This analysis is embedded within the Canadian Longitudinal Study on Aging (CLSA), the largest prospective Canadian cohort study comprised of 51,338 community-dwelling individuals, aged 45–85 years at recruitment and followed every 3 years for 20 years or until death [21]. Detailed descriptions of the study population, protocols and data collection have been reported and may be accessed online [22]. Participants in the CLSA Comprehensive cohort (n = 30,097) were randomly selected from within 50 km of one of 11 data collection sites in seven provinces, and provided information on biological, medical, psychological, social, and lifestyle at in-home interviews. Of these, 27,170 (90.3%) provided blood samples and extensive laboratory measures, including clinical chemistry, genetics and metabolomic analysis, were performed on cryopreserved blood and plasma [23]. The current study is based on the CLSA Comprehensive cohort datasets for baseline (2012–2015) and the first two follow-ups (Follow-up 1; FUP1, 2015–2018 and Follow-up 2; FUP2, 2018–2022). This study was conducted on 12,934 participants according to sample availability and UGT genotypes (Supplementary Fig. 1).

The study included an equal number of male and female participants. Both overall and sex-based analyses were conducted, and we report the sex-related influence of UGT-deficiency on several age-related measures.

Details of treatment of blood samples and genotypic procedures can be accessed online [24]. The occurrence of UGT2B17 and UGT2B28 deletions was established based on the tag SNPs rs2708666 and rs11249532, which are in linkage disequilibrium with the respective deletions [9] and imputed from the genomic data as previously described [15]. The UGT2B17-deficient group (absence of both gene copies)were homozygous carriers of UGT2B28 gene. The UGT2B28-deficient group (e.g., absence of both gene copies) were homozygous carriers of UGT2B17 gene. The UGT-proficient group (reference group) were homozygous carriers of both genes as previously described [15]. The following analyses use the UGT-deficient participants as the exposure; their germline genetic status does not change with time so only ascertained at baseline.

We considered two outcomes related to lifespan: longevity defined a priori as survival age ≥ 75, and all-cause mortality defined as death from any cause. Participants who reached age 75 or more were classified as reaching longevity. This cutoff was used to distinguish mid-to-late and advanced stages of aging. Mortality status and date of death were obtained through participant status dataset given in CLSA FUP2 release.

Healthspan was characterized from a disease burden perspective, by comparing chronic conditions prevalence, incidence, and co-occurrence between UGT-deficient and UGT-proficient individuals. This analysis aimed to identify differences in multimorbidity patterns and disease accumulation across groups.

In CLSA, participants were asked to self-report in questionnaires their disease status for 35 chronic conditions that lasted or were expected to last at least 6 months and that were diagnosed by a health professional [25]. Identification of chronic conditions was further supplemented in the CLSA Comprehensive cohort by clinical measures (e.g. medication) and biomarkers. We sought to include diseases reported at baseline that fit into several criteria for inclusion, namely association with mortality, lower quality of life and/or chronic conditions with the highest disability-adjusted life-years (DALY) burden [26], previous association to UGTs [27, 28] and a generalized use in multimorbidity studies [29,30,31]. Chronic conditions were then grouped into 18 classes with sufficient pathophysiological similarity following recommendations [32] and International Classification of Diseases 10th revision (ICD-10) codes. These groupings were intended to balance clinical interpretability, statistical power, and feasibility in population-based analyses. Groupings included cardiovascular disease (e.g. heart attack), heart failure (e.g. congestive heart failure), stroke and transient ischemic attack, hypertension, diabetes, obesity, thyroid disorder (e.g. under- or over-active thyroid gland), osteoporosis, arthritis and/or rheumatoid arthritis (e.g. osteoarthritis), chronic musculoskeletal conditions causing pain or limitation (e.g. back problems), dementia and/or Alzheimer’s disease or memory problems, depression and/or anxiety, neurological problems (e.g. migraine, multiple sclerosis), kidney disease or failure, colon problem (e.g. bowel disorders/incontinence), stomach problem (e.g. stomach ulcer), chronic lung problems (e.g. asthma), cancer. Obesity was included as a chronic condition and was classified according to a body mass index (BMI) greater than 30 kg/m2. The list of individual conditions included in each class is provided in Supplementary Table 1. Multimorbidity was defined as having two or more than two prevalent diseases at CLSA baseline.

Allostatic load (AL) allows for the quantification of physiological dysregulation caused by life stressors that lead to molecular alterations at the cellular level, leading to accelerated aging and the development of diseases, premature morbidity and all-cause mortality [33,34,35]. The AL index was developed based on Statistic Canada guidelines applied to CLSA data as previously reported [36, 37]. It was established for each participant based on the commonly used quartile method considering the upper or lower quartile of the sample distribution for each of the nine following biomarkers at baseline: cardiometabolic and inflammatory biomarkers comprising total cholesterol, high density lipoprotein (HDL), glycated hemoglobin, waist-to-hip ratio, average systolic blood pressure, average diastolic blood pressure, average pulse rate, C-reactive protein and albumin [38, 39]. The AL index ranged from 0 to 9 and was the sum of all biomarkers per participant calculated as follows. One point was given for HDL and albumin if the participant’s value was below the 25th percentile whereas one point was given for values at or above the 75th [39]. AL index was treated as a continuous variable and dichotomized into high vs low, with high AL defined as an AL index score greater than the median.

To capture medication use, CLSA participants were asked to present all regularly scheduled or taken medications, which were then mapped to the World Health Organization (WHO) anatomical therapeutic chemical (ATC) classification. Polypharmacy was defined as five or more ATC Level 5 codes reported by participant at baseline.

Sociodemographic and health behavior factors known to influence the outcomes of interest were selected from the CLSA baseline dataset as covariates, namely age as a continuous variable, self-reported sex (male or female), education level (less than secondary school graduation; secondary school graduation, no post-secondary education; some post-secondary education and post-secondary degree/diploma) and marital status (single, never married or never lived with a partner; married/living with a partner in a common-law relationship, widowed, divorced and separated). Health behaviors included smoking status (current, former and never smoker) and alcohol intake (current, occasional and non-drinkers) as previously described [15]. BMI (kg/m2) was determined through measurement of height (m) and weight (kg) collected by trained CLSA following standard operating procedures and used as continuous variable. To control for chronic conditions, we assessed the individuals for presence or absence of at least one of the 18 classes of chronic conditions used in the exploration of multimorbidity. The number of medications was included as a continuous variable representing the total count of medications ATC Level 5 codes reported by the participants at baseline.

Descriptive statistics were used to summarize variables. For categorical variables, counts and proportions (percentages) were calculated. For continuous variables, the mean and standard deviation (SD), as well as the median and interquartile range (IQR), were reported. All the data were reported for each of the three study groups (UGT-proficient, UGT2B17-deficient and UGT2B28-deficient), both overall and by age (45–64 years and ≥ 65 years) and sex (female and male). Chi-square test for categorical variables and one-way analysis of variance F-tests for continuous variables were performed to compare UGT-proficient participants (reference group) with each of the UGT-deficient groups. When comparing prevalence of chronic conditions between groups, a false discovery rate (FDR) correction was applied to account for multiple pairwise comparisons performed in the baseline analyses. Raw P-values obtained from the statistical tests were adjusted using the Benjamini–Hochberg procedure, implemented from the "stats" package in R. This approach controls the expected proportion of false-positive findings among the significant results of pairwise comparisons. Raw P-values were reported in all results tables, whereas FDR-adjusted P-values were provided only where indicated.

Multivariable Cox proportional hazards models were used to estimate the association (Hazard Ratio (HR)) of each UGT-deficient group to all-cause mortality. In these models, age was used as the time scale and delayed entry was handled by specifying each participant’s age at study enrollment as the entry time and age at death or censoring as the exit time. Kaplan–Meier survival curves were constructed to provide a descriptive overview of time-to-event distributions across groups. Delayed entry (left truncation) was accounted for where applicable. Differences between groups were assessed using the global Score test derived from the Cox proportional hazards model. In the presence of left truncation, the Score test is considered analogous to the log-rank test and represents a statistically appropriate alternative to the classic log-rank approach [40]. The survival probabilities at ≥ 75 years were obtained from the all-cause mortality Cox models using the survfit() function. We used three regression models: the simplest model was adjusted for sex (and age as time scale); the semi-adjusted model was additionally adjusted for education (4 categories), smoking (3 categories), alcohol consumption (3 categories) and BMI (continuous); and the fully adjusted model was additionally adjusted for the presence of chronic conditions and number of medications. We also explored the interaction between UGT-deficiency and sex, by including an interaction term totaling 6 models.

Given that fewer incident cases than anticipated were observed for several issues during follow-up, limiting statistical power for prospective analyses, multivariable logistic regressions were conducted to estimate the cross-sectional associations (odds ratio (OR)) of UGT-deficiencies to individuals' diseases, multimorbidity, and AL. In these models, age was treated as a continuous variable, and the 18 classes of chronic conditions were excluded from the multimorbidity models. For individual diseases, we used an adjusted model with a reduced number of covariates (age, smoking (3 categories) and alcohol consumption (3 categories)) to avoid adjusting for potential intermediates.

Analyses were conducted using available data. Missing data was less than 5% for all variables and no imputation was performed. For descriptive purposes, occasional and current smokers were combined to comply with small cell size disclosure requirements. To assess the robustness of our findings, we conducted sensitivity analyses across outcomes. For multimorbidity and AL, we tested alternative definitions of multimorbidity (based on individual diseases, 3 diseases and ≥ 4 diseases) and AL as a categorical variable using clinical cut-offs. The survival and survminer packages were used for the Kaplan–Meier and Cox regression analyses and stats (base package) for the logistic regression models. Analyses were conducted using R (version 4.4.0) and P < 0.05 was considered as statistically significant.