Recent European guidelines recommend in patients with atherosclerotic cardiovascular disease to achieve a reduction of low-density lipoprotein-cholesterol of at least 50% if the baseline low-density lipoprotein-cholesterol level is between 1.8 and 3.5 mmol/L. Systematic reviews have associated a given statin/dose combination with a fixed percentage low-density lipoprotein-cholesterol response. Algorithms for detecting cases and estimating the prevalence of familial hypercholesterolaemia often rely on such fixed percentage reductions.
We used data from 915 coronary patients participating in the EUROASPIRE V study in whom atorvastatin or rosuvastatin therapy was initiated at hospital discharge and who were still using these drugs at the same dose at a follow-up visit 6 or more months later. Pre and on-treatment low-density lipoprotein-cholesterol levels were compared across the full low-density lipoprotein-cholesterol range. The prevalence of FH was estimated using the Dutch Lipid Clinic Network criteria, once using observed pre-treatment low-density lipoprotein-cholesterol and once using imputed pre-treatment low-density lipoprotein-cholesterol by following the common strategy of applying fixed correction factors to on-treatment low-density lipoprotein-cholesterol. Inter-individual variation in the low-density lipoprotein-cholesterol response to a fixed statin and dose was considerable, with a strong inverse relation of percentage reductions to pre-treatment low-density lipoprotein-cholesterol. The percentage low-density lipoprotein-cholesterol response was markedly lower at the left end of the pre-treatment low-density lipoprotein-cholesterol range especially for levels less than 3 mmol/L. The estimated prevalence of familial hypercholesterolaemia was 2% if using observed pre-treatment low-density lipoprotein-cholesterol and 10% when using imputed low-density lipoprotein-cholesterol.
The inter-individual variation in the percentage low-density lipoprotein-cholesterol response to a given dose of a statin is largely dependent on the pre-treatment level: the lower the pre-treatment low-density lipoprotein-cholesterol level the smaller the percentage low-density lipoprotein-cholesterol reduction. The use of uniform correction factors to estimate pre-treatment low-density lipoprotein-cholesterol is not justified.
Introduction
In patients with established atherosclerotic cardiovascular disease (ASCVD), controlling levels of low-density lipoprotein (LDL) cholesterol, as part of a comprehensive programme of cardiovascular prevention and rehabilitation, is essential to reduce the risk of recurrent events and improve survival. In order to achieve evidence-based LDL-cholesterol targets, the use of lipid-lowering drug therapies (LLTs) is recommended in all patients with ASCVD. According to recent European guidelines, a LDL-cholesterol level of less than 1.8 mmol/L should be achieved in these patients or, if the baseline LDL-cholesterol level is between 1.8 and 3.5 mmol/L a LDL-cholesterol reduction of at least 50%.1,2 Despite the availability and the use of potent lipid-modifying therapies, the EUROASPIRE surveys (European Action on Secondary and Primary Prevention by Intervention to Reduce Events) have documented that although the management of dyslipidaemia in clinical practice is improving over time, it is still suboptimal in coronary patients.3,4 Reasons for failing to achieve recommended targets are numerous, including adverse lifestyle, therapeutic inertia, use of less efficient cholesterol-reducing drugs, inadequate dose titration and low adherence but also inadequate perception of elevated LDL-cholesterol as a primary risk factor by patients as well as by physicians.5
In addition, several studies have reported on a considerable inter-individual variation in LDL-cholesterol response to a given dose of a statin or a PCSK9 inhibitor.6,–8 In an analysis of the VOYAGER database, the standard deviation of the percent LDL-cholesterol reduction ranged from 13% to 18% across statins and doses.6 Although 80 mg of atorvastatin daily is associated with a 55% LDL-cholesterol reduction in meta-analyses, 43% of patients on a daily dose of 80 mg atorvastatin in VOYAGER failed to achieve a 50% LDL-cholesterol reduction.6,9
In this study, based on real-world clinical data, we tested whether a given dose of a given statin results in a uniform percentage reduction in LDL-cholesterol across the entire pre-treatment LDL-cholesterol spectrum, as suggested by meta-analyses of statin trials.9,–11
Screening for familial hypercholesterolaemia (FH) in patients with premature ASCVD is recommended.1 To aid diagnosis, several algorithms have been proposed, all of which are based on the patients' pre-treatment LDL-cholesterol level.12,–14 As the latter are often not readily available in daily practice, correction factors have been developed from on-treatment LDL-cholesterol, for the specific type and dose of the current LLT to simulate pre-treatment LDL-cholesterol.15 As these correction factors are derived from back-transforming the fixed percentage changes as described by meta-analyses of clinical trials irrespective of pre-treatment levels, we also investigated to what extent the use of such imputed pre-treatment LDL-cholesterol values is justified and how this may affect the estimation of FH prevalence.
Methods
Study population and methodology
The design, methodology and main results of the EUROASPIRE V study have been published previously.4 In summary, 8261 male and female coronary patients aged 18–80 years from 27 European countries were examined and interviewed at least 6 months but not more than 2 years (average 1.1 years) following hospitalisation for one of the following first or recurrent clinical diagnoses or treatments: (a) elective or emergency coronary artery bypass grafting; (b) elective or emergency percutaneous coronary intervention; (c) acute myocardial infarction; (d) acute myocardial ischaemia. For the present analyses, patients were identified who, according to their medical record, had not been on LLTs prior to their recruiting event and in whom LDL-cholesterol levels were available, the latter considered as the ‘observed pre-treatment LDL-cholesterol levels’. Among them a particular subgroup was selected of 915 patients having been discharged from hospital on either atorvastatin 20, 40 or 80 mg/day or rosuvastatin 10 or 20 mg/day and who reported still taking these drugs at the same dose at the time of the follow-up visit. We only included patients in these analyses who reported 90% or greater compliance with the intake of these drugs. Venous blood samples obtained at the survey visit were analysed at the central laboratory (National Institute for Health and Welfare, Helsinki) according to standardised methods. The laboratory takes part in the lipid standardisation program organized by the Centers for Disease Control and Prevention, Atlanta, Georgia, USA and fulfils the requirements of the standard SFS-EN ISO/IEC 17025:2005. Serum LDL-cholesterol values (‘on-treatment LDL-cholesterol levels’) were calculated according to the Friedewald formula for triglyceride levels less than 4.5 mmol/L. As recommended by Besseling et al.,15 we derived ‘imputed pre-treatment LDL-cholesterol levels’ from these on-treatment LDL-cholesterol levels, through multiplication by a factor 100/(100–x) where x stands for the expected percentage changes published from the meta-analysis conducted by Law et al.9 In this comprehensive meta-analysis, the absolute and percentage LDL-cholesterol reductions were standardised to a usual pre-treatment LDL-cholesterol of 4.8 mmol/L9. Finally, we defined ‘probable or definite’ FH as a Dutch Lipid Clinic Network (DLCN) score of six or more.13 No information on corneal arcus and tendinous xanthomata was available.
Statistical methods
Distributions of pre and on-treatment LDL-cholesterol values as well as absolute and percentage changes were characterised by their means and standard deviations. The predicted percent LDL-cholesterol change at a fixed pre-treatment LDL-cholesterol level of 4.8 mmol/L was obtained from a linear regression model. Analyses were undertaken using SAS statistical software (release 9.4).
Ethical procedures
National coordinators for EUROASPIRE V were responsible for obtaining approvals from local ethics committees. Informed consent was obtained from each participant by means of a signed declaration.
Results
Low-density lipoprotein-cholesterol (LDL-C) changes in 467 patients initiated with atorvastatin 40 mg/day or rosuvastatin 20 mg/day.
Percentage low-density lipoprotein-cholesterol (LDL-C) changes in 467 patients initiated with atorvastatin 40 mg/day or rosuvastatin 20 mg/day.
Percentage low-density lipoprotein-cholesterol (LDL-C) changes in 467 patients initiated with atorvastatin 40 mg/day or rosuvastatin 20 mg/day according to pre-treatment LDL-C.
Observed and expected changes in LDL-cholesterol levels according to equipotent statin doses.
| Type of statin and dose | |||
|---|---|---|---|
| atorva 20 mg or rosuva 10 mg N = 161 | atorva 40 mg or rosuva 20 mg N = 467 | atorva 80 mg N = 287 | |
| Expected % change in LDL-cholesterola | –43% | –49% | –55% |
| Pre-treatment LDL-cholesterol (mmol/L), mean (SD) | 3.12 (0.96) | 3.44 (1.17) | 3.37 (1.12) |
| Post-treatment LDL-cholesterol (mmol/L), mean (SD) | 2.35 (0.86) | 2.20 (0.80) | 1.92 (0.66) |
| Absolute change in LDL-cholesterol (mmol/L), mean (SD) | –0.77 (1.06) | –1.24 (1.16) | –1.45 (1.09) |
| Percent change in LDL-cholesterol (%), mean (SD) | –20% (32%) | –31% (29%) | –38% (26%) |
| Mean change in LDL-cholesterol/mean pre-treatment LDL-cholesterolb | –25% | –36% | –43% |
| Predictedc at pre-treatment LDL-cholesterol of 4.8 mmol/L | |||
| Absolute change in LDL-cholesterol (mmol/L) | –1.96 | –2.27 | –2.59 |
| Absolute change in LDL-cholesterol (mmol/L)/4.8 mmol/Lb | –41% | –47% | –54% |
| Type of statin and dose | |||
|---|---|---|---|
| atorva 20 mg or rosuva 10 mg N = 161 | atorva 40 mg or rosuva 20 mg N = 467 | atorva 80 mg N = 287 | |
| Expected % change in LDL-cholesterola | –43% | –49% | –55% |
| Pre-treatment LDL-cholesterol (mmol/L), mean (SD) | 3.12 (0.96) | 3.44 (1.17) | 3.37 (1.12) |
| Post-treatment LDL-cholesterol (mmol/L), mean (SD) | 2.35 (0.86) | 2.20 (0.80) | 1.92 (0.66) |
| Absolute change in LDL-cholesterol (mmol/L), mean (SD) | –0.77 (1.06) | –1.24 (1.16) | –1.45 (1.09) |
| Percent change in LDL-cholesterol (%), mean (SD) | –20% (32%) | –31% (29%) | –38% (26%) |
| Mean change in LDL-cholesterol/mean pre-treatment LDL-cholesterolb | –25% | –36% | –43% |
| Predictedc at pre-treatment LDL-cholesterol of 4.8 mmol/L | |||
| Absolute change in LDL-cholesterol (mmol/L) | –1.96 | –2.27 | –2.59 |
| Absolute change in LDL-cholesterol (mmol/L)/4.8 mmol/Lb | –41% | –47% | –54% |
LDL: low-density lipoprotein.
According to Law et al.9
Expressed as percentage.
Predicted from linear model.
Observed and expected changes in LDL-cholesterol levels according to equipotent statin doses.
| Type of statin and dose | |||
|---|---|---|---|
| atorva 20 mg or rosuva 10 mg N = 161 | atorva 40 mg or rosuva 20 mg N = 467 | atorva 80 mg N = 287 | |
| Expected % change in LDL-cholesterola | –43% | –49% | –55% |
| Pre-treatment LDL-cholesterol (mmol/L), mean (SD) | 3.12 (0.96) | 3.44 (1.17) | 3.37 (1.12) |
| Post-treatment LDL-cholesterol (mmol/L), mean (SD) | 2.35 (0.86) | 2.20 (0.80) | 1.92 (0.66) |
| Absolute change in LDL-cholesterol (mmol/L), mean (SD) | –0.77 (1.06) | –1.24 (1.16) | –1.45 (1.09) |
| Percent change in LDL-cholesterol (%), mean (SD) | –20% (32%) | –31% (29%) | –38% (26%) |
| Mean change in LDL-cholesterol/mean pre-treatment LDL-cholesterolb | –25% | –36% | –43% |
| Predictedc at pre-treatment LDL-cholesterol of 4.8 mmol/L | |||
| Absolute change in LDL-cholesterol (mmol/L) | –1.96 | –2.27 | –2.59 |
| Absolute change in LDL-cholesterol (mmol/L)/4.8 mmol/Lb | –41% | –47% | –54% |
| Type of statin and dose | |||
|---|---|---|---|
| atorva 20 mg or rosuva 10 mg N = 161 | atorva 40 mg or rosuva 20 mg N = 467 | atorva 80 mg N = 287 | |
| Expected % change in LDL-cholesterola | –43% | –49% | –55% |
| Pre-treatment LDL-cholesterol (mmol/L), mean (SD) | 3.12 (0.96) | 3.44 (1.17) | 3.37 (1.12) |
| Post-treatment LDL-cholesterol (mmol/L), mean (SD) | 2.35 (0.86) | 2.20 (0.80) | 1.92 (0.66) |
| Absolute change in LDL-cholesterol (mmol/L), mean (SD) | –0.77 (1.06) | –1.24 (1.16) | –1.45 (1.09) |
| Percent change in LDL-cholesterol (%), mean (SD) | –20% (32%) | –31% (29%) | –38% (26%) |
| Mean change in LDL-cholesterol/mean pre-treatment LDL-cholesterolb | –25% | –36% | –43% |
| Predictedc at pre-treatment LDL-cholesterol of 4.8 mmol/L | |||
| Absolute change in LDL-cholesterol (mmol/L) | –1.96 | –2.27 | –2.59 |
| Absolute change in LDL-cholesterol (mmol/L)/4.8 mmol/Lb | –41% | –47% | –54% |
LDL: low-density lipoprotein.
According to Law et al.9
Expressed as percentage.
Predicted from linear model.
The prevalence of probable or definite FH according to the Dutch Lipid Clinic Network criteria in the total group of 915 patients was 2% (18/915) if based on observed pre-treatment LDL-cholesterol levels.13 However, when imputed LDL-cholesterol levels were used instead of observed pre-treatment LDL-cholesterol levels, this prevalence rose to 10% (89/915). Moreover, only 10% of these 89 patients were actually considered as probable or definite FH cases based on their observed pre-treatment LDL-cholesterol levels.
Discussion
Meta-analyses of clinical trials of LLTs have clearly indicated that a range of statins in varying doses are very effective in lowering LDL-cholesterol and reducing cardiovascular risk.9,10 Clinical guidelines established in the past decades have mainly focused on absolute risk reductions in order to achieve specific recommended LDL-cholesterol targets. Given the evidence from pooling randomised trials of patients with known atherosclerotic disease, suggesting that the relative reduction in LDL-cholesterol adds incremental prognostic value over the LDL-cholesterol attained, more recent guidelines advocate the use of percentage reductions from a baseline LDL-cholesterol before treatment was initiated.16 Achieving a reduction of at least 50% in LDL-cholesterol in very high-risk patients has been re-commended in Europe if the baseline LDL-cholesterol level is between 1.8 and 3.5 mmol/L.1,2 High-intensity statins will therefore be needed in the vast majority of these patients.17
The present data confirm the considerable variability between patients in their LDL-cholesterol response to the same type and dose of a statin. This inter-individual variability may reflect unidentified changes in concomitant drug therapies, errors in drug administration or mistakes in blood sample handling or in assays in the laboratory; the genetic determinants of the untreated LDL-cholesterol level and of the responsiveness of LDL-cholesterol to a given dose of a statin have been well documented in Mendelian randomization-based studies.18,,,–22 Behavioural factors such as lifestyles or therapeutic compliance may also explain part of the observed heterogeneity in statin response. In order to control for the latter source of variation, we only included patients in our analyses who reported a 90% or greater compliance with taking their LLTs.
Our analyses also clearly demonstrate that a substantial fraction of the inter-individual variability in percentage LDL-cholesterol response to treatment is directly related to differences in baseline LDL-cholesterol values. In contrast to the constant percentage changes in LDL-cholesterol suggested in systematic reviews of statin trials, we found that the percentage change is not fixed but varies substantially according to pre-treatment LDL-cholesterol, with systematically lower percentage reductions in the lower pre-treatment LDL-cholesterol range. For instance, the mean percentage LDL-cholesterol reduction in response to atorvastatin 40 mg/day or rosuvastatin 20 mg/day was 31%, substantially less than the predicted 48–49% from the literature.9 This observed reduction was even lower than 20% for baseline LDL-cholesterol levels below 3 mmol/L. This discrepancy between the observed and predicted LDL-cholesterol response to statins is mainly due to the fact that baseline LDL-cholesterol levels in our sample of stable ASCVD, left untreated until the recruiting event occurred, were relatively low (3.4 mmol/L on average) in comparison to the much higher baseline LDL-cholesterol levels of participants included in the, mainly older, statin trials. In the meta-analysis of Law et al., predicted LDL-cholesterol reductions were standardised to a pre-treatment LDL-cholesterol concentration of 4.8 mmol/L.9 In our analyses, the 47% percentage LDL-cholesterol change estimated from a linear model predicted for a pre-treatment level of 4.8 mmol/L was very close to the expectation from Law's meta-analysis. In line with our observations, low baseline LDL-cholesterol levels in the VOYAGER study were found to be a strong predictor of suboptimal response analysis: 14% of the patients with a baseline LDL-cholesterol of less than 3.6 mmol/L had a suboptimal response compared to just 3% of the patients with a baseline LDL-cholesterol greater than 5.2 mmol/L.6
Screening for FH in patients with premature ASCVD is recommended. Evidently, in all currently available screening models most weight is given to untreated LDL-cholesterol levels.12,13 However, the large majority of ASCVD patients are on LLTs and many are unaware of their pre-treatment lipid levels. This may also be the case for their treating physicians faced with incomplete (electronic) health records in terms of historical LDL-cholesterol levels. To illustrate further the clinical relevance of our findings, we have shown that the imputation of pre-treatment LDL-cholesterol values using unique correction factors calculated from fixed percentage reductions found in the literature and applied to on-treatment LDL-cholesterol, is not justified and inevitably leads to a substantial overestimation of the number of individuals with potential FH. In our analyses, the likelihood of being identified as having probable or definite FH according to the commonly used DLCN criteria was found to be five times higher when using imputed pre-treatment LDL-cholesterol. The imputation method has been validated in patients with confirmed FH but again, at very high baseline LDL-cholesterol levels of 6.3 mmol/L on average.23
Fixed percentage LDL-cholesterol reductions have also been assumed in the context of pharmaco-economic evaluations of lipid-modifying treatments and in modelling studies at the population level.24 Our results indicate that the cost-effectiveness of LLTs may have been overestimated in populations with LDL-cholesterol levels at lower parts of the spectrum.
More advanced statistical techniques for imputing unknown pre-treatment measurements have been proposed.25 However, estimates from parametric models cannot account for the substantial inter-individual variation of the LDL-cholesterol response to LLTs.
Our study has some limitations. It is a retrospective analysis of a subset of the EUROASPIRE V survey but this was a good existing database for studying the research question using real-world data. In our sample of patients not on LLTs before the recruiting event, we interpreted the LDL-cholesterol levels at hospitalisation found in the medical notes as the ‘observed pre-treatment LDL-cholesterol levels’. Lipid levels may fall after an acute coronary syndrome or surgery and we do not know when LDL-cholesterol was measured during hospitalisation although it was likely to be at the time of admission. However, we did not observe differences in pre-treatment LDL-cholesterol levels across the four diagnostic categories we selected in EUROASPIRE V (results not presented). Also, in the time period of 6 months to 2 years between the start of LLTs and the survey visit, factors other than LLTs may have influenced LDL-cholesterol values, but it is rather unlikely that any such bias would differentially influence our analyses. Finally, pre-treatment LDL-cholesterol levels were taken from the medical notes and therefore were based on local laboratory data while on-treatment blood samples were analysed in one central laboratory. However, given the good internal and external quality assessment of the EUROASPIRE V central laboratory, measurement bias in the lower ranges of the LDL-cholesterol spectrum seems unlikely.
In conclusion, the percentage LDL-cholesterol response to a given dose of a statin is not fixed across the whole pre-treatment LDL-cholesterol range. Using a specific LDL-cholesterol goal in addition to the percentage reduction remains warranted to address further the treatment gap in dyslipidemia. For patients treated with LLTs, the use of uniform correction factors to estimate pre-treatment LDL-cholesterol levels is not justified. As a consequence, the prevalence of FH according to common screening algorithm models is largely overestimated. We strongly recommend that all adults should have at least one LDL-cholesterol level available in their routine medical record because this ‘untreated’ LDL-cholesterol value is so important for later therapeutic and diagnostic purposes. The results from this study also emphasise the need for a stepwise approach in reaching LDL-cholesterol goals by identifying those patients who may need dose adjustments and/or a combination of lipid-lowering drugs.
Author contribution
DDB and GDB contributed to the design, data acquisition, data analysis, interpretation, drafting and revision of the manuscript. KK, LR and DW contributed to the design, data acquisition and interpretation and critically revised the manuscript. DDS, ŽR, EC and LT contributed to the interpretation and critically revised the manuscript. All authors gave final approval and agreed to be accountable for all aspects of the work ensuring integrity and accuracy.
Acknowledgements
EORP Oversight Committee, Executive Committee and Steering Committee of the EUROASPIRE V study. Data collection was conducted by EORP with E Fiorucci as project officer, V Missiamenou and F Larras as data managers. The authors are grateful to the administrative staff, physicians, nurses and other personnel in the hospitals in which the survey was carried out and to all participating patients.
Declaration of conflicting interests
The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: there were no competing interests for D De Bacquer, D De Smedt, E Clays, D Wood and G De Backer. Ž Reiner has received honoraria for lectures from Sanofi Aventis and Akcea. L Tokgözoğlu has had financial interests, arrangements or affiliations with Actelion, Amgen, Sanofi, Pfizer, Novonordisk, MSD, Recordati, Kowa, Abbott, Novartis, Mylan, Bayer, Servier and Sanovel. K Kotseva received consultancy fees from Amgen. L Rydén has received honoraria and research grants from Amgen, Bayer AG, Boehringer Ingelheim, MSD, Novo Nordisk and Sanofi.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: the EUROASPIRE V survey was carried out under the auspices of the European Society of Cardiology, EURObservational Research Programme. The survey was supported through research grants to the European Society of Cardiology from Amgen, Eli Lilly, Sanofi, Pfizer, Ferrer and Novo Nordisk. The sponsors had no role in the study design, data collection, data analysis, data interpretation or decision to publish the manuscript.
Comments