Background

Cardiovascular disease (CVD) is a leading cause of death and disability [1, 2]. Elevated levels of low-density lipoprotein cholesterol (LDL-C) are a major contributor to atherosclerosis leading to subsequent CVD events. Numerous clinical trials of lipid lowering drugs have found that reducing LDL-C levels substantially reduces the risk of CVD [3–5] suggesting a strong direct relationship between plasma LDL-C levels and CVD outcomes [6, 7]. Many people, however, still have residual CVD risk and suffer from CVD events despite significant LDL-C lowering. In addition to LDL-C, other risk factors are likely to influence residual cardiovascular risk. Among these, lipoprotein(a) [Lp(a)], has been proposed to be independently associated with CVD [8].

Lp(a) is an low density lipoprotein (LDL) particle which is attached to the polypeptide, apolipoprotein(a) [apo(a)] [9]. Apo(a) exists in multiple forms or ‘kringles’, which give rise to different Lp(a) isoforms. Apo(a) is also believed to be responsible for the anti-fibrinolytic properties of Lp(a) [9]. Further biomechanisms behind the Lp(a) and CVD relationship may also involve prothrombotic or proatherosclerotic processes, or a combination of the two [10].

Lp(a) may be measured using a variety of different assays. However, the reliability of many of the assays is questionable, due to their poor abilities at detecting the multiple molecular isoforms of Lp(a). Consequently, some assays (isoform dependent) that measure Lp(a) mass cannot distinguish between high and low molecular weight apo(a) isoforms, whilst others (isoform independent) can. However, to our knowledge, at present there appear to be no Lp(a) assays that are both isoform independent and suited for use clinical laboratories [11]. This problem has led to poor standardisation and comparability with respect to the Lp(a) values recorded by different assays, which in turn hampers comparisons between trials assessing the relationship between Lp(a) and CVD [12]. Despite this clinical trials have shown that Lp(a) is a risk factor in patients on long-term statin treatment [13, 14]. Evidence from the Atherothrombosis Intervention in Metabolic Syndrome with Low HDL/High Triglycerides: Impact on Global Health Outcomes (AIM-HIGH) trial suggests that Lp(a) is a predictor of CVD events in patients with normal LDL-C levels [15] and recent studies have suggested that elevated Lp(a) levels like elevated LDL-C, could be associated with premature CVD [8]. Extensive research exists to support an association between Lp(a) and CVD events with respect to the primary prevention of events in the general population [16, 17]. This relationship appears to be independent of LDL-C, other lipid levels such as high density lipoprotein (HDL) and the presence of other cardiovascular risk factors [18]. Some evidence from pooled analyses of prospective studies [19–21] suggests a potential association between Lp(a) and risk of CVD among high risk and secondary prevention population. The availability of new data from recently published clinical studies has prompted the need for a more contemporary systematic review. This review assesses the relationship between Lp(a) and CVD outcomes, with particular emphasis on high cardiovascular risk populations (high risk primary prevention and secondary prevention). The review also focuses on the best available evidence from studies that used multivariable analysis methods to control for the effect of confounding variables.

Methods

To reduce the risks of bias and error, this review adhered to a pre-specified protocol and methods recommended by the Cochrane Collaboration [22], and the Centre for Reviews and Dissemination (York, United Kingdom) [23] which are widely regarded as ‘gold standard’ methodologies.

This review included prospective studies, which assessed the relationship between Lp(a) and CVD outcomes. These studies included randomised controlled trials (RCTs), cohort studies and nested case-control studies. Eligible populations were any adult (≥18 years) population regardless of baseline CVD risk, gender, age and ethnicity. Studies had to follow patients for at least 12 months. No restrictions were placed on the CVD outcome or the type of Lp(a) assay. However, studies were required to use a multivariable analytical approach, which assessed the effect of Lp(a) on CVD outcomes after adjusting for other confounding factors. At a minimum the analysis had to adjust for baseline age and gender; but this restriction was relaxed for studies in gender and age subgroups or in nested case-control studies where cases and controls were matched on age and gender. Studies were excluded from the review if they fail to clearly report effect sizes based on relevant multivariable analyses.

Extensive literature searches were performed using search strategies developed by an Information Specialist (full strategies are available in Additional file 1). A total of six electronic databases were searched from inception to 31 December 2015 including: MEDLINE, Embase, Medline In-Process & Daily Update, Cochrane Central Register of Controlled Trials (CENTRAL), Database of Abstracts of Reviews of Effects (DARE) and the Cochrane Database of Systematic Reviews (CDSR). Search strategies were refined and adapted according to the configuration and requirements of each database. The final strategies combined relevant search terms comprising indexed keywords (e.g. Medical Subject Headings, MeSH and EMTREE) and free text terms appearing in the title and/or abstract of database records. Search terms were identified through discussion between the review team, by scanning background literature and ‘key articles’ already known to the review team, and by browsing database thesauri. Literature searches were not limited by date, language or publication status. Supplementary searches were undertaken in two trials registers (National Institutes of Health [NIH] ClinicalTrials.gov and International Standard Randomised Controlled Trial Number [ISRCTN] Registry) and conference abstracts from four major cardiovascular disease conferences (European Atherosclerosis Society Congress; European Society of Cardiology Congress; American College of Cardiology Annual Scientific Session; and American Heart Association Annual Scientific Sessions, for years 2011-2015). The reference lists of included studies and systematic reviews were checked for further studies. Identified references were downloaded in Endnote X6 software (Thomson Reuters, New York) for further assessment and handling, and duplicate records were removed.

The study selection process was performed by two reviewers working independently. Data were extracted into a specifically developed spreadsheet in Excel 2010 (Microsoft Corporation, Redmond, Washington). One reviewer extracted the study data and a second reviewer independently reviewed the data against the original paper for completeness and accuracy. Data were extracted on the baseline population (e.g. race and previous CVD events), Lp(a) assay (e.g. isoform independence), CVD outcomes, statistical analysis methods (details of the type of multivariable model and the variables included) and effect sizes for the relationship between Lp(a) and CVD outcomes. The methodological quality (risk of bias) of each study was assessed using the criteria of the Quality in Prognostic Studies (QUIPS) tool [24]. The quality assessments were performed independently by two reviewers. Any discrepancies between reviewers during data extraction or quality assessments were resolved through consensus or consultation with a third reviewer.

Meta-analysis was not possible due to heterogeneity in the CVD outcomes, populations and statistical analysis methods. Studies have been summarized in a narrative synthesis accompanied by data tables. It was not possible to plot data on Forest plots due to the absence of necessary data. Effect sizes for Lp(a) are reported as odds ratios (ORs), hazard ratios (HRs) or adjusted Lp(a) levels with accompanying 95 % confidence intervals (CIs) or means/interquartile ranges (IQR). Studies are grouped according to CVD outcome, population and variables included in the multivariable model(s). The term “positive” association refers to an increase in Lp(a) or higher Lp(a) levels resulting in an increased risk of CVD outcomes. Similarly a “negative” association refers to an increase in Lp(a) or higher Lp(a) levels resulting in a decreased risk of CVD outcomes.

Results

Literature searches of electronic databases and other sources including hand searching retrieved 3837 titles/abstracts through December 2015. After de-duplication, a total of 2850 titles/abstracts were screened, and 2189 papers were excluded as having no relevance to the review. Full papers of 312 potentially relevant references were selected for further examination. Of these, 197 papers were excluded after further examination for the following reasons: do not report relevant prognostic factors (20 papers), not relevant outcome (32 papers), not relevant study design (103 papers), and no clearly reported multivariable analysis (42 papers). A total of 60 studies (115 papers) met the criteria for inclusion in the review.

A summary of the identification and selection of studies for inclusion in this review is presented in Fig. 1, in accordance with the PRISMA [25].

Fig. 1
figure 1

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram for the selection of studies

Full size image

The 60 studies included ten RCTs, 37 prospective cohort studies and 13 nested case control studies. A summary of the studies is reported in Table 1 and further information about individual studies is available in an Additional file 2: Table S1.

Table 1 Summary of characteristics across included studies (60 studies)
Full size table

Overall, across all 60 studies, the level of bias was assessed as moderate. However 13 studies were assessed as having a high risk of bias and five studies as a low risk of bias; 11 studies failed to report sufficient detail so as to allow a full assessment of the risk of bias. Those studies that were of high risk of bias often had methodological issues within the QUIPS domains 2 and 3 concerning study attrition and prognostic factor [i.e. Lp(a)] measurement respectively. Reporting across studies was not always sufficiently detailed to allow a judgement to be made and 16 studies were reported as not having enough information to make a judgement for at least one of the six QUIPS criteria. Further details of the risk of bias assessments for the individual studies are available in a Additional file 3: Table S2.

Primary prevention studies

The majority of the identified studies (39 studies) were carried out in participants from the general population, i.e. did not select patients based on their baseline history or risk of CVD events. Further details of the characteristics of the studies in the general population are reported in Table 2. These included four RCTs, 25 prospective cohort studies and 10 nested case-control studies. Thirteen studies [26–38] were conducted in males and two in females [39, 40]. Specific ethnic groups were used in some studies including populations from South Korea [41, 42], Native American Indians (from the USA) [43], Japan [44, 45] and Taiwan [46]. Follow-up in the studies tended to be longer than in the high risk and secondary prevention populations, with 20 out of 39 studies (52.3 %) having a follow-up of 5 to ≤ 10 years and 11 out of 39 studies; 28.2 %) following participants for over 10 yrs. The longest follow-up period was 20 years in the ARIC study [47]. The risk of bias across the 39 studies was assessed as low in two studies (5.4 %); moderate in 22 studies (84.0 %), high in seven studies (17.9 %) and there was insufficient information to make an assessment in eight (20.5 %) studies. Only 12 (30.8 %) studies used assays that were reported as isoform independent and five (12.8 %) used assays on fresh plasma samples. The majority of studies used a Cox proportional hazards model (20 studies; 51.3 %); other models included logistic regression (11 studies; 28.2 %), conditional logistic regression (seven studies; 17.9 %), and both conditional and unconditional logistic regression (one study; 2.6 %).

Table 2 Summary of primary prevention studies in the general population (39 studies)
Full size table

Half of the studies included LDL-C (19 studies; 48.7 %) as a covariate in the multivariable model and all but four of these studies (Justification for the Use of Statins in Primary Prevention: An Intervention Trial Evaluating Rosuvastatin trial [JUPITER] [14], PRospective du l’Infarctus MyocardE; prospective epidemiological study of myocardial infarction [PRIME] [34], Invecchiare in Chianti [InCHIANTI] Study [48] and Atherosclerosis Risk in Communities [ARIC] [49]) reported significant positive associations between Lp(a) and CVD. The JUPITER study [14] and PRIME [34] each reported two negative associations between Lp(a) and CVD events, but in both cases these were for subgroup analyses and the results were not statistically significant. In the case of the JUPITER study [14] the authors concluded that among white participants treated with potent statin therapy, Lp(a) concentrations (at baseline and on-statin) were a significant determinant of residual risk with respect to CVD events; and in the PRIME study [34] that increased baseline Lp(a) levels (considered as the Lp(a) cholesterol content) were significantly associated with the risk for MI and angina pectoris, especially in men with high LDL-C. In the ARIC study [49] a significant negative association with Lp(a) mass was reported for a subgroup analysis of ischemic strokes in white females, when comparing >6.6 to ≤ 14.6 mg/dL Lp(a) [Quintile 4] versus 0.1 to ≤ 1.6/dL Lp(a) [Quintile 1]. However, this result was based on a small number of events suggesting that it may not be robust and was in contrast to the overall trend of the other primary and subgroup analyses, which suggested that overall a positive association existed between Lp(a) mass and CVD events. Overall, the authors concluded that Lp(a) mass was positively associated with CVD events, but that it appeared stronger in blacks compared to whites. The InCHIANTI study [48] did not report any statistically significant associations after following patients longitudinally for six years, but did find evidence of a link between prevalent peripheral arterial disease (PAD) and Lp(a). The authors concluded that Lp(a) concentration was an independent predictor of PAD in the cross-sectional evaluation, but that further larger, longer duration, prospective studies are needed to establish a longitudinal association.

Among the remaining half of the studies which did not include LDL-C as a covariate in their multivariable models, a further four studies (Physician’s Health Study [PHS] [33, 50], Italian Longitudinal Study on Aging [ILSA] [51] Uppsala Longitudinal Study of Adult Men [USLM] [26] and Women’s Health Initiative Observational Study/Hormones and Biomarkers Predicting Stroke Studies[WHI-OS/HaBPS] [39]) reported some negative associations between Lp(a) mass and CVD events. Negative statistically insignificant associations in the USLM study [26] were reported for the relationship between Lp(a) mass and intracerebral haemorrhage, but the authors concluded that high serum Lp(a) level independently predicted fatal and non-fatal stroke/transient ischemic attack (TIA) in a population of middle-aged men followed for 32 years. In the ILSA study [51] no overall statistically significant association was found between high Lp(a) levels and the risk of all-cause mortality, cumulative fatal–nonfatal stroke, and cumulative fatal–nonfatal coronary artery disease (CAD) events. However, the authors reported that high Lp(a) levels were an independent and significant predictor of non-fatal CAD events after 6.3 years in an elderly (65 to 84 years) population [51]. No association was also reported for Lp(a) concentration and ischemic stroke in the WHI-OS/HaBPS [39] study, where the authors concluded that they found no significant relationship between Lp(a) and ischemic stroke in postmenopausal women. However, the methods used to measure Lp(a) were not well described in this study and so their reliability was unclear. The PHS study [33] also found no evidence that Lp(a) levels were a significant predictor of PAD in men.

High risk primary prevention studies

Seven studies assessed the relationship between Lp(a) and CVD outcomes in populations at high risk of CVD events, but who had not as yet experienced a CVD event (Agewall 2002; [52] Choices for Healthy Outcomes in Caring for End Stage Renal Disease study [CHOICE]; [53] Cleveland Clinic Hemodialysis Cohort; [54] Diamant Alpin Collaborative Dialysis Cohort; [55] Japan Diabetes Complications Study [JDCS]; [56] Koda 1999; [57] and Zimmermann 1999 [58]). Follow-up in the studies ranged from 2 years [14, 55, 57] to 7.8 years [56] and the sample size ranged from 118 [59] to 1494 [60] participants. These studies included patients with hypertension [52, 55], dialysis patients [55, 57, 58, 61, 62] and patients with diabetes [55–57, 63]. Five studies were in mixed gender populations [54–58, 61, 64], with one study in older (aged 56 to 77 yrs) males [52]. Two of the studies were in Japanese populations [56, 57]. A summary of the characteristics and effect sizes for these studies is shown in Table 3.

Table 3 Summary of primary prevention studies in high risk populations (7 studies)
Full size table

The overall quality of the six prospective cohort studies [53–55, 57, 58, 64] and one RCT [65] was mixed, with the risk of bias assessed as high in four studies [52, 55–57], moderate in two studies [54, 61] and low in one study [58].

CVD outcomes assessed in the studies were individual and composite outcomes including coronary heart disease (CHD) [56], non-fatal myocardial infarction (MI) [52, 55], cardiovascular death [52, 55, 57], atherosclerotic events [53, 54], non-haemorrhagic stroke [54, 65], heart failure (HF) [58], TIA [56] and all deaths [57, 58]. Six studies used Cox proportional hazards models [54–56, 58, 61, 64], with one study reporting results from stepwise multiple logistic and Cox regression analyses [54]. and one reporting a multiple logistic regression model [57]. Age was considered as a variable in all but one study [54], and similarly gender was considered in all but one study [52]. All of the studies included additional variables. These varied in type and number across the studies and studies reported different results for models adjusted with different groups of variables. However, only one study included LDL-C in their model [54]. This study (Cleveland Clinic Hemodialysis Cohort) [54] found that baseline Lp(a) concentration was a significant independent risk factor (p = 0.001) for clinical events attributed to atherosclerotic cardiovascular disease in patients receiving chronic haemodialysis treatment of end-stage renal disease. The only study to report the use of an isoform independent Lp(a) assay (CHOICE [53]) similarly reported that high Lp(a) levels (≥52.5 nmol/L) predicted a 30 % to 40 % increased risk of elevated atherosclerotic cardiovascular disease (ASCVD) in dialysis patients, but that the association of ASCVD with low molecular weight LMW isoforms (increased risk of 60 % to 90 %) was stronger than the association with high Lp(a) concentration.

In half of the studies Lp(a) levels were included as a continuous variable and the other half of the studies used categorical data. Of the eight studies, only one failed to find a significant relationship between Lp(a) and CVD outcomes (Zimmermann 1999 [58]). This study in haemodialysis patients reported that serum Lp(a) concentration was a significant predictor in univariate analyses, but that significance was lost when Lp(a) concentration was included in multivariable models for death and cardiovascular death [58]. This study was carried out in stable haemodialysis patients measuring outcomes at two years and the authors suggested that Lp(a) was only involved in an acute phase reaction. All of the remaining studies showed a positive association, i.e. that increased Lp(a) levels increased the risk of CVD events; hazard ratios (HRs) ranged from 1.16 to 2.97.

Secondary prevention studies

Fourteen studies assessed the relationship between Lp(a) and CVD outcomes in patients with previous CVD events (Scandinavian Simvastatin Survival Study [4S Study]; [66] Atherothrombosis Intervention in Metabolic Syndrome With Low HDL/High Triglycerides: Impact on Global Health Outcome trial [AIM–HIGH]; [67] Ezhov 2014; [68] Global Evaluation of New Events and Restenosis After Stent Implantation [GENERATION] study; [69] Heart and Estrogen/progestin Replacement Study [HERS]; [70] Ikenaga 2011; [71] Park 2015; [72] Konishi 2013; [45] Kwon 2015; [73] Long-Term Intervention with Pravastatin in Ischaemic Disease [LIPID] study; [13] Rapamycin-Eluting Stent Evaluated At Rotterdam Cardiology Hospital study [RESEARCH]; [74] Rosengren 1990; [75] Treating to New Targets [TNT] study; [76] and Wehinger 1999 [77]). Follow-up in the studies ranged from 1 year [77] to 8.5 ± 3.5 years [68] and the sample size ranged from 115 [75] to 4444 [66] participants. These studies included patients with previous MI/CHD disease [63], [66, 70, 75, 76] CAD; [73] stent placement after symptomatic CAD [63, 72, 77], patients undergoing percutaneous coronary intervention (PCI) [45, 71, 78], history of CVD [13, 60, 67], patients after successful coronary artery bypass graft (CABG) [68] and patients hospitalised for stable angina, non-ST elevation acute coronary syndromes (NSTACS) or ST-segment elevation myocardial infarction (STEMI) [63, 69]. The majority of studies were in mixed gender populations with the exception of two studies in men [69, 75] and one in elderly postmenopausal women [70]. Two of the studies were in Japanese populations [45, 71] and two in Korean populations [72, 73]. A summary of the characteristics and effect sizes for these studies is shown in Table 4.

Table 4 Summary of studies in secondary prevention (14 studies)
Full size table

The overall quality of the eight prospective cohort studies [45, 68, 69, 71–74, 77], one nested case-control study [75] and five RCTs [13, 66, 67, 70, 76] was mixed, with the risk of bias assessed as moderate in seven studies [13, 66, 72–75, 77] low in two studies [67, 68] and high in two studies [45, 76]. There was insufficient information to make a proper assessment in the remaining three studies due to poor reporting [69–71].

CVD outcomes assessed in the studies were individual and combined outcomes including the following: CHD [70], non-fatal MI [13, 68–70, 75], cardiovascular death [13, 75, 76], [69, 70, 79] stroke [67], MACE [66, 71–74], hospitalisation for angina [68, 69], MI [70, 72, 76] TIA [67] angiographic restenosis [72, 77] and all deaths [45, 66]. Twelve studies used Cox proportional hazards models [13, 67–74, 76, 77, 80], and two studies used logistic regression [66, 75]. All of the studies considered age and gender in their analyses. The other variables used in the analyses differed in type and number across the studies and on occasions the studies reported results for models adjusted for different groups of variables. Only two studies included LDL-C level in their model (HERS [70] and Park 2015 [70, 72]) and binary restenosis and 3 yr adverse clinical outcomes in an Asian population (Lp(a) > 50 mg/dL (versus Lp(a) ≤ 50 mg/dL [reference]: OR 2.88, 95 % CI: 1.37 to 6.07) [72]. In addition, one further study (Kwon 2015 [73]) included the presence of hyperlipidemia (defined as LDL-C ≥ 130 mg/dL) as a variable in the multivariable model. This study also concluded that elevated Lp(a) is associated with worse outcomes (MACE) in type 2 diabetics patients with symptomatic CAD (Tertile 1 [median 4.7 mg/dL; reference] versus Tertile 2 [median 13.5 mg/dL]: HR 1.54, 95 % CI: 0.68 to 3.50 and Tertile 1 [median 4.7 mg/dL; reference] versus Tertile 3 [median 38.8 mg/dL]: HR 2.89, 95 % CI: 1.37 to 6.08).

Five studies (HERS; [70] Ezhov 2014; [68] LIPID; [13] Kwon 2015; [73] and RESEARCH [74]) reported the use of an isoform independent Lp(a) assay. All of these studies concluded that increased baseline [13, 68, 70, 72] or follow-up [13] Lp(a) concentration was a significant and independent risk factor for CVD events including recurrent CHD [70], MACE [68, 73], total CHD events [13] and prognosis after PCI [74]. HERS [70] reported that increased baseline Lp(a) concentrations (≥ 25.4 mg/dL) of Lp(a) were associated with significant and independent increased in CHD (HRs between 1.01 and 1.31). Ezhov 2014 [68] reported that stable CHD patients with Lp(a) ≥30 mg/dL were at a significantly greater risk of cardiovascular death and MI (HR 2.98, 95 % confidence interval [CI]: 1.76 to 5.03) and cardiovascular death, MI, hospitalization for recurrent or unstable angina and repeat revascularization (HR 3.47, 95 % CI: 2.48 to 4.85), than patients with Lp(a) values <30 mg/dL. The LIPID [13] study as reported that increased baseline Lp(a) concentrations were independently associated with an increased risk of total CHD events (p < 0.001), total cardiovascular disease events (p = 0.002), and coronary events (p = 0.03). The authors also reported that the greatest risk occurred at Lp(a) concentrations >73 mg/dL (upper decile) and that an increase in Lp(a) concentration at 1 year was associated with an increased risk of total CHD events and total cardiovascular disease events (both p = 0.002). The RESEARCH [74] study concluded that there was a significant and independent association between Lp(a) concentrations before PCI and a higher risk of MACE at 1-year follow-up (HR 3.1, 95 % CI: 1.1 to 8.6 for the highest versus [≥ 65.2 nmol/L] the lowest tertile [<9.8 nmol/L]). However, the authors reported that this association weakened and lost significance with long-term follow-up.

Overall, half of the studies included Lp(a) level as a continuous variable and the other half of the studies used categorical data. Where Lp(a) was assessed as a categorical value, the thresholds for the categories differed between studies. Across all of the 14 studies, 91 effect sizes were reported for the association between Lp(a) and CVD events, 36 were statistically significant, showing a positive association, between increased Lp(a) levels and the risk of CVD events (HRs ranged from 0.75 to 3.7). Only one study reported a negative association (Wehringer 1999 [77]), which was not statistically significant. This study concluded that elevated Lp(a) mass did not influence the one-year clinical and angiographic outcome after stent placement [77].

The remaining 53 analyses suggested that Lp(a) predicted CV outcomes (HRs ranged from 1.21 to 3.7), though not all of the effect sizes were statistically significant. This included results from five studies (4S study; [66] GENERATION; [69] HERS; [70] LIPID; [13, 74] and RESEARCH [74]). Although some non-significant results for subgroup and sensitivity analyses were reported in these five studies, all concluded that high Lp(a) concentrations (range ≥ 25 to 65.2 mg/dL) were significant predictors of CVD events, including in populations of postmenopausal women with CHD [70], patients with previous CHD [66], stable CHD [13], stable and unstable coronary syndromes after coronary stenting [69] and short term progress after PCI [74].

Discussion

This review presents contemporary evidence examining the extent of relationship between Lp(a) levels and CVD outcomes. The strengths of the review include the adherence to validated rigorous systematic review methodology. This is also one of the first systematic reviews to examine high risk primary prevention patients and secondary prevention populations separately, while only focusing on evidence from multivariable analyses which took into consideration potential confounders.

Our review suggests that evidence is available to support an independent positive association between Lp(a) and the risk of future CVD events both in the general population and in high risk populations, such as those with diabetes, hypertension, or on dialysis. Evidence also exists to support the positive independent association of Lp(a) mass with CVD events in secondary prevention populations. The number of studies for high risk primary prevention populations and secondary prevention populations was limited.

Our findings confirm previous reviews of primary prevention studies; [16, 18] including a review by the Emerging Risk Factors Collaboration published in 2009. This review of 37 prospective studies (n = 1,40,956) reported that Lp(a) modelled as continuous and categorical variables, was an independent risk factor for coronary heart disease death, nonfatal MI, and stroke [16]. The review used individual patient data from the included studies and focused on the primary prevention of coronary heart disease, stroke and non-vascular mortality. In comparison, our review was based on study level data and included a much broader range of CVD outcomes and encompasses primary prevention in high risk population and secondary prevention populations.

Our review found four out of 57 studies which concluded that Lp(a) mass was not a predictor of subsequent events. This included three primary prevention studies [33, 39, 51] in the general population and one in a high risk population [58]. The study in a high risk population reported that in stable haemodialysis patients measuring outcomes at two years, Lp(a) mass was related to the development of overall death and cardiovascular death, but suggested that it was involved in an acute phase reaction and successful treatment of the inflammatory condition may improve long-term survival and so explain the lack of an association with mortality in these patients [58]. The primary prevention studies in the general population concluded that there was no overall statistically significant association between Lp(a) mass and the risk of all-cause mortality, cumulative fatal–nonfatal stroke, and cumulative fatal–nonfatal CAD events [51], ischemic stroke in postmenopausal women [39] and peripheral arterial disease in men [33]. This may have been due to inadequate measurement methods for Lp(a) in comparison with other primary prevention studies, although methods were not always clearly described with respect to their isoform independence [39].

With respect to the relationship between Lp(a) mass and the secondary prevention of CVD outcomes, our review found some evidence to support a positive association in agreement with previous reviews in this population [81, 82] including a recent review by O’Donoghue [20]. Unlike our review, O’Donoghue and colleagues had access to individual patient data from three studies (PEACE – Prevention of Events with Angiotensin Converting Enzyme Inhibition; [83] CARE - Cholesterol and Recurrent Event); [84] and PROVE IT–TIMI 22 - Pravastatin or Atorvastatin Evaluation and Infection Therapy–Thrombolysis In Myocardial Infarction 22 [85] trials) and were able to pool these data. When combined with in some cases unpublished data from eight previous studies (FATS - Familial Atherosclerosis Treatment Study; [86] 4S; [66] HERS; [70] GENERATION; [69] Saely 2006; [87] Skinner 1997; [88] Stubbs 1998; [89] and AIM-HIGH [67]) the authors also found a significant association between Lp(a) and the risk of future myocardial infarction, MACE (OR 1.40, 95 % CI: 1.15 to 1.71). This suggests that although our review did not necessarily have access to all of these trial data, similar conclusions were evident.

Only one study [77] in our review concluded that elevated Lp(a) levels were not predictive of CVD events after stent placement. In this case, Lp(a) mass was found not to influence the one-year clinical and angiographic outcome after stent placement, but the study did not use a standardised assay for Lp(a) level determination and may have been confounded by the use of antithrombotic drugs post stent placement [77]. Previous reviews in secondary prevention have similarly reported a positive association between Lp(a) and CVD event for populations including those who have experienced a stroke [81], and patients who have experienced in-stent restenosis after coronary stenting [82]. However, unlike our review these reviews have not focused specifically on multivariable studies to control for confounding factors.

The relationship between Lp(a) mass and stroke is of particular interest clinically and our review suggests that there is evidence to suggest a significant positive relationship between Lp(a) and non-haemorrhagic stroke in high risk [54, 65] and secondary prevention [67] populations, in agreement with the findings of other recent reviews [16]. However, the risk relationship in the general population was not as clearly defined, with the suggestion that an effect is only present in certain subgroups of the population. For instance, Lp(a) appeared to independently predict fatal and non-fatal stroke/TIA in middle-aged men [26], but not for ischemic stroke in postmenopausal women [39]. The effects of Lp(a) levels on stroke including ischemic stroke were not as well investigated however, and there is a need for more well designed studies to look at the specific effects of Lp(a) with regard to the different stroke subtypes including ischemic stroke.

The identification of Lp(a) as a potential risk factor for CVD/CHD risk prompts the question as to whether this would be an appropriate biomarker for risk stratification and screening [90]. To date, RCTs have not been performed in patients with elevated Lp(a) levels that were randomized to a therapy, primarily due to lack of therapeutic agents developed specifically to lower Lp(a). In addition, there is also a lack of clinical trial evidence to show that Lp(a) reduction (independent of effects on LDL-C) lowers CVD event risk. Only half of the studies in our review included LDL-C as a variable in the multivariable model. Despite this, some studies have suggested a genetic basis for a link between Lp(a) and CVD. Evidence from multiple genome wide association [91, 92] and Mendelian randomization studies [93], suggested that LPA gene variants (encoding Lp(a) lipoprotein) were strongly associated with both an increased level of Lp(a) lipoprotein and an increased risk of coronary disease. However, a more recent Mendelian randomization study has suggested that Lp(a) promotes CVD through atherosclerotic stenosis whereas possible prothrombotic effects appear less influential [94]. In addition, evidence suggests that Lp(a)-lowering therapy such as a Lp(a) apheresis with immunabsorption against human apo(a) results in a significant improvement in the stenosis of the coronary arteries without evidence of other major changes in the lipid profile [95].

In addition, Lp(a) may enhance risk discrimination and reclassification; a recent study on the predictive role of Lp(a) in long-term (15 years) CVD outcomes in general community showed that the net reclassification improvement afforded by Lp(a) was as high as 39.6 % in intermediate-risk group and indicated that Lp(a) can modify clinical risk assessment [96, 97]. The 2015 National Lipid Association recommendations for patient-centered management of dyslipidemia suggest that the presence of Lp(a) levels of 50 mg/dL or more may warrant moving a patient into a higher risk category [98]. Similarly, the European Atherosclerosis Society consensus panel recommends screening for elevated Lp(a) in those at intermediate or high CVD/CHD risk and a desirable level <50 mg/dL as a function of global cardiovascular risk [18]. This would also suggest that current and future treatments which reduce Lp(a) levels, such as apheresis [95, 99], antisense therapy which targets Apo(a) [100] and PCSK9 inhibitors [10, 101, 102] could provide additional benefit in the treatment of patients at risk of CVD beyond LDL-C lowering [18], and evidence of such additional benefit beyond LDL-C lowering should be investigated further in ongoing and future trials [103].

The analysis within our review was limited by the inability to carry out statistical pooling/meta-analysis. This was not possible due to the considerable variation in outcomes, modelling and Lp(a) assays used. Little can also be concluded about the concurrent effects of LDL-C and Lp(a) as very few of the included studies considered both LDL-C and Lp(a) as model variables. Poor reporting of study methodology in some of the studies also hampered the conduct of the review, particularly during the study selection process and the risk of bias assessment. In addition, in some cases relevant data for studies were not available from individual study publications and were only available in pooled analyses from groups of trialists, which were not eligible for inclusion in our review.

Issues with Lp(a) measurement were also problematic and hampered interpretation as has been noted by previous review authors [11, 104]. The methods used to measure Lp(a) mass were also poorly reported. At present research suggests that there are no commercially available assays that are completely and truly insensitive to the variability in Lp(a) particle mass, and so the development of assays which are mass-insensitive are key to the future interpretation of Lp(a) risk prediction studies [104]. However, for the purposes of this review we have used the authors’ description to classify whether studies were isoform dependent or independent. In the majority of cases the classification was either not reported or unclear. However, the authors of one study (Cho 2010 [46]) clearly reported that the test used was isoform dependent and 17 studies that the test used was isoform independent (CCHS; [97] Chin-Shan Community Cardiovascular Cohort Study; [42] CHOICE; [53] CHS; [105] D’Angelo 2006; [106] EPIC; [107] Ezhov 2014; [68] FHS; [108] GRIPS; [28] HERS; [70] HPFS; [32] LIPID; [13] Lipid Research Clinics Coronary Primary Prevention trial; [35] NHS; [109] RESEARCH; [74] Saely 2006; [87] and WHS [40]). No obvious differences between these two sets of studies were evident in terms of the significance and direction of effects for the relationship. All of the studies of high risk prevention and secondary prevention population that reported the use of an isoform independent test concluded that Lp(a) was an independent risk factor for CVD events. Future studies should ensure that their methodology, including Lp(a) assay methods is clearly reported, given the potential issues relating to the reliability of Lp(a) measurement and comparability between assays.

Conclusions

There is evidence to suggest that increased Lp(a) levels are associated with modest increases in the risk of future CVD in both of the lower and higher risk populations reviewed. Therapies that provide Lp(a) lowering in addition to LDL-C lowering such as PCSK9 inhibitors and antisense therapy which targets Apo(a), should be investigated for additional benefit in these populations beyond the expected benefits of the LDL-C lowering.