Association of PCSK9 inhibitors with mortality: insights from a retrospective cohort analysis

Abstract

Aims

Proprotein convertase subtilisin/kexin type 9 (PCSK9) inhibitors are effective in reducing cardiovascular events, but their impact on all-cause mortality and medical utilization compared to statins is unclear. This study investigated PCSK9 inhibitor use and its impact on mortality and medical utilization vs. statins, using TriNetX database data with up to 9 years of follow-up.

Methods and results

This retrospective cohort study analysed TriNetX data spanning 1 July 2015, to 31 December 2023, including 79 194 PCSK9 inhibitor users (alirocumab, evolocumab, inclisiran) and 5 437 513 statin users with hyperlipidaemia. The primary outcomes were all-cause mortality and medical utilization, including hospital inpatient services, emergency department visits, critical care, and mechanical ventilation. Propensity score matching showed that PCSK9 inhibitor use was associated with a 28.3% lower risk of all-cause mortality [adjusted hazard ratio (aHR) 0.717, 95% confidence interval (CI): 0.673–0.763] and significant reductions in medical utilization (hospital inpatient services usage: aHR 0.692, 95% CI: 0.664–0.721; emergency department services: aHR 0.756, 95% CI: 0.726–0.788; critical care services: aHR 0.619, 95% CI: 0.578–0.664; and mechanical ventilation: aHR 0.537, 95% CI: 0.484–0.596) compared to statins. These findings were consistent across various demographics and clinical subgroups. The sensitivity analyses supported the robustness of the findings.

Conclusion

PCSK9 inhibitors significantly reduced all-cause mortality and medical utilization compared to statins, suggesting their important role in dyslipidaemia management, particularly for statin-naïve or intolerant patients. Further research, including randomized controlled trials, is needed to confirm these findings and explore the underlying mechanisms.

Introduction

Proprotein convertase subtilisin/kexin type 9 inhibitors (PCSK9i), namely alirocumab, evolocumab, and inclisiran currently used on the market, have emerged as significant advances in lipid-lowering therapy, particularly for patients with atherosclerotic cardiovascular disease (ASCVD) who are either intolerant to or inadequately managed by statins.¹ Systematic reviews and meta-analyses have shown PCSK9i effectively reduced the incidence of major adverse cardiovascular events (MACE), including myocardial infarction, stroke, and coronary revascularization.^2–5 However, these studies consistently found no significant reduction in all-cause or cardiovascular mortality, even with the substantial lowering of low-density lipoprotein-cholestrol (LDL-C) with additional high-intensity statins and ezetimibe.^3,4 The impact of PCSK9i on mortality remains a topic of ongoing investigation and debate.

Two key factors might account for the observed divergence between cardiovascular outcomes and mortality rates in the previous meta-analyses.^2–4 First, this could be due to the relatively short follow-up periods employed in the included randomized controlled trials (RCTs).² The median follow-up duration of 1.56 years may not suffice to reveal mortality benefits.² The benefits regarding mortality may become more apparent in longer-term follow-ups, as the cumulative risk reduction for major cardiovascular events may translate into improved survival rates.⁶ This delayed benefit underscores the importance of extending the duration of follow-up in cardiovascular RCTs to fully capture the impact of interventions on mortality.⁷

Second, the relatively low baseline LDL-C levels could obscure potential mortality benefits. A meta-regression within a meta-analysis revealed a significant association between intensive LDL-C lowering and reduced all-cause and cardiovascular mortality, specifically in trials where baseline LDL-C exceeded 100 mg/dL.⁸ Notably, trials with baseline LDL-C levels at or above 160 mg/dL demonstrated the most substantial reductions in all-cause mortality.⁸ It is plausible that patients with higher initial LDL-C levels, such as those with familial hypercholesterolaemia, might experience a more pronounced reduction in mortality. This highlights the necessity for further research with an extended follow-up period to pinpoint specific patient subgroups that may show the most significant mortality benefits from PCSK9i treatment.

Statin intolerance, with an estimated global prevalence of 9.1%, poses a significant barrier to effectively managing ASCVD and preventing subsequent cardiovascular events.⁹ Traditionally, rechallenging patients with either the same or a different statin following a washout period has been advocated as a viable strategy.¹⁰ Recently, the adoption of PCSK9i has been recognized as an alternative and efficacious approach according to consensus.¹⁰ However, robust, direct comparative evidence on mortality benefits between statins and PCSK9i is scarce. Research comparing medical utilization across these treatments is also limited.

To bridge this knowledge gap, we leveraged the comprehensive TrinetX clinical database to explore the longitudinal relationship between PCSK9i use and mortality, in comparison with statin use. We extended our follow-up period to up to 10 years and categorized patients based on their baseline LDL-C levels to provide a nuanced analysis.

Methods

Study design and data source

We conducted this study as a retrospective cohort analysis, leveraging the vast resources of TriNetX, the leading global platform for real-world data and evidence in life sciences and healthcare. TriNetX provides access to de-identified electronic health records from over 250 million individuals across more than 120 healthcare organizations worldwide.¹¹ For detailed information about TriNetX, please visit their website: https://trinetx.com/?mc_cid=7e2ecd5bc5&mc_eid=%5BUNIQID%5D. To ensure data integrity and reliability, TriNetX follows a stringent standardized framework focusing on conformance, completeness, and plausibility, underscoring its role in numerous prestigious studies.^12,13 Our study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.

Data extraction and analysis were conducted in January 2024, centring on the US Collaborative Network, a segment of TriNetX that includes 61 healthcare organizations. Our analysis was confined to data collected from 1 July 2015, to 31 December 2023, aligning with our research objectives.

Ethics statement

The TriNetX platform exemplifies data privacy adherence, strictly complying with the Health Insurance Portability and Accountability Act (HIPAA) and the General Data Protection Regulation (GDPR). It has received a notable waiver from the Western Institutional Review Board (WIRB), recognizing its commitment to utilizing only de-identified data. This emphasizes its goal of aggregating data and generating statistical summaries without jeopardizing individual privacy. Additionally, the Institutional Review Board of Chung Shan Medical University Hospital has approved our study's use of the TriNetX platform (Approval No: CS2-21176).¹⁴

Study subjects

In our study, we selected participants with at least two medical visits and diagnosed hyperlipidaemia (defined by International Classification of Diseases, Tenth Revision, Clinical Modification, ICD-10-CM code E78) at least twice during the study period.

These individuals were stratified into two cohorts. The first cohort, the ‘PCSK9i users,’ comprised individuals prescribed with PCSK9i, namely alirocumab (as recognized by the National Library of Medicine's normalized naming system for generic and branded drugs, RxNorm code 1659152), evolocumab (RxNorm code 1665684), and inclisiran (RxNorm code 2588243). For this group, the initiation of PCSK9i therapy marked their index date. In contrast, the cohort labelled ‘statin users’ comprised individuals who were newly prescribed statins (as confirmed by the Anatomical Therapeutic Chemical, ATC code C10AA). Notably, these participants had not received a statin prescription before 2015 (new users design). The index date for this group was marked by their inaugural statin prescription. We meticulously excluded those who transitioned between the two drug categories after the index date (switchers) or anyone below 18 years of age. Additionally, individuals who passed away on or before the index date were not included in our analysis.

Outcomes

Our primary outcome was all-cause mortality. The secondary outcomes focused on medical utilization, including hospital inpatient services (defined by Current Procedural Terminology, CPT codes 1013659, 1013699, or 1013729, or when the visit type was classified as inpatient for hospital inpatient services), emergency department services (CPT code 1013711), critical care services (CPT code 1013729), and mechanical ventilation (CPT codes 31500, 1015098, 1022227, or ICD-10-Procedure Coding System, ICD-10-PCS codes 5A1935Z, 5A1945Z, 5A1955Z, 0BH17EZ, 0BH18EZ, 0BH13EZ, or ICD-9-CM code 39.65). We monitored both cohorts over a period of 5 years, starting 30 days after the index date, to assess the risk of outcomes.

Covariates

In our study, we conducted a detailed analysis accounting for a broad range of covariates to mitigate potential confounders. We included demographic factors such as age, sex, race, and socioeconomic status. Lifestyle determinants like nicotine dependence or tobacco use and alcohol consumption were also considered. We assessed healthcare utilization across various services and meticulously categorized comorbidities using the ICD-10 classification. Medication use and pertinent laboratory results were also integrated into our analysis.

Statistical analyses

In our endeavour to address potential confounding, we employed TriNetX's built-in capability to generate propensity scores and conducted 1:1 matching using greedy nearest neighbour matching, as detailed in https://live.trinetx.com/tnx/study/195246/analytics/65af0e612c986f4bde38f9b5/outcomes/results (Supplementary material online, S1). During the matching phase, we introduced a calliper equal to 0.1 pooled standard deviations for variables like age at index, gender, race, socioeconomic indicators, lifestyle determinants, medical utilization, comorbidities, medication usage, and laboratory results. The comparability between the cohorts, both pre- and post-matching, was ascertained through standardized mean differences (SMD); an SMD under 0.1 was a testament to the balanced nature of our cohorts.

To extrapolate the likelihood of outcomes, we employed the Kaplan-Meier analysis. The hazard ratio (HR) alongside its confidence intervals (CI) and the test for proportionality were computed through the R's Survival package (version 3.2-3) by logging into the TriNetX platform Help Center:https://support.trinetx.com/hc/en-us/articles/360053133594-How-does-TriNetX-test-for-proportionality-on-a-hazard-ratio or referring to https://socialsciences.mcmaster.ca/jfox/Books/Companion/appendices/Appendix-Cox-Regression.pdf. The Log-Rank test, executed within the TriNetX environment, determined the potential variances in survival trajectories across the cohorts.

Seven subgroup analyses were performed based on sex (male or female), age groups (ranging from 18 to 64 years, or those 65 years and above), racial backgrounds (White vs. Black/African American), low-density lipoprotein (LDL) cholesterol levels (<130 or ≥130 mg/dL), PCSK9i medications mechanisms of action¹⁵ (monoclonal antibodies, like alirocumab or evolocumab, or those prescribed siRNA inhibitors, namely inclisiran), prior statin usage status, and prior comorbid MACE status.

To enhance the validity and consistency of our results, we performed three sensitivity analyses. First, we excluded patients who experienced acute coronary syndromes, cerebral infarctions, or underwent coronary therapeutic services and procedures from three months prior to the index date up to 1 month following the index date. Second, we adopted an intention-to-treat design to minimize any risk of bias that may arise from comparing groups with differing prognostic variables. Finally, in clinical practice, PCSK9i and statins are often prescribed concurrently, so we also included this clinical scenario in our analysis.

Results

Participant characteristics

We identified a cohort consisting of 79 194 individuals on PCSK9i and 5 437 513 individuals on statins. After applying propensity score matching (PSM), we successfully paired 42 285 PCSK9i users with an equal number of statin users. The methodology and flow of participant selection are depicted in Figure 1. The median follow-up duration for the PCSK9i cohort was 661 days (interquartile range, IQR: 919), and for the statins cohort was 1111 days (IQR: 1299).

Table 1 presents the characteristics of PCSK9i and statin users, both before and after PSM. Initially, PCSK9i users were generally older (mean age 65.9 vs. 63.0 for statin users) and more likely to be white (75.9% vs. 68.4%). After matching, age and racial distribution were closely aligned between groups, with mean ages of 65.9 for PCSK9i users and 65.9 for statin users, and white individuals comprising 75.9% of PCSK9i users and 76.3% of statin users. The standardized differences for all matched characteristics were below 0.1, indicating a high level of similarity between groups, except for variations in laboratory results. These adjustments ensure comparability between groups for assessing the impact of PCSK9i vs. statins on clinical outcomes.

Mortality risk and medical utilization risk

In our analysis, the utilization of PCSK9i compared to statin therapy demonstrated a significant reduction in all-cause mortality (Figure 2 and Table 2) and medical utilization. Specifically, PCSK9i users exhibited a 28.3% lower risk of all-cause mortality [adjusted hazard ratio (aHR) 0.717, 95% CI: 0.673–0.763, log-rank test P < 0.001], a notable decrease in hospital inpatient services usage (aHR 0.692, 95% CI: 0.664–0.721), emergency department services (aHR 0.756, 95% CI: 0.726–0.788), critical care services (aHR 0.619, 95% CI: 0.578–0.664), and mechanical ventilation (aHR 0.537, 95% CI: 0.484–0.596). The mean LDL cholesterol levels were similar between PCSK9i users and statin users, with values of 88.81 ± 51.58 mg/dL and 88.19 ± 37.67 mg/dL, respectively. A t-test showed no statistically significant difference between the two groups (t = 1.593, P = 0.111), indicating comparable LDL cholesterol control in users of PCSK9i and statins.

Subgroup analyses

Our study found PCSK9i significantly lower all-cause mortality and medical utilization compared to statins, exhibiting consistent benefits across various groups including gender (Supplementary material online, Table S1), age groups (Supplementary material online, Table S2), racial categories (Supplementary material online, Table S3), and LDL-C levels (Supplementary material online, Table S4). We conducted a subgroup analysis which demonstrated that the combination of PCSK9 inhibitors, with or without prior statin use, was associated with lower all-cause mortality and reduced medical utilization (Supplementary material online, Table S5). While monoclonal antibody PCSK9i notably decreased mortality and medical utilization in comparison to statins, inclisiran demonstrated less consistent effects with wide confidence intervals, suggesting distinct mechanisms of action on the cardiovascular system (Supplementary material online, Table S6). The efficacy of PCSK9i was similar regardless of recent MACE occurrence within the past year (Supplementary material online, Table S7). Figure 3 presents the forest plot for all-cause mortality across subgroup analyses.

Sensitivity analysis

Our findings remained consistent across different models after adjusting for multiple confounders (Supplementary material online, Table S8). Segmenting the follow-up into four periods (30 days to 1, 2, 3, and 4 years post-index date) affirmed the stability of our results (Supplementary material online, Table S9). Moreover, our conclusions remained valid after excluding patients with recent cardiovascular events or coronary interventions shortly after PCSK9i use (Supplementary material online, Table S10) or using an intention-to-treat design (Supplementary material online, Table S11). PCSK9i users who also took statins had a slightly lower risk of mortality compared to those who used statins alone, though this difference was not statistically significant (Supplementary material online, Table S12).

Discussion

The study revealed a substantial 28.3% reduction in all-cause mortality for patients newly treated with PCSK9i compared to those newly initiated on statins. Furthermore, it observed significant reductions in the utilization of hospital inpatient services, emergency department visits, critical care, and mechanical ventilation among PCSK9i users. The robustness of these findings was validated through detailed subgroup and sensitivity analyses, underscoring the potential of PCSK9i as a first-line therapy for managing dyslipidaemia.

Our study contributes to the existing literature on the long-term safety and benefits of PCSK9i, which remain inadequately characterized. The FOURIER (Further Cardiovascular Outcomes Research With PCSK9 Inhibition in Subjects With Elevated Risk) trial, with a median follow-up of 2.2 years, showed that evolocumab significantly reduced the risk of major cardiovascular events by 15% for the primary composite outcome and 20% for the key secondary outcome.⁶ However, no significant reduction in cardiovascular mortality was observed. The FOURIER-OLE (FOURIER Open-Label Extension) study, which extended the follow-up to a median of 5 years, confirmed the long-term safety and efficacy of evolocumab. The study showed a 23% lower risk of cardiovascular death among patients initially randomized to evolocumab compared to those assigned to placebo and then treating with evolocumab in the open-label period. This reduction was observed despite 63% of all subjects in the original placebo group achieving an LDL-C level of less than 40 mg/dL at 12 weeks after starting evolocumab.¹⁶ Evolocumab appears to have a legacy effect on cardiovascular mortality similar to that observed with statins.¹⁷

Similarly, the ODYSSEY OUTCOMES (Evaluation of Cardiovascular Outcomes After an Acute Coronary Syndrome During Treatment With Alirocumab) trial, with a mean follow-up of 2.8 years, demonstrated a lower all-cause mortality risk (HR = 0.85; 95% CI: 0.73–0.98) in patients with recent acute coronary syndrome, although subgroup analyses showed non-significant reductions in cardiovascular and non-cardiovascular deaths.⁷ Furthermore, in a pre-specified follow-up analysis of patients eligible for follow-up beyond 3 years, the benefit of alirocumab on all-cause mortality (HR = 0.78; 95% CI: 0.65–0.94) appeared more prominent.¹⁸ Our study's findings of reduced 5-year all-cause mortality further support the evidence that PCSK9i provide mortality benefits, regardless of baseline LDL level, prior statin use, monoclonal antibodies or SiRNA use, or a history of recent MACE. Therefore, a longer duration of PCSK9i treatment is necessary to achieve mortality benefits, highlighting the importance of early and proactive inclusion of high-risk patients in lipid-lowering treatment plans.

The significant reduction observed with PCSK9i compared to statins highlights the importance of the degree of LDL-lowering achieved. Statins have been the cornerstone of lipid-lowering therapy; however, a notable proportion of patients do not reach target LDL-C levels with statins alone or experience adverse effects, necessitating alternative treatments.^19,20 The FOURIER-OLE demonstrated that mean LDL-C levels dropped from 92 mg/dL at baseline to 30 mg/dL at 12 weeks after initiating evolocumab, representing a least squares mean reduction of 58.4% from baseline.¹⁶ The ODYSSEY OUTCOMES trial showed mean absolute LDL-C levels decreased from 122.8 mg/dL at baseline to 48.3 mg/dL at week 24 after starting alirocumab, reflecting a 61% reduction from baseline to 6 months.^18,21 In our study, after matching, the mean baseline LDL-C level remained higher in PCSK9i users (137.8 ± 55.61 mg/dL) compared to statin users (129.8 ± 45.15 mg/dL). Following treatment, LDL-C levels were similarly reduced in both groups, with PCSK9i users showing an LDL-C level of 88.81 ± 51.58 mg/dL and statin users 88.19 ± 37.67 mg/dL. This superior LDL-C lowering efficacy of PCSK9i may translate to a further decrease in cardiovascular events.²² Our study suggests that PCSK9i could potentially be used as first-line therapy for patients with dyslipidaemia, especially those with severe hypercholesterolemia or statin intolerance.

The observed reduction in medical resource utilization among PCSK9i users can be attributed to several key factors. First, PCSK9is significantly lowers LDL-C levels, thereby reducing cardiovascular events. The FOURIER trial demonstrated that evolocumab reduced cardiovascular events such as myocardial infarction, stroke, and the need for urgent or elective coronary revascularization.⁶ Additionally, the ODYSSEY OUTCOMES trial found that alirocumab significantly reduced MACE, including coronary heart disease events, cardiovascular events, non-fatal myocardial infarction, fatal or non-fatal ischaemic stroke, unstable angina requiring hospitalization, and ischaemia-driven coronary revascularization procedures.^7,23 The consistent reduction in MACEs and improved cardiovascular outcomes among PCSK9i users likely contribute to decreased hospital admissions, reduced emergency service needs, and lower utilization of critical care services and mechanical ventilation, as severe cardiovascular incidents often necessitate such interventions.

Limitations

Our study presents several limitations that warrant consideration. First, the use of real-world data within a retrospective cohort design inherently carries the risk of biases and confounding factors that may not be fully mitigated by PSM. The observational nature of this study limits our capacity to establish causality, particularly in comparison to RCTs. Moreover, the applicability of our findings may be constrained to specific populations or healthcare settings, given the database-specific nature of our cohort. Second, despite the comprehensiveness of the TriNetX database, it may not capture all relevant clinical details or health behaviours that significantly influence outcomes, such as diet, physical activity levels, or adherence to prescribed therapies. This lack of detailed personal health information could limit the depth of analysis regarding the interaction between PCSK9i or statin therapy and mortality or medical utilization. Third, the study's reliance on electronic health records and administrative data may introduce inaccuracies due to coding errors or documentation variability across healthcare organizations. Such discrepancies could impact patient cohort identification, outcome classification, and result interpretation. Fourth, the reason for the statin discontinuation in the PCSK9i group is not available in the database. Fifth, while the results were consistent between monoclonal antibody and siRNA treatments, the different mechanisms and observation periods associated with siRNA may necessitate extended follow-up to ascertain its definitive effectiveness. Lastly, our analysis does not provide mortality benefits related to specific diseases, such as cardiovascular or cancer mortality, due to the absence of detailed cause-of-death information in the database.

Future directions

Despite these challenges, our study's strength lies in its exploration of long-term mortality benefits associated with PCSK9i therapy, a previously underexamined area. We conducted multiple subgroup analyses to test various populations and performed several sensitivity analyses to account for different scenarios. These analyses provided a comprehensive understanding of the robustness and applicability of our findings across diverse groups and conditions, ensuring reliability and generalizability. Our findings highlight the significance of PCSK9i as a crucial option for managing dyslipidaemia, including for patients new to or intolerant of statin therapy. Future research should validate these results through prospective studies and RCTs, and investigate the mechanisms behind PCSK9 inhibitors’ benefits on mortality and morbidity to fully leverage their clinical utility.

Conclusions and implication

Our study, utilizing the comprehensive TriNetX database, revealed a significant association between the use of PCSK9i and reductions in all-cause mortality, hospital inpatient service utilization, emergency department visits, critical care, and mechanical ventilation, regardless of baseline LDL level, prior statin use, or a history of recent MACE. These findings suggest a potential role for PCSK9 inhibitors in dyslipidaemia management strategies, particularly for patients who are either statin-naïve or intolerant.

Funding

This work was supported by the National Science and Technology Council, Taiwan (grant number NSTC 112-2314-B-650-001-MY3, EDAHJ112009, EDAHS112032, EDAHS112024).

Conflict of interest: The authors have declared no conflicts of interests for this article.

Data availability

The data underlying this article are available in TriNetX, at https://live.trinetx.com/tnx/study/195246/analytics/65af0e612c986f4bde38f9b5/outcomes/results

Author contribution

C.H.H., S.I.W., and J.C.C.W. contributed to the design and implementation of the research. C.H.H. and S.I.W. developed the study flowchart, and S.I.W. performed the analyses. F.S.F., H.J.L., and J.C.C.W. verified the analytical methods. C.H.H. and S.I.W. wrote the draft manuscript with input from all authors. All authors discussed the results and contributed to the final manuscript.

References

Author notes