Patient selection

This study was based on data from Danish healthcare registries from 1 January 1995 till 1 January 2019. We used the National Patient Registry, which covers all Danish hospitals and contains data on all in- and outpatient contacts since 1977, and the National Presciption Registry, which contains data on all filled prescriptions at community pharmacies in Denmark since 1994 (Pottegård et al. 2017).

We used the National Patient Registry to establish a cohort of patients ≥ 18 years of age with a first-time diagnosis of ALD cirrhosis between 1 January 2000 and 1 January 2019; all diagnoses were identified by ICD-10 codes (Schmidt et al. 2015). Patients with a diagnosis of ALD cirrhosis between 1995 and 1999 were excluded. Patients entered the cohort on their ‘ALD date‘, which was defined as either the earliest discharge date of a hospital admission or the date of the first outpatient visit coded with an ALD cirrhosis diagnosis code (whichever came first). Patients were excluded if they had filled a prescription for lactulose before the ALD date. Usage of lactulose after the ALD date was identified using data from the National Presciption Registry. Of note, lactulose can be purchased over-the-counter, and we could not identify such purchases. Lactulose is a low-cost drug, with prices at about 20 euros per liter and for some patients, it is possible to have a subsidy which further decreases the cost (Kaplan 2021).

Study design Prevalence and cumulative incidence of lactulose use

All patients with ALD cirrhosis within our cohort were included to analyze the prevalence of lactulose use and the cumulative incidence of lactulose use from the ALD date. The ALD date represents the first possible date after ALD diagnosis on which patients were able to fill a prescription. (Fig. 1).

Fig. 1

Illustration of the study design

Figure 1 Prevalence and cumulative incidence of lactulose use: All patients with a diagnosis of ALD cirrhosis from 1 January 2000 to 1 January 2019 were followed from the date patients enters the cohort (ALD date) until end of follow-up to study prevalence and cumulative incidence of first-time lactulose use after ALD cirrhosis diagnosis. We excluded patients who had used lactulose before their ALD date. The prevalence at time t is the proportion of lactulose users at time t among those remaining in the cohort at time t. The cumulative incidence at time t is the proportion of patients who have initiated lactulose use up to time t among those who were in the cohort at the beginning of follow-up. Mortality: Cases and matched controls were followed from index date until death, censoring or end of follow-up. Patients who emigrated from Denmark were censored on the emigration date and were not allowed to re-enter the cohort if they later immigrated back to Denmark.

Initiation of lactulose in patients with ALD cirrhosis

In this analysis we used a nested case-control design, i.e., a case-control study was nested within our cohort of patients with ALD cirrhosis. The purpose of this design was to match cases (patients who initiate lactulose) and controls (patients who have not [yet] initiated lactulose) on time since cirrhosis diagnosis, which is a very strong predictor of mortality(Jepsen et al. 2010). With this study design, all cohort members are followed from the date of their ALD diagnosis, and time is expressed as ‘time since cirrhosis diagnosis’. When a member fills his or her first prescription for lactulose, he or she is identified as a ‘case’ (labeled “case #1”), and one other cohort member is selected at random as a ‘control’ (labeled “control #1”). The patient labeled “case #1” is removed from the cohort because he/she is no longer at risk of initiating lactulose, and all the others are followed onwards. Later, another patient fills his or her first prescription for lactulose and becomes “case #2”, and simultaneously “control #2” is drawn at random from the cohort. Then “case #2” is removed from the cohort, and the others are followed onwards, and so on until there are no more cohort members at risk of initiating lactulose (or follow-up ends, which happened for our cohort on 1 January 2019) (Fig. 1). After that, we have equal-sized groups of cases and controls and multiple ‘risk sets’, each of them consisting of two patients with ALD matched on time since cirrhosis diagnosis (case #1 and control #1 make up one risk set, etc.). The two members of a risk set share the same “index date”, which is the date on which the set’s case filled his or her first prescription for lactulose. The sampling procedure described here is called ‘risk set sampling’, also known as ‘incidence density sampling’ (Langholz and Clayton 1994). It is possible for a patient with ALD to be selected as a control in multiple risk sets, and it is also possible that a patient with ALD was not in contact with the healthcare system between their date of ALD diagnosis and their index date.

Mortality

This analysis included all lactulose users and their matched controls, in whom we analyzed the impact of lactulose usage on mortality from the index date.

Index date variables

Patient information on the index date was based on data from the National Prescription Registry and the National Patient Registry. It included diagnoses of chronic obstructive pulmonary disease (COPD), diabetes, cardiovascular disease (CVD), severe liver disease (defined as having a diagnosis code for one or more of the following: ascites, portal hypertension, spontaneous bacterial peritonitis or gastrointestinal bleeding) (Jepsen et al. 2008), HCC, and other cancers (Jepsen et al. 2008).

The date defining COPD was the earliest hospital diagnosis code of COPD. The date defining diabetes was the earliest of the following: a hospital diagnosis code of any type of diabetes or prescription of antidiabetic medicine. The date defining CVD was the earliest of the following: a hospital diagnosis code of arterial hypertension, atherosclerosis, or ischemic heart disease. The date defining severe liver disease was the earliest of the following: a hospital diagnosis code of either ascites, portal hypertension, hepatorenal syndrome, spontaneous bacterial peritonitis or gastoesophageal varices with or without bleeding (McPherson et al. 2016). The date defining HCC was the earliest hospital diagnosis code of HCC. The date defining other cancer was the earliest hospital diagnosis code of any type of cancer except HCC. All codes are listed in Supplementary Table 1.

Statistical analysis Prevalence of lactulose use

The prevalence of lactulose use was computed for each day between the ALD date and end of follow-up as the number of lactulose users in the ALD cirrhosis cohort on a given day divided by the total number of patients in the ALD cirrhosis cohort on that same day.

Cumulative incidence of lactulose use

The cumulative incidence of lactulose use was computed for each day between the ALD date and end of follow-up using the cumulative incidence function with death as a competing risk.

Initiation of lactulose use in patients with ALD cirrhosis

We used conditional logistic regression to study the association between sex, age and the defined index date variables and the incidence rate ratio of initiating lactulose treatment.

Mortality

We used Cox regression to examine the association between initiating lactulose and all-cause mortality, adjusting for sex, age and the above-mentioned index date variables. Cases and controls were followed from their index date until death or until censorship at the end of follow-up, 1 January 2019, whichever came first.