Non-alcoholic fatty liver disease (NAFLD) and the cardiovascular disease (CVD) risk categories in primary care: is there an association?

We conducted a cross sectional study among patients attending two primary care clinics from a university health centre in Selangor, Malaysia between July 2017 to July 2018. The conduct of the study is shown in Fig. 1. We included participants 18 years and older who have at least one risk factors for NAFLD or CVD. These include abnormalities of liver enzymes (ALT ≥ 34 U/L, GGT > 60 U/L), cholesterol components (TC ≥ 5.0 mmol/L, LDL-C ≥ 3.1, TG ≥ 1.7 mmol/L, HDL-C ≤ 1.0 mmol/L in men or HDL-C ≤ 1.3 mmol/L in women), impaired fasting glucose > 5.6 mmol/L or random glucose > 7.8 mmol/L or known diabetes mellitus (DM), elevated HbA1c > 7.0%, elevated blood pressure (BP) ≥130/80 or known hypertension or abnormal waist circumference (WC) ≥80 cm in women or ≥ 90 cm in men. Patients with the following criteria were excluded: (a) previously diagnosed with chronic liver disease (b) history of alcohol intake of > 140 g per week for male and > 70 g per week for female and (c) pregnant prior to the study or during the study period.

The 10-year CVD risk for each participant was calculated using Framingham Risk Score (FRS) general CVD risk score 2008 [20]. It allows the calculation of probability risk of CVD events (coronary artery, cerebrovascular and peripheral vascular events). The score was calculated separately for men and women using online calculator. This tool has been calibrated and validated to be used in primary care settings among Malaysian adults [21]. The variables included in the tools were age, total cholesterol, HDL, systolic blood pressure, diabetes and smoking status. The score was transformed into 10-year CVD risk that categorized as low risk < 10%, moderate risk 10–20% and high risk > 20% [21].

Sample size was calculated using OpenEpi Tool Version 3.01, single proportion formula for “Sample Size for a Proportion or Descriptive Study” from http://​openepi.​com/​SampleSize/​SSPropor.​htm

The formula for the calculation is:

n = sample size.

DEFF = design effect (for cluster surveys).

N = population size (for finite population correction factor or fpc).

p = hypothesized % frequency of outcome factor in the population.

Z² 1 − α/2 = confidence interval.

d = desired precision.

The population size (N) was estimated based on the total number of patients receiving care in the primary care clinics in a year i.e. 8000 (as per registry from the information technology unit). The percentage (%) of the prevalence of NAFLD was estimated as 22.7% based on a previous study done in Malaysia [14]. In this study, the study population was adult patients attending health screening in a private hospital. In the absence of studies conducted in primary care setting, this is the closest to resemble primary care setting. Therefore, this percentage was taken as the hypothesized frequency. The confidence interval was taken as 95%. Based on these assumptions, the minimum required sample needed for this study was 261 patients. After considering additional 30% of non-eligibility and dropout rate, our study targeted to approach 339 patients.

Patients aged ≥18 years who attended the two primary care clinics during the data collection days were consecutively approached by the researcher in the waiting area after they have received their registration numbers. They were given a Patient Information Leaflet about the study in Malay or English languages and written informed consent was obtained. Patients who agreed were screened according to the eligibility criteria. These were done through physical examinations, tracing the medical records for medical history and blood investigations. Those who were eligible were recruited into the study.

Data was collected from July to September 2017 by the same researcher throughout the data collection period. One nurse from each clinic was trained on how to do the anthropometry measurement procedures according to the study protocol. This was to ensure standardisation and to minimise the variability of data collection. A structured proforma which consisted of six sections was used to collect data from participants. The six sections were (a) socio-demographic information (b) medical history (c) anthropometrics data (d) biochemistry results (e) ultrasound findings and (f) Framingham Risk Score.

All the collected data were entered and analysed using Statistical Package for Social Sciences (SPSS) version 23 (SPSS, Inc., Chicago, IL, USA). Categorical variables were described in numbers and percentage whereas continuous variables were expressed as mean with standard deviation (SD). Inferential analysis was conducted to compare socio-demographic and clinical characteristics of participants with and without NAFLD. Simple logistic regression (SLogR) was used as preliminary analysis to identify the significant factors for NAFLD. Variables with p < 0.25 were included in the Multiple Logistic Regression (MLogR) to determine the independent associated factors for NAFLD after adjusting for the confounders. A p-value of < 0.05 was considered significant.

Out of 340 patients who were approached, 316 agreed to participate. Out of 316, 296 (87.1%) fulfilled the eligibility criteria. However, 33 were excluded from analysis due to not completing the ultrasound investigation giving reasons such as ‘busy with commitments’, ‘being overseas’, ‘no transportation’, or withdrew from the study; eight had incomplete blood investigation results and one participant was found to be pregnant during the reminder call for blood taking appointment. Therefore, the total number included in the final analysis was 263 (77.4%).

Out of 263 participants, 143 (54.4%) of them was found to have NAFLD with 95% CI (48,60%). Socio-demographics and clinical characteristics of those with and without NAFLD are shown in Table 1. Univariate analysis using Chi Square test found several factors to be significantly associated with NAFLD. These factors were age, gender, occupational status, T2DM, dyslipidaemia, BMI, waist circumference, blood pressure, fasting blood sugar, ALT, GGT and HDL-cholesterol.

Figure 2 illustrates the prevalence of NAFLD according to FRS categories. Participants with high FRS category had a greater prevalence of NAFLD (p = 0.025). The mean FRS was also calculated and compared between the two groups. It was found that the mean FRS was significantly higher in individuals with NAFLD compared to those without NAFLD (17.38 ± 12.35 vs. 12.35 ± 12.89, p = 0.003).

Variables with p-value < 0.25 from SLogR were included in the MLogR. These included age, gender, employment status, BMI, WC, systolic blood pressure, HDL-C, FPG, ALT, GGT and FRS categories. Table 2 shows the results of the MLogR. Five factors were found to be independently associated with NAFLD. These were being employed (Adj. OR 2.44, 95%CI 1.46,4.70), BMI of obese category (Adj. OR 2.89, 95%CI 1.21,6.91), elevated FPG (Adj OR 2.79, 95%CI 1.44,5.43), elevated ALT (Adj OR 3.70, 95%CI 1.85,7.44) and high FRS category (Adj OR 2.82, 95%CI 1.28,6.23). Post regression analysis showed that the model reasonably fits well. The receiver operating characteristic (ROC) curve gave an area under the curve of 0.779 (95%CI 0.723,0.835) which indicated that the model could accurately discriminate 77.9% of the cases.

To the best of our knowledge, this study was the first prospective study in Malaysia to determine the prevalence of NAFLD in the primary care setting. Previous studies were conducted in secondary care settings. This study was also the first that has established the association of 10-year CVD risk categories and NAFLD in Malaysia.

In this study, the overall prevalence of NAFLD was reported as 54.4% (95% CI 48, 60%). Our finding is comparable with other local studies conducted among individuals with risk factors. The prevalence of NAFLD among participants with T2DM in a university hospital setting was 49.6% [15]. Our finding is also comparable to another study among participants with dyslipidaemia in another university hospital setting where the prevalence was 56.7% [16]. The prevalence of NAFLD among patients with MetS was even higher (82.8%) as reported by another study in a public hospital [22]. However, the prevalence of NAFLD in a general population attending health screening at a private hospital was lower at 22.7% compared to our study [14]. This is understandable as their study included individuals with and without risk factors for NAFLD or CVD.

In comparison with studies in the Asian region and the West, the prevalence of NAFLD in our study was also comparable with studies among individuals with risk factors. In India, the prevalence was 56.5% among individuals with T2DM attending diabetic clinics [23]. This was supported by a meta-analysis of 17 studies on the prevalence of NAFLD in patients with T2DM where the overall prevalence was 54% (95%CI 45,64%) [24]. In another study conducted among patients with T2DM in a Brazilian tertiary hospital, the prevalence was 69.4% [25]. In contrast to our study, the prevalence of NAFLD in the general population was found to be generally lower. A population-based cross sectional study in Thailand showed a prevalence of 22.9% [26]. In the US, the prevalence was 31% in a multi-ethnic population study [27]. Another population study in Italy showed that the prevalence of NAFLD was 22.6% [28]. Overall, the prevalence of NAFLD in our study which included individuals with at least 1 risk factor for NAFLD or CVD was shown to be higher when compared to studies done in the general population. Our finding was comparable to studies which included individuals with risk factors.

With regards to FRS categories, this study showed that the prevalence of NAFLD in the high FRS risk group was significantly higher compared to the moderate and low risk groups (65.5% vs. 55.4% vs. 46.3%, p = 0.025). Out of 143 participants with NAFLD, 39.9, 21.6 and 38.5% were in the low, moderate and high risk categories, respectively. Our findings were comparable to a study in Korea which showed that out of 5769 participants with NAFLD; 62.1, 28.4 and 9.5% were in the low, moderate and high risk categories, respectively [29]. However, their study used the FRS 10-year coronary artery disease (CAD) risk prediction tool.

Regarding factors associated with NAFLD, our study found five factors to be significantly associated. These included being employed, BMI of obese category, elevated FPG, elevated ALT and high FRS category.

Participants who were currently employed have a higher odd of having NAFLD (Adj. OR 2.44, 95% CI 1.46, 4.70). In contrast, a study in China reported that those who were employed were less likely to have NAFLD (OR 0.69, p = 0.008) [30]. Another study conducted among police officers in China found that high occupational stress (HR = 1.727, 95% CI = 1.405–2.124) and high personal strain (HR = 3.602, 95% CI = 1.912–6.787)) were independent predictors for NAFLD [31]. However, in our study, data was not collected with regards to the type of occupation. Therefore, further deduction could not be made and more studies are needed to explore the type of occupation which may be associated with NAFLD.

With regards to BMI, our study shows that participants who were obese (BMI ≥ 27.5 kg/m²) had 2.89 times the odds of having NAFLD compared to non-obese individuals (Adj. OR 2.89, 95%CI 1.21, 6.91). Similar finding was also observed from an Asian study which reported that overweight individuals (BMI > 25 kg/m²) were more likely to have NAFLD; OR 1.05 (95%CI 1.004,1.09) p = 0.031 [32]. In a study conducted in Malaysia among overweight individuals with BMI > 23 kg/m², the odds of having NAFLD was also higher compared to those with normal BMI; OR 14.66 (95%CI 9.62,22.33) p < 0.001 [14]. These findings suggested that obese and overweight individuals are more likely to have NAFLD. This might be explained by the increased in fatty acid metabolism in overweight and obesity that lead to accumulation of triglyceride in the liver [33].

In our study, individuals with elevated FPG (≥ 5.6 mmol/L) had 2.79 times the odds of having NAFLD compared to those with normal FPG (Adj OR 2.79, 95%CI 1.44, 5.43). Similarly, in Sri Lanka, elevated FPG of ≥ 5.6 mmol/L was found to be associated with NAFLD; OR 1.7 (95%CI 1.39,2.08) p < 0.001 [13]. This is also comparable to a study in China which showed that elevated FPG (≥6.1 mmol/L) was independently associated with NAFLD; OR 3.324 (95%CI 1.89,5.85; p < 0.001) [34]. Elevated FPG indicates abnormal glucose metabolism or insulin resistance. This condition has been thought to be the influencing factor or active metabolic factor for fat deposition in the liver [2].

It was found from our study that those individuals with elevated ALT (> 34 U/L) had an increased odds of having NAFLD by 3.70 times compared to those with normal ALT level (Adj OR 3.70, 95%CI 1.85, 7.44). This is consistent with several studies which found that elevated ALT was significantly associated with NAFLD with slightly different upper limit values as per local references. In Taiwan, ALT (> 40 IU/L) was significantly associated with NAFLD; OR 5.66 (95%CI 3.99,8.01) p < 0.001 [11]. While in Sri Lanka, individuals with high ALT (defined as twice the upper limit of normal) had increased odds of having NAFLD compared to those with normal ALT level; OR 2.28 (95%CI 1.32,3.94) p = 0.003 [13].

In terms of CVD risk categories, our study shows that individuals in the high FRS CVD risk category had an increased odd of having NAFLD by 2.82 times compared to those in the low risk category (Adj OR 2.82, 95%CI 1.28,6.23). The study by Choi et al. found that NAFLD was strongly associated with moderate FRS CAD risk category (OR: 1.26 95%CI 1.11,1.42, p < 0.001) and high FRS CAD risk category (OR: 1.35 95%CI 1.10,1.65, p < 0.001) [29]. In a longitudinal cohort study in the US, multivariate analysis showed that FRS was the only variable significantly associated with new onset CAD (OR = 1.13, 95% CI = 1.05–1.21; p = 0.001) [7]. Patients with NAFLD was found to have a higher 10-year CAD risk than the general population of the same age and gender [7] . Therefore, the findings of this study underscore the significant association between NAFLD and CVD risk.

The strength of this study includes prospective diagnosis of NAFLD using ultrasound scan by a radiologist. This study was also the first that has established the association of 10-year CVD risk categories and NAFLD in Malaysia.

There are a few limitations of this study. Firstly, the cross-sectional study design would not be able to draw causal relationship between various factors associated with NAFLD. Secondly, the use of non-probability sampling method may be prone to sampling bias. Therefore, efforts were made to reduce sampling bias by ensuring patients who attended the clinics on data collection days were approached consecutively and invited to participate. Thirdly, this study was conducted in a university primary care setting where majority of the participants were Malays. Therefore, the results may not be generalised to other primary care settings where the population served might be different. It is also acknowledged that medications status was not highlighted in this study hence it might contribute to possible confounding factor to the final analysis. Finally, this study only included two main categories of factors that are associated with NAFLD i.e. the socio-demographics and clinical factors. Other factors that may be associated with NAFLD such as psychological and environmental factors were not included in this study. Therefore, the results of the multivariate analysis should be interpreted with caution.

This study purposely recruited patients with ≥ 1 risk factors for NAFLD or CVD rather than those with no risk factor. These patients were chosen because they should be targeted for screening of NAFLD in view of limited resources in the primary care clinics. Currently, there is no guideline on routine screening for NAFLD even in targeted groups. According to the American Association of the Study for Liver Disease (AASLD), recommendation for routine screening for NAFLD is not made at this point as there are still ongoing studies to better understand NAFLD in terms of its natural history, diagnosis and treatment [3]. However, the findings of our study should strengthen the body of evidence to suggest targeted screening for NAFLD in individuals with obesity, elevated FPG, elevated ALT and high FRS category.

Figure 3 illustrates the proposed algorithm for screening of NAFLD in the target groups.

Patients with at least one CVD or NAFLD risk factor should be risk stratified using the 10-year FRS general CVD risk score. If they are found to have high FRS, or obese or have elevated FPG or elevated ALT, they are recommended to have a liver ultrasound to screen for NAFLD. If the patients are found to have NAFLD, then the severity of the condition should be assessed using scoring such as NAFLD Fibrosis Score (NFS) to identify those who might be referred for liver biopsy. Regardless of their NAFLD status or NFS score, these patients should be targeted for aggressive lifestyle intervention and risk factor management.