COVID-19 Severity and Mortality Among Chronic Liver Disease Patients

A Systematic Review and Meta-Analysis

Ramya Nagarajan, MD; Yuvaraj Krishnamoorthy, MD; Sathish Rajaa, MD; Vishnu Shankar Hariharan, MD


Prev Chronic Dis. 2022;19(8):e53 

In This Article


This was a systematic review and meta-analysis of observational studies and was performed according to the Preferred Items for Systematic Reviews and Meta-Analyses (PRISMA) statement guidelines.[14] The study protocol was registered in the PROSPERO database (registration ID: CRD42021291761).

Eligibility Criteria

We included studies with any of the following study designs: prospective or retrospective cohort, case control, and cross-sectional. Only published full-text studies were included; conference abstracts, unpublished data, and gray literature were excluded. Studies conducted among COVID-19 patients were included; studies among COVID-19 patients with comorbidities other than CLD were excluded.

Studies reporting the COVID-19 outcomes among CLD and non-CLD patients were included. CLD patients are diagnosed with the condition by clinical examination, laboratory or radiologic examination, or all 3 investigations. The CLD conditions most commonly found in COVID-19 patients included in our review were cirrhosis, viral hepatitis, NAFLD, and MAFLD. Studies reporting the diagnosis of CLD based on previous medical records were also included in the review.

Outcomes were the 1) severity of COVID-19 and 2) mortality due to COVID-19. The severity of the COVID-19 condition can be graded based on any of the following patient criteria: respiratory rate >30 breaths/min; oxygen saturation (SpO2) <93%; oxygenation index (PaO2/FiO2) ≤300 mm Hg; intensive care unit stay required; or mechanical ventilation.[15] Studies reporting any of the outcomes mentioned above were included in our review.

Search Strategy

We conducted a comprehensive, systematic, and extensive search in the electronic databases Medline, EMBASE, ScienceDirect, Google Scholar, and Cochrane Library. We selected the terms required for the search during the protocol stage. We used both the medical subject headings (MeSH) and free-text words while searching these databases. The keywords and their synonyms were searched using appropriate truncations, wildcards, and proximity searching. The terms used to search were "liver disease"/exp OR "hepatic disease":ti,ab OR "hepatic disorder":ti,ab OR "hepatopathy":ti,ab OR "liver cell disease":ti,ab OR "liver disease":ti,ab OR "liver diseases":ti,ab OR "liver disorder":ti,ab OR "liver illness":ti,ab) AND "coronavirus disease 2019"/exp AND ("mortality"/exp OR "excess mortality" OR "mortality" OR "mortality model" OR "disease severity"/exp OR "disease severity" OR "illness severity" OR "severity, illness" OR "cause of death"/exp OR "cause of death" OR "cause, death" OR "death cause" OR "death caused" OR "mortality cause" OR "death"/exp OR "death" OR "mortality". We also searched for crucial concepts using corresponding subject headings in each database. The last search was carried out by combining the individual search results using appropriate Boolean operators ("OR" and "AND"). The search was narrowed down using the available filters on the type of studies. We restricted the search from the inception of the pandemic to February 2022 and published in English only (Supplementary Table 1 available at: Bibliographies of the retrieved articles were also hand-searched to identify any themes missed during the database search.

Study selection process

This process involved 3 stages:

  1. Primary screening: Two independent investigators (R.N. and Y.K.) performed preliminary screening of title, abstract, and keywords by executing the literature search. Full-text articles were retrieved for the studies shortlisted on the basis of the eligibility criteria.

  2. Secondary screening: The same 2 investigators (R.N. and Y.K.) screened the full text of these retrieved studies and assessed them against the review's eligibility criteria. Studies that satisfied all the eligibility criteria concerning design, participants, exposure, and outcome were included.

  3. Finalizing the study selection: Disagreements during the screening process between the investigators were resolved. A final consensus on the inclusion of studies was reached with the help of another investigator (S.R.).

Data Extraction

Data were extracted manually from the included studies using a structured data extraction form that was developed and pilot tested during the protocol stage. We extracted the following data: general information, such as author and year of publication; information related to methods, such as study design, setting, sample size, sampling strategy, study participants, inclusion and exclusion criteria, outcome assessment method, and quality-related information; and information related to outcomes, such as patients' severity of disease and mortality rates. Data were entered by the investigator (S.R.), and the entry was double checked by the secondary investigator (V.H.).

Risk of Bias Assessment

Two independent investigators (S.R. and V.H.) used the Newcastle-Ottawa Scale to assess the risk of bias and quality of nonrandomized studies in meta-analyses under 3 domains: selection, comparability, and outcome.[16] The quality of the study was graded as good, fair, or poor based on the scores obtained under each domain.

Data Synthesis

We used Stata version 16 (StataCorp LLC) to conduct the meta-analysis. Because all outcomes were dichotomous, the number of events and participants in each group were entered to obtain the pooled effect estimate in terms of odds ratios (ORs) with 95% CIs and prediction intervals (PIs). We used the random-effects model with the restricted maximum likelihood method to calculate the weights of individual studies[17] because of the clinical and methodologic heterogeneity among the included studies. We used the command meta esize to compute the summary statistic; it automatically adjusts for zero cells by adding 0.5 to all cells in a 2-by-2 table that contains a zero value while computing the summary statistic. Evidence of between-study variance due to heterogeneity was assessed through the χ2 test of heterogeneity and I2 statistics to quantify the inconsistency. I2 less than 25% is mild, 25% to 75% is moderate, and more than 75% is considered substantial heterogeneity.[17] Study-specific and pooled estimates were graphically represented through a forest plot. We also performed a sensitivity analysis to assess the robustness of the results by removing the studies one at a time and checking for any significant variation in the results. We also performed subgroup analysis on the basis of each type of CLD.

We conducted univariable meta-regression with the study-level characteristics using the metareg package in Stata. Publication bias was assessed for each outcome using the funnel plot and Doi plot for visual interpretation and Egger test and Luis Furuya-Kanamori asymmetry index (LFK index) for statistical interpretation.[18] Asymmetry of the funnel plot and Doi plot and P value less than .10 in the Egger test indicates the possibility of publication bias. On the basis of the LFK index value, the possibility of publication bias was classified as no asymmetry (value within ±1), minor asymmetry (value out of ±1 but within ±2), and major asymmetry (value more than ±2).[18]