Article Text
PDF
Abstract
Objective Some data suggest a positive association between non-alcoholic fatty liver disease (NAFLD) and incident major adverse cardiovascular events (MACEs). However, data are lacking from large cohorts with liver histology, which remains the gold standard for staging NAFLD severity.
Design This population-based cohort included all Swedish adults with histologically confirmed NAFLD and without cardiovascular disease (CVD) at baseline (1966–2016, n=10 422). NAFLD was defined from prospectively recorded histopathology and categorised as simple steatosis, non-fibrotic steatohepatitis, non-cirrhotic fibrosis and cirrhosis. Patients with NAFLD were matched to ≤5 population controls without NAFLD or CVD, by age, sex, calendar year and county (n=46 517). Using Cox proportional hazards modelling, we calculated multivariable adjusted HRs (aHRs) and 95% CIs for MACE outcomes (ie, ischaemic heart disease (IHD), stroke, congestive heart failure (CHF) or cardiovascular (CV) mortality).
Results Over a median of 13.6 years, incident MACE was confirmed in 2850 patients with NAFLD and 10 648 controls. Patients with NAFLD had higher incidence of MACE than controls (24.3 vs 16.0/1000 person-years (PY); difference=8.3/1000 PY; aHR 1.63, 95% CI 1.56 to 1.70), including higher rates of IHD (difference=4.2/1000 PY; aHR 1.64, 95% CI 1.54 to 1.75), CHF (difference=3.3/1000 PY; aHR 1.75, 95% CI 1.63 to 1.87), stroke (difference=2.4/1000 PY; aHR 1.58, 95% CI 1.46 to 1.71) and CV mortality (difference=1.2/1000 PY; aHR 1.37, 95% CI 1.27 to 1.48). Rates of incident MACE increased progressively with worsening NAFLD severity (ptrend=0.02), with the highest incidence observed with cirrhosis (difference vs controls=27.2/1000 PY; aHR 2.15, 95% CI 1.77 to 2.61).
Conclusion Compared with matched population controls, patients with biopsy-proven NAFLD had significantly higher incidence of MACE, including IHD, stroke, CHF and CV mortality. Excess risk was evident across all stages of NAFLD and increased with worsening disease severity.
- epidemiology
- fibrosis
- cardiovascular disease
Data availability statement
No data are available. No additional data are available due to Swedish regulations.
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
Some—but not all—prior studies suggest that non-alcoholic fatty liver disease (NAFLD) is associated with an increased risk of developing fatal and non-fatal major adverse cardiovascular events (MACEs).
However, the precise risk of developing MACE across the full NAFLD histological spectrum is unknown.
What are the new findings?
Among adults in Sweden with biopsy-confirmed NAFLD and no underlying cardiovascular disease (CVD), rates of incident MACE outcomes were significantly elevated, as compared with matched population controls without NAFLD or CVD.
Significant excess risk was found across all stages of NAFLD and increased progressively with worsening NAFLD histological severity.
NAFLD was also associated with significant excess risk of individual MACE outcomes, including ischaemic heart disease, stroke, congestive heart failure and cardiovascular mortality.
How might it impact on clinical practice in the foreseeable future?
These findings underscore the importance of conducting careful cardiovascular risk stratification in patients with NAFLD.
Public health efforts focused on the primary prevention of CVD in patients with NAFLD should be prioritised.
Introduction
Non-alcoholic fatty liver disease (NAFLD) represents a leading cause of chronic liver disease in the USA and Europe, where it affects over 100 million adults.1 2 Between 20% and 33% of patients with NAFLD develop progressive steatohepatitis (non-fibrotic steatohepatitis (NASH)) with fibrosis, which in turn can lead to cirrhosis, decompensated liver disease and liver-related mortality.3–7 Observational data have also linked NAFLD to the development of cardiovascular disease (CVD), including subclinical atherosclerosis,8 9 carotid atherosclerosis,10 cardiac diastolic dysfunction,11 subclinical myocardial infarction (MI) or stroke,8 and certain overt cardiovascular (CV) events, including acute MI.12 Based on this body of evidence, both the American Association for the Study of Liver Disease and the European Association for the Study of the Liver recommend that patients with NAFLD undergo careful CV screening.13 14
Despite this, data are still limited regarding the precise magnitude of CVD risk associated with NAFLD, and whether certain high-risk patients with NAFLD need targeted interventions to prevent CVD remains controversial.15 16 While two recent meta-analyses found a 1.6-fold to 1.8-fold higher risk of incident CVD in patients with NAFLD, compared with healthy controls,12 17 other data are conflicting,18 and published studies have either been small in size with few recorded events19 20 or have relied on surrogate definitions of NAFLD, which may introduce misclassification, and which cannot identify steatohepatitis (NASH) or stage fibrosis.12 17 18 Moreover, very few prior studies have compared major CVD events in patients with confirmed NAFLD to the general population, which is important for quantifying real-world risk.
We leveraged a nationwide histopathology cohort composed of all adults in Sweden with biopsy-confirmed NAFLD and matched general population controls, to examine the incidence of fatal and non-fatal major adverse cardiovascular events (MACEs), according to the presence and histological severity of NAFLD.
Methods
We conducted a population-based, matched cohort study using the established Epidemiology Strengthened by Histopathology Reports in Sweden (ESPRESSO) cohort. Briefly, ESPRESSO includes prospectively recorded liver histopathology submitted to all 28 pathology departments in Sweden (1965–2016) and thus is complete for the entire country.21 Each liver biopsy report includes the unique personal identity number assigned to all Swedish residents, the biopsy date and location, and describes topography within the liver and morphology, using the Systematised Nomenclature of Medicine system. We linked ESPRESSO to validated, nationwide registers with comprehensive data regarding demographics, comorbidities, prescribed medications, incident CV outcomes and death. Informed consent was waived as the study was register-based.22
For the current study, we identified all liver histopathology specimens submitted between 1966 and 2016 that confirmed an NAFLD diagnosis, from adults aged ≥18 years without another competing aetiology of liver disease and without CVD at the index date (ie, date of liver biopsy) using a validated International Classification of Diseases (ICD) algorithm (online supplemental figure S1 and eMethods).23 Consistent with prior work,23 we excluded anyone with a first recorded primary diagnosis or the second of two secondary diagnoses of another aetiology of liver disease, prior alcohol abuse/misuse, liver transplantation, CVD or emigration from Sweden, recorded on or before the index date (online supplemental table S1). In our previous validation of NAFLD cases in ESPRESSO,23 this methodology yielded a positive predictive value (PPV) of 92%. Patients meeting criteria for NAFLD were subsequently categorised into four histological groups that we also have validated,23 including simple steatosis (PPV=90%), NASH without fibrosis (PPV=87%), non-cirrhotic fibrosis (PPV=93%) and cirrhosis (PPV=97%) (see online supplemental appendix for definitions).23 We then matched each patient with NAFLD to up to five general population controls without recorded NAFLD or CVD, according to age, sex, calendar year and county of residence. General population controls were derived from the Total Population Register,24 and identical exclusion criteria were applied (online supplemental figure S1).
Supplemental material
Supplemental material
Outcomes and covariates were ascertained from the validated Patient Register, which prospectively records data from all inpatient and outpatient medical facilities in Sweden, including hospital discharge diagnoses (since 1964) and specialty outpatient care (since 2001), with established PPVs for clinical diagnoses between 85% and 95%, including for CV and liver-related diseases.25 The primary outcome was incident MACE, a composite outcome defined by ≥1 primary or secondary inpatient or outpatient ICD diagnosis for incident ischaemic heart disease (IHD) including acute MI, congestive heart failure (CHF), stroke or CV mortality (online supplemental table S2). Secondary outcomes included each individual MACE outcome. We also conducted a sensitivity analysis using a more stringent definition of MACE that required ≥2 primary inpatient diagnoses and which has demonstrated PPVs of 98%–100% for acute MI26 27 and 95% for CHF.28 Mortality was ascertained from the Total Population Register, which records 93% of all deaths in Sweden within 10 days and the remaining 7% of deaths within 30 days, and also from the Cause of Death Register.29
We collected detailed information regarding demographics, education, clinical comorbidities and prescription medication use (online supplemental table S3 and eMethods). Briefly, age at the index date (ie, the biopsy date among patients with NAFLD or the corresponding matching date among controls), sex, date of birth and emigration status were ascertained from the Total Population Register,24 and education level was obtained from the longitudinal integrated database for health insurance and labour market studies.30 Clinical comorbidities were extracted from the validated Patient Register using established ICD algorithms25 (online supplemental table S3 and online supplemental eMethods), and we also identified the number of hospitalisations for each patient in the 1 year preceding the index date. The Total Population Register also includes data from the Multigenerational Register regarding first-degree family members, which permitted us to ascertain a first-degree family history of early CVD diagnosed before 50 years of age. We also collected detailed data regarding use of the following medications: statins, other lipid-lowering agents, low-dose aspirin (<163 mg), other antiplatelets, anticoagulants, antidiabetic medications and antihypertensive agents.31 Medication use was ascertained from the Prescribed Drug Register, a well-validated and virtually complete nationwide register31 that includes prospectively recorded data for all dispensed prescriptions from Swedish pharmacies since July 2005.
Statistical analysis
Our primary analyses examined rates of incident MACE according to the presence and histological severity of NAFLD, compared with matched controls. Follow-up began at the index date (or corresponding matching date for controls) and continued to the first diagnosis of incident MACE, non-CV death, emigration or end of follow-up (31 December 2016). For all outcomes, we calculated incidence rates and absolute rate differences, together with 95% CIs. Using cause-specific Cox proportional hazard regression models, we estimated multivariable adjusted HRs (aHRs) and 95% CIs for incident MACE and defined other non-CV causes of death as competing events.32 For analyses of non-fatal secondary outcomes, all-cause mortality was considered a competing event. The fully adjusted multivariable model accounted for matching factors (ie, age, sex, calendar year and county) and a priori-selected confounders defined up to and including the index date (ie, diabetes, obesity, hypertension, dyslipidaemia, chronic kidney disease, a first-degree family history of early CVD before 50 years of age, education and number of inpatient hospitalisations in the 1 year before the index date), as well as diagnoses of alcohol abuse/misuse recorded during follow-up, consistent with our prior work23 (online supplemental table S3 and eMethods). The proportional hazards assumption was assessed by examining the relationship between Schoenfeld residuals and time.
To better characterise the gradient of risk associated with NAFLD histological severity and to minimise potential confounding from misclassification or from the underlying indication for liver biopsy, we excluded controls and restricted the cohort to patients with NAFLD, with simple steatosis as the comparator, consistent with prior work.23 In stratified analyses, we examined the associations between NAFLD and MACE outcomes according to age, sex, duration of follow-up, country of birth and metabolic comorbidities, and we tested the significance of effect modification using the contrast test statistic.33
To assess potential confounding related to shared genetic or early environmental factors, including a first-degree family history of early CVD, we identified all patients with NAFLD with ≥1 full sibling without recorded NAFLD or CVD before the start of follow-up and applied the same exclusion criteria as our primary analyses. We then compared each patient with NAFLD with his or her full sibling(s), after conditioning on matching set within family and further adjusting for all covariates in the multivariable model.
We conducted several sensitivity analyses to test the robustness of our results. First, because a widely used NAFLD histological scoring system was published in 2005,34 the same year that comprehensive prescription medication data were first available, we restricted the cohort to patients with index date ≥1 January 2006 and constructed models further adjusting for time-varying use of relevant medications (ie, aspirin, other antiplatelets, statins, other lipid-lowering agents, antidiabetic agents, antihypertensive medications and anticoagulants). Second, we repeated our primary analysis using the aforementioned, more stringent definition of MACE. Third, because we lacked detailed data regarding smoking, we constructed models further adjusting for chronic obstructive pulmonary disease (COPD) as a proxy for heavy smoking. Fourth, to address potential reverse causation, we excluded patients with incident MACE recorded within <90 days or within <2 years (with follow-up starting at day 90 or at 2 years, respectively). Finally, using an array-based approach for observational studies,35 we tested the robustness of our models to unmeasured confounding.
Analyses were conducted using R software V.3.6.1 (R Foundation for Statistical Computing, Vienna, Austria; Survival Package V.2.44 (https://CRAN.R-project.org/package=survival)). A two-sided p value of <0.05 was considered statistically significant.
Patient and public involvement
No patients were involved in setting the research question or the outcome measures. However, patients were involved in the establishment of the overall ESPRESSO cohort, which formed the foundation of this work. No patients were asked to advise on interpretation or writing up of results. The results of this research will be disseminated to patients by press release.
Results
Table 1 outlines the baseline characteristics of 10 422 adults with histologically confirmed NAFLD and 46 517 matched general population controls, at the index date. Among patients with NAFLD, the average age at the index biopsy was 52 years, and 45% were female. Simple steatosis was found in 7144 (68.5%), while 1183 (11.4%) had NASH without fibrosis; 1554 (14.9%) had non-cirrhotic fibrosis; and 541 (5.2%) had cirrhosis. Patients with NAFLD were more likely than population controls to have diabetes, obesity, hypertension, dyslipidaemia, chronic kidney disease and a first-degree family history of early CVD.
Baseline characteristics of adults with histologically confirmed NAFLD and matched population controls, at the index date (1966–2016)*
Major adverse cardiovascular event
Over a median of 13.6 years of follow-up, we confirmed 2850 incident MACE outcomes among patients with NAFLD (24.3/1000 person-years (PY)) and 10 648 among population controls (16.0/1000 PY), corresponding to an absolute rate difference of 8.3/1000 PY (95% CI 7.4 to 9.3) (figure 1 and table 2). After multivariable adjustment, patients with NAFLD had a significant, 1.63-fold higher rate of developing incident MACE, compared with controls (95% CI 1.56 to 1.70). Furthermore, rates of MACE outcomes increased monotonically with worsening NAFLD severity (ptrend=0.02) (figure 1 and table 2). Specifically, compared with controls, the absolute rate differences and corresponding aHRs were significantly elevated in patients with both simple steatosis (7.0/1000 PY; aHR 1.58, 95% CI 1.50 to 1.67) and NASH without fibrosis (8.1/1000 PY; aHR=1.52, 95% CI=1.32–1.75), and they were further amplified in patients with non-cirrhotic fibrosis (11.1/1000 PY; aHR 1.67, 95% CI 1.47 to 1.89) and cirrhosis (27.2/1000 PY; aHR 2.15, 95% CI 1.77 to 2.61).
Cumulative incidence of MACE outcomes according to the presence and histological severity* of NAFLD. *MACE was defined as a composite endpoint that included incident ischaemic heart disease (including acute myocardial infarction), congestive heart failure, stroke and cardiovascular mortality. NAFLD histological severity was defined in four validated categories: simple steatosis, NASH without fibrosis, non-cirrhotic fibrosis and cirrhosis (please see online supplemental appendix). MACE, major adverse cardiovascular event; NAFLD, non-alcoholic fatty liver disease; NASH, non-alcoholic steatohepatitis.
Incident MACEs among adults with histologically confirmed NAFLD* and matched population comparators
In stratified analyses, the significant, the positive association between NAFLD and incident MACE outcomes appeared stronger among women (aHR 1.74, 95% CI 1.63 to 1.86) compared with men (aHR 1.55, 95% CI 1.46 to 1.64; p-interaction=0.009) and also among patients diagnosed with NAFLD at younger ages (aHR at 18–39 years=2.52, 95% CI 2.15 to 2.95; aHR at 40–59 years=1.61, 95% CI 1.51 to 1.72; aHR at ≥60 years=1.51, 95% CI 1.42 to 1.62; p-interaction <0.001) (online supplemental table S4). Additionally, this association also appeared stronger in patients with a family history of early CVD, compared with those without, although the numbers of events in this subgroup were small (p-interaction=0.0048).
Individual mace outcomes
Compared with controls, patients with NAFLD had significantly higher rates of each secondary MACE outcome, including IHD (difference=4.2/1000 PY; aHR 1.64, 95% CI 1.54 to 1.75), CHF (difference=3.3/1000 PY; aHR 1.75, 95% CI 1.63 to 1.87), stroke (difference=2.4/1000 PY; aHR 1.58, 95% CI 1.46 to 1.71) and CV mortality (difference=1.2/1000 PY; aHR 1.37, 95% CI 1.27 to 1.48) (table 3). In analyses focused on incident acute MI, the pattern of association was similar (difference=1.6/1000 PY; aHR 1.47, 95% CI 1.35 to 1.61) (online supplemental table S5).
Individual MACE outcomes among adults with histologically confirmed NAFLD* and matched population comparators
Next, we evaluated secondary MACE outcomes according to NAFLD histological categories (table 3). Compared with controls, the absolute excess rates of developing IHD, CHF, stroke and CV mortality were significantly elevated in patients with simple steatosis (differences=3.9, 2.7, 2.0 and 0.8/1000 PY, respectively) and non-fibrotic NASH (differences=3.7, 3.2, 2.4 and 1.6/1000 PY, respectively), and they increased progressively in patients with non-cirrhotic fibrosis (differences=4.9, 4.9, 3.5 and 1.8/1000 PY, respectively) and even further with cirrhosis (differences=9.7, 11.2, 7.1 and 8.4/1000 PY, respectively).
NAFLD-only subgroup
After restricting the population to patients with biopsy-confirmed NAFLD, the absolute rates of incident MACE increased with worsening NAFLD histological severity, with the highest rates observed with cirrhosis (online supplemental table S6). Specifically, compared with simple steatosis, the absolute excess rates of MACE outcomes with non-fibrotic NASH, non-cirrhotic fibrosis and cirrhosis were 1.1, 4.1 and 20.2/1000 PY, respectively. After multivariable adjustment, this translated to a 27% higher rate of incident MACE outcomes in patients with cirrhosis, compared with those with simple steatosis (aHR 1.27, 95% CI 1.08 to 1.49).
Sibling analyses
To address potential residual confounding related to shared genetic and intrafamilial factors, including a family history of early CVD, we rematched 4763 patients with NAFLD to 9128 full-sibling comparators who were alive at the index date and without a recorded diagnosis of NAFLD or CVD (online supplemental table S7). Consistent with our primary analysis, patients with NAFLD demonstrated significantly higher rates of incident MACE, compared with full sibling controls (difference=6.0/1000 PY; aHR 1.73, 95% CI 1.54 to 1.93). These significant, positive associations persisted for each of the individual MACE outcomes, including IHD (difference=2.3/1000 PY; aHR 1.47, 95% CI 1.27 to 1.70), CHF (difference=2.4/1000 PY; aHR 1.94, 95% CI 1.56 to 2.42), stroke (difference=2.3/1000 PY; aHR 2.02, 95% CI 1.63 to 2.49) and CV mortality (difference=1.2/1000 PY; aHR 1.78, 95% CI 1.32 to 2.38). Consistent with our primary analyses, rates of incident MACE outcomes were significantly elevated across all NAFLD histological categories, with the highest rates found in patients with cirrhosis (online supplemental table S7).
Sensitivity analyses
Our findings were robust across all sensitivity analyses, including after restricting the cohort to patients with NAFLD with index biopsy ≥1 January 2006 (n=2201) and matched controls (n=8737), and further adjusting for all covariates including time-varying use of relevant CV medications (aHRMACE with NAFLD vs controls=1.27, 95% CI 1.04 to 1.55, which was highly similar when compared with the primary analysis from 2006 to 2016 (aHR 1.30, 95% CI 1.09 to 1.55; online supplemental table S4). We also applied a more stringent definition of MACE, and our results were not materially altered (aHR 1.55, 95% CI 1.48 to 1.63; online supplemental table S8). Similarly, our findings persisted after further adjusting for COPD (aHRMACE with NAFLD vs controls=1.62, 95% CI 1.55 to 1.69) and also after excluding any patient with a diagnosis of MACE recorded within <90 days of follow-up (aHR 1.58, 95% CI 1.51 to 1.66; online supplemental table S9) or within <2 years (aHR 1.58, 95% CI 1.50 to 1.66; online supplemental table S10). Finally, we tested the potential sensitivity of our model to unmeasured confounding and found that an unmeasured confounder would need to be simultaneously very strongly associated with risk of incident MACE (ie, aHR ≥3.5) and be highly imbalanced between patients with NAFLD and controls (ie, ≥50% difference in prevalence) to fully attenuate our results (see online supplemental table S11 for details of this analysis and results).
Discussion
In this population-based cohort composed of all Swedish adults with biopsy-confirmed NAFLD and matched general population controls, NAFLD was associated with significantly higher rates of both fatal and non-fatal MACE outcomes, including an increased incidence of IHD, MI, CHF, stroke and CV mortality. Compared with controls, patients with NAFLD had a 63% higher relative risk of developing MACE and an absolute excess rate of 8.3 per 1000 PY, which corresponds over 10 years to one additional incident MACE outcome per each 12 patients with NAFLD. Moreover, rates of both fatal and non-fatal CV events increased with worsening NAFLD histological severity, particularly among patients with progressive fibrosis. Although the highest overall risk was found with cirrhosis, we nevertheless observed significant excess risk in patients with earlier stages of NAFLD, including those with simple steatosis and non-fibrotic NASH. Importantly, our findings were robust across numerous sensitivity analyses, including after accounting for use of relevant medications, other causes of death and after comparing patients with NAFLD with their full siblings, to address important genetic and intrafamilial factors.
Our results confirm and extend prior literature, including a recent meta-analysis that similarly found a 1.6-fold higher rate of incident CV events in patients with NAFLD, compared with controls.12 However, others have found null associations, particularly for CV mortality,36 37 and the vast majority of prior studies have ascertained NAFLD using surrogate ultrasonographic or laboratory parameters, or administrative codes, which are imprecise and cannot identify NASH or stage fibrosis. This includes a recent European primary care cohort study that found a null association between NAFLD diagnoses and rates of incident MI or stroke18; however, that prior study had well-described limitations,16 38 including reliance on billing codes to define NAFLD, which may introduce substantial misclassification. Among the few published NAFLD histology cohorts, to date,19 20 39 all have been limited by small sample sizes (largest n=603) with few recorded outcomes and imprecise estimates of risk across NAFLD histological categories. In contrast, the current study benefitted from a comprehensive, nationwide population with biopsy-confirmed NAFLD, and included a larger number of patients and more recorded MACE outcomes (n=2850) than all prior histology cohort studies combined.19 20 39
Preclinical evidence supports a link between NAFLD and the development of CVD. While the precise mechanisms are undefined, NAFLD may contribute to CVD by exacerbating insulin resistance, atherogenic dyslipidaemia, oxidative stress16 40 or intestinal dysbiosis,41 by stimulating platelet activation42 or by promoting the release of proinflammatory, vasoactive cytokines.43–45 Alternatively, a recent clinical study of 60 adults with biopsy-confirmed NAFLD found that progressive NAFLD was associated with higher circulating levels of the gut microbiota-derived metabolite, trimethylamine-N-oxide,46 which contributes to platelet hyperreactivity and thrombosis, and which has been linked to both fatal and non-fatal CVD events.47 Finally, NAFLD may also contribute to CVD risk through dysregulation of diverse bioactive lipid pathways.48
It is currently thought that progressive NAFLD fibrosis is the most important histological predictor of clinical outcomes, including CV events13 49; however, robust data to support this hypothesis are limited. Our findings confirm this association in a nationwide population with biopsy-confirmed NAFLD, and our large sample size allowed us to study differences in outcomes between groups of patients with varying degrees of NAFLD severity, something that was not possible in previous smaller studies. We found that even simple steatosis was associated with higher rates of MACE, which supports recent data demonstrating that simple steatosis may not be a benign condition.23 50 Moreover, compared with patients with simple steatosis, those with non-cirrhotic fibrosis and cirrhosis had significantly higher absolute excess rates of developing MACE outcomes (4.1 and 20.2 per 1000 PY, respectively), which corresponds over 10 years to one additional incident MACE outcome per every 24 patients diagnosed with non-cirrhotic fibrosis, and per every 5 patients diagnosed with cirrhosis, compared with those with simple steatosis. In contrast, compared with simple steatosis, the absolute excess rate of developing incident MACE was not significantly elevated in patients with non-fibrotic NASH (1.1 per 1000 PY). Taken together, those findings support the need to closely examine and modify CVD risk factors in patients with NAFLD fibrosis or cirrhosis. Additionally, further research is needed to determine if personalised CV risk stratification tools might offer meaningful benefits for certain high-risk patients with biopsy-confirmed NAFLD.
It has been hypothesised that the link between NAFLD and MACE outcomes could be explained by features of the metabolic syndrome. Consistent with other administrative cohorts, the recorded prevalence of metabolic comorbidities in this dataset was relatively low, which could result in unmeasured confounding. However, the PPVs for clinical diagnoses in the Swedish Registers is high (ie, between 85% and 95%25), and we found consistent associations in patients with diabetes or any metabolic comorbidity, as compared with controls with the same diagnoses. Moreover, while it is possible that patients with NAFLD were more likely to have diagnoses of metabolic comorbidities compared with controls, our findings remained broadly similar in the NAFLD-only subgroup analysis. Finally, published evidence from the general population indicates that the full metabolic syndrome contributes modestly to the risk of developing incident MACE (pooled relative risk=1.78, 95% CI 1.58 to 2.00),51 and our sensitivity analysis demonstrated that the association for NAFLD is very robust to unmeasured confounding. Specifically, a confounder would need to have both an aHR of ≥3.5 for incident MACE and simultaneously have a ≥50% difference in prevalence between patients with NAFLD and controls, to fully attenuate our results. Thus, our findings suggest that the excess risk of MACE observed with NAFLD is unlikely to be fully explained by features of the metabolic syndrome alone.
Strengths of this study include a comprehensive, nationwide population with prospectively recorded histopathology, together with strict, validated definitions of both NAFLD23 and confounding variables, in registers with near-complete follow-up for the entire population. Our larger sample size permitted calculation of more precise risk estimates across the full NAFLD histological spectrum, and enabled us to account for established CVD risk factors, relevant medications and important intrafamilial factors. We also applied robust techniques to address potential bias related to residual confounding, the underlying indication for liver biopsy and non-CV causes of death, while accounting for ranges of time between liver biopsy and MACE outcomes helped to minimise reverse causation.
Several limitations should be addressed. First, this was a retrospective study with NAFLD defined histologically, and not all patients with NAFLD undergo biopsy, which could introduce selection bias. However, our case distribution across histological groups accords with previous smaller studies,20 and our risk estimates are highly consistent with recent meta-analyses12 17 that primarily included cohorts in which NAFLD was defined using non-invasive laboratory or imaging approaches, which underscores the generalisability of our findings. Second, while it is possible that our controls could have included patients with undiagnosed NAFLD, such misclassification would most likely have attenuated, rather than exaggerated, our results. Moreover, our findings were consistent in the NAFLD-only subgroup analysis, in which all patients underwent liver biopsy. Third, despite careful matching, some residual confounding is possible, as we lacked detailed laboratory data or information regarding smoking, precise alcohol consumption or body mass index. Patients with NAFLD also had a higher prevalence of diabetes and metabolic syndrome than controls, and future studies are needed that focus on CVD risk in patients with NAFLD and diabetes. Nevertheless, our findings were robust after multivariable adjustment for important CV risk factors, including clinical comorbidities, relevant medication use, subsequent alcohol use disorders and COPD, and even a first-degree family history of early CVD, an important risk factor that has not been considered in prior NAFLD cohort studies. Moreover, after rematching patients with NAFLD with full siblings to address shared intrafamilial and early-life factors, the associations for MACE outcomes were confirmed. Fourth, it is possible that the observed association between NAFLD histology and CVD outcomes is mediated by established CVD risk factors; thus, future studies are needed in large cohorts that include both liver histopathology and detailed clinical and anthropometric assessments. Fifth, despite the well-described limitations of histopathology, our prior validation study in this cohort has demonstrated the accuracy of our exposure definitions.23 Finally, the Swedish population is primarily Caucasian, highlighting the need for additional research in more diverse, large-scale histology cohorts.
In conclusion, within a large, population-based cohort of 10 422 adults with biopsy-confirmed NAFLD and matched population controls without underlying CVD, NAFLD was associated with significantly higher incidence of both fatal and non-fatal MACE outcomes, including IHD, acute MI, CHF, stroke and CV mortality. Our study provides quantitative estimates regarding the risks of developing incident MACE according to the presence and histological severity of NAFLD, for the first time on a nationwide scale. We observed significantly higher rates of fatal and non-fatal MACE outcomes across all stages of NAFLD, and the magnitude of risk increased progressively with worsening NAFLD histological severity. Thus, our results provide support for improved CV screening and risk stratification strategies in high-risk patients with NAFLD fibrosis and cirrhosis who may be most likely to benefit from primary CVD prevention.
Data availability statement
No data are available. No additional data are available due to Swedish regulations.
Ethics statements
Patient consent for publication
Ethics approval
ESPRESSO was approved by the Stockholm Ethics Board on 27 August 2014 (number 2014/1287-31/4).
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors Guarantor: the corresponding author (TGS) had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Acquisition of data: JFL. Analysis: BR. Writing the first draft of the manuscript: TGS and JFL.
Study concept and design, interpretation of data, critical revision of the manuscript for important intellectual content and approval of final version: all coauthors.
Funding TGS was supported by NIH K23 DK122104, the Harvard University Center for AIDS Research and the Dana Farber/Harvard Cancer Center GI SPORE. HH was supported by grants from Region Stockholm (postdoctoral appointment). JFL was supported by the Karolinska Institutet.
Competing interests JFL coordinates an unrelated study on behalf of the Swedish IBD Quality Register that has received funding from Janssen Corporation. TGS has received research funding from Amgen and has received consulting fees from Aetion for work unrelated to this manuscript. HH reports research grants to his institution from AstraZeneca, Pfizer, Merck, EchoSens, Intercept and Gilead, and board advisory for Bristol-Myers Squibb and Gilead. JS reports ownership in companies providing services to Itrim, Amgen, Janssen, Novo Nordisk, Eli Lilly, Boehringer, Bayer, Pfizer and AstraZeneca, outside the submitted work. The remaining authors have no disclosures and no conflicts of interest to disclose.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.