cirrhosis on fracture risk and outcomes.
► Our data lack clinical risk scores, lesion characteristics, biochemical measures, and lifestyles of patients with fracture and patients with liver cirrhosis.► Detailed lifestyle data including smoking, alcohol drinking, and physical activity were not available.ethical approval Insurance reimbursement claims used in this study were decoded with patients' identification scrambled for further research access.This study was conducted in accordance with the Helsinki Declaration.Although NHRI regulations do not require informed consent because patient identification has been decoded for privacy, this study was also approved by Taipei Medical University's Joint Institutional Review Board (TMU-JIRB-201705063; TMU-JIRB-201705084; TMU-JIRB-201506001; TMU-JIRB-201404070).

Table 2
Risk of fracture events for cohorts with and without liver cirrhosis *

Table 3
Characteristics of patients with fracture with and without liver cirrhosis

Table 4
Adverse events after fracture in patients with and without liver cirrhosis

Table 5
Liver cirrhosis associated with post-fracture adverse events in the stratification analysis by age, sex and type of fracture *Any adverse events included 30-day in-hospital mortality, septicaemia, and acute renal failure.†Controlled for all covariates listed in table 3. CI, confidence interval; OR, odds ratio.