Safety and efficacy of anti-PCSK9 antibodies: a meta-analysis of 25 randomized, controlled trials

Acbstract Background Inhibition of proprotein convertase subtilisin/kexin type 9 (PCSK9) has been intensively studied to lower low-density lipoprotein cholesterol (LDL-C) levels. The purpose of this meta-analysis was to evaluate the safety and efficacy of anti-PCSK9 antibodies in randomized, controlled trials (RCTs). Methods PubMed, EMBASE, CENTRAL databases, and recent conferences were searched. Safety outcomes were rates of common adverse events. Efficacy outcomes included percentages of LDL-C lowering and other lipid changes compared with placebo and ezetimibe, respectively. Results Twenty-five RCTs encompassing 12,200 patients were included. The rates of common adverse events were firstly reported in our study by pooling together all evidence in RCTs, showing largely no significant difference between anti-PCSK9 antibodies and placebo (or ezetimibe), except that alirocumab was associated with reduced rates of death (relative risk (RR): 0.43, 95 % confidence interval (CI): 0.19 to 0.96, P = 0.04) and an increased rate of injection-site reactions (RR: 1.48, 95 % CI: 1.05 to 2.09, P = 0.02); evolocumab reduced the rate of abnormal liver function (RR: 0.43, 95 % CI: 0.20 to 0.93, P = 0.03), both compared with placebo. No significant difference in safety outcomes was detected between monthly 420 mg and biweekly 140 mg evolocumab treatments. Monthly 420 mg evolocumab treatment significantly reduced LDL-C by −54.6 % (95 % CI: −58.7 to −50.5 %) and by absolute −78.9 mg/dl (95 % CI: −88.9 to −68.9 mg/dl) versus placebo, and by −36.3 % (95 % CI: −38.8 to −33.9 %) versus ezetimibe, and increased high-density lipoprotein cholesterol (HDL-C) by 7.6 % (95 % CI: 5.7 to 9.5 %) versus placebo and 6.4 % (95 % CI: 4.3 to 8.4 %) versus ezetimibe. An equal or even greater change was observed following biweekly 140 mg administration. Significant and favorable changes were also detected in other lipids following evolocumab treatment. Biweekly 50 to 150 mg alirocumab lowered LDL-C by −52.6 % (95 % CI: −58.2 to −47.0 %) versus placebo, by −29.9 % (95 % CI: −32.9 to −26.9 %) versus ezetimibe, and increased HDL-C by 8.0 % (95 % CI: 4.2 to 11.7 %) versus placebo. Conclusions Evolocumab and alirocumab were safe and well-tolerated from our most-powered analyses. Both antibodies substantially reduced the LDL-C level by over 50 %, increased the HDL-C level, and resulted in favorable changes in other lipids. Electronic supplementary material The online version of this article (doi:10.1186/s12916-015-0358-8) contains supplementary material, which is available to authorized users.


Background
Hypercholesterolemia is a major risk factor for cardiovascular disease (CVD) [1]. The introduction of statins has substantially reduced CVD events around the world and is recommended as first-line therapy for CVD management [2]. However, a necessity for other lipid-lowering (especially low-density lipoprotein cholesterol (LDL-C) lowering) agents still exists because some patients cannot tolerate statins due to adverse events, or cannot achieve intensive LDL-C lowering because of extremely high baseline LDL-C levels, or patients with very high risk of CVD events need more intensive lowering therapy [3].
The role of proprotein convertase subtilisin/kexin type 9 (PCSK9) in cholesterol regulation has been established since PCSK9 mutations were first discovered in autosomal dominant hypercholesterolemia (ADH) in 2003 [4]. PCSK9 binds to LDL receptors (LDLR) and facilitates the degradation of LDLRs [5] and thus leads to LDL-C increase, indicating great therapeutic potential. Therefore, inhibiting PCSK9 by monoclonal antibodies [6,7], small interfering RNA [8], and small molecule inhibitors [9] has been evaluated to lower LDL-C levels in human studies during the last few years. However, a comprehensive analysis of the safety of anti-PCSK9 antibodies is absent, and efficacy outcomes on lipid profiles are not uniformly consistent. Therefore, we performed a comprehensive review of the current available evidence to address the safety (to provide the exact rates of common adverse events) and the efficacy (to determine the exact extent of lipid changing effect) of anti-PCSK9 antibodies.

Literature search
We sought to identify all randomized, controlled trials (RCTs) evaluating the safety and efficacy of PCSK9 monoclonal antibodies. We searched PubMed, EMBASE, and the Cochrane Central Register of Controlled Trials (CEN-TRAL) from their inception to 6 October 2014, using the following search terms and key words: ' AMG 145', 'evolocumab', 'SAR236553', 'REGN727x', 'SAR236553/REGN727', 'alirocumab' and 'PCSK9'. Reference lists of the identified reports and relevant reviews were manually checked. Major conference proceedings were searched to retrieve unpublished studies until the end of the American Heart Association (AHA) scientific sessions on 20 November 2014. We did not apply any restriction on languages.

Study selection
Eligibility assessment was performed by two investigators (XZ and QZ). Studies were included if they: 1) were RCTs; 2) involved human subjects; 3) evaluated the safety and efficacy of an anti-PCSK9 antibody (evolocumab or alirocumab); and 4) reported mean differences with corresponding confidence intervals (CIs) or provided data necessary to calculate such. We did not restrict the type of study populations. We excluded animal studies, studies which were not randomized, and studies using other anti-PCSK9 antibodies, such as bococizumab, or PCSK9 inhibitors such as small interfering RNA because of the limited number of trials published regarding these PCSK9 inhibitors.

Outcomes
The safety outcomes were rates of common adverse events, and the primary efficacy endpoints were percent and absolute reductions in LDL-C following anti-PCSK9 antibody treatment. Secondary outcomes included: 1) LDL-C reduction at 52 weeks follow-up for evolocumab; 2) other lipid profile changes stratified by treatment dosages and durations of follow-up.

Data collection
Data were abstracted independently by two reviewers (XZ and QZ) using a standardized data extraction form. When there were disagreements, a third reviewer (LZ) checked the data. The following information was extracted: trial name/first author, year of publication, number of patients, duration of follow-up, age, gender, race, diabetes mellitus, coronary heart disease (CHD), PCSK9 level and all lipid profiles at baseline. Patient profile and background lipid-lowering therapy, treatments and doses in each study were also recorded. For safety endpoints, we extracted the number of events of interest and total number of patients in each group. For efficacy outcomes, as a priority, we extracted the mean differences and their corresponding 95 % CIs or standard errors (SEs) of anti-PCSK9 antibody versus placebo or ezetimibe for each lipid items. Alternatively, mean changes and 95 % CIs (or SEs) from baseline after either anti-PCSK9 antibody or placebo (or ezetimibe) treatments were extracted, thereafter mean differences of anti-PCSK9 antibody versus controls were calculated.

Quality assessment
We followed the Cochrane Collaboration's tool to assess the risk of bias of included trials. Random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias) and other sources of bias were included in the assessment independently performed by two reviewers (QZ and LZ).

Statistical analysis
For all efficacy outcomes, the mean differences following anti-PCSK9 treatment versus placebo or ezetimibe were pooled across studies using the DerSimonian-Laird random-effects models. Comparisons of anti-PCSK9 antibodies with placebo or ezetimibe were performed separately and stratified by dosages of antibodies. Adverse event rates were also pooled with random-effects models. Trials in which the endpoint was not detected in any of the treatment groups were excluded in the analysis of that endpoint. For studies in which only one of the groups had no event of interest, the estimate of treatment effect and its confidence interval were calculated after adding 0.5 to each cell of the 2 × 2 table for the trial [10,11]. We used the I 2 statistic to assess the consistency across studies, with 25 %, 50 %, and 75 % indicating low, moderate, and high degrees of heterogeneity respectively. Meanwhile, the χ2-based Q test was applied, and a P >0.10 suggests significant heterogeneity. Begg's test and Egger's test were performed to assess publication bias. Sensitivity analyses were carried out by omitting one study at one time to evaluate the consistency of the results.
In the LAPLACE-2 trial [7], efficacy data comparing evolocumab and ezetimibe were only reported in five subgroups stratified by background lipid-lowering therapies. We combined the results from these subgroups into a single group using the formulae recommended by the Cochrane Collaboration [12]. All analyses were conducted with the STATA version 11.0 software (STATA Corporation, College Station, TX, USA). The metaanalysis was in line with recommendations from the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (Additional file 1).
Baseline characteristics of individual trials are shown in Table 1 and Table S1 and S2 (in Additional file 2). Several studies did not report age, lipids or PCSK9 level in the overall population. Therefore, we presented these characteristics in control populations (placebo or ezetimibe) in these studies given the significantly similar baseline values between the anti-PCSK9 treatment group and controls. All randomized trials included were published or presented in major conferences between 2012 and 2014. The mean age ranged from 31 to 62 years, and the percentage of women from 37 to 74; over 80 % of the patients were white. Regarding evolocumab, all trials were followed up for 12 weeks, except the OSLER and DESCARTES trials [25,30], which were followed up for 52 weeks. With regard to alirocumab, most trials were followed up for 24 weeks except three phase 2 trials which were followed-up for 8 to 12 weeks [36,37,39]. All included RCTs had a low risk of bias, as detailed in Table S3 (in Additional file 2).

Safety outcomes of evolocumab
The pooled estimate for overall incidence of any treatment emergent adverse events (TEAEs) was 52.2 % (95 %     Table 3). The event rates at 52-week follow-up following evolocumab are also reported in Table 2.  Table S4). There was significant heterogeneity in both comparisons (I 2 = 80.4 % and 93.9 %, respectively). Biweekly administration of 140 mg evolocumab led to even greater reduction than 420 mg monthly treatment, both of which reduced the LDL-C level by over 50 %. The effect is likely dose-dependent with the same frequency of administration. Likewise, in absolute level changes, 420 mg monthly and 140 mg biweekly dosing lowered LDL-C by −78.9 mg/dl (95 % CI: −88.9 to −68.9 mg/dl) and −81.6 mg/dl (95 % CI: −92.0 to −71.1 mg/dl), respectively (Additional file 2: Figure S1 and Table S4). Compared with ezetimibe, significant lowering of LDL-C also occurred in all evolocumab dosages at week 12. Monthly 420 mg and biweekly 140 mg evolocumab administration reduced LDL-C level by −36.3 % (95 % CI: −38.8 to −33.9 %) and −38.2 % (95 % CI: −41.5 to −34.5 %), respectively, versus ezetimibe ( Fig. 3 and Additional file 2: Table S4). No significant heterogeneity was detected in the comparisons (I 2 = 0 and 28.4 %, respectively). Fewer studies reported absolute changes of LDL-C level versus ezetimibe; meta-analyses of these studies demonstrated largely similar but less remarkable results compared with those versus placebo.

Safety outcomes of alirocumab
LDL-C percent and absolute changes at the mean of weeks 10 and 12 following evolocumab treatments versus placebo or ezetimibe were all significant and similar to those changes at week 12 (Additional file 2: Table S5).

Other efficacy outcomes of evolocumab
All dosages except monthly 280 mg evolocumab treatment significantly increased high-density lipoprotein cholesterol (HDL-C) levels at week 12 compared with placebo. The HDL-C level was increased by 7.6 % (95 % CI: 5.7 to 9.5 %) and 6.9 % (95 % CI: 5.4 to 8.4 %) by monthly 420 mg and biweekly 140 mg evolocumab treatment, respectively ( Fig. 4 and Additional file 2: Table S6). No significant heterogeneity was detected in the comparisons (I 2 = 23.3 % and 0, respectively). These two dosages of evolocumab also increased the HDL-C level compared with ezetimibe by 6.  (Table 5 and Additional file 2: Figures  S2 to S5, Tables S7 to S10). Similar results were detected following biweekly 140 mg evolocumab treatment.
The free PCSK9 level was decreased by any dosage of evolocumab treatment. At week 12, monthly 420 mg and biweekly 140 mg treatments lowered the PCSK9 level by −44.0 % (95 % CI: −53.9 to −34.2 %) and −60.9 % (95 % CI: −83.9 to −37.9 %) versus placebo, respectively (Table 5   Table 3 Adverse event rates at 12-week follow-up following different dosages of evolocumab treatments  and Additional file 2: Table S16). Significant heterogeneity was detected. Similar results were obtained at the mean of weeks 10 and 12, and largely similar but less remarkable results were achieved when compared with ezetimibe (Additional file 2). Two RCTs reported efficacy outcomes of monthly 420 mg treatment at 52 weeks follow-up. Likewise, all the comparisons were significant (Additional file 2: Figure S11).
HDL-C level was increased by 8.0 % (95 % CI: 4.2 to 11.7 %) following biweekly 50 to 150 mg treatment and by 7.4 % (95 % CI: 3.8 to 11.1 %) after monthly 150 to 300 mg administration. No significant heterogeneity was found (I 2 = 0 for both comparisons) (Fig. 5B). Metaanalyses of other efficiency outcomes demonstrated reduction in TC, non-HDL-C, ApoB, and Lp(a), and an increase in ApoA1 following alirocumab treatment, which are shown in Table S17 (in Additional file 2).
No significant publication bias was found in most analyses, detailed in each table in Additional file 2. Sensitivity analyses did not generated inconsistent results.

Discussion
For the first time we provide in our study the rates of common adverse events following PCSK-9 antibody treatments by enrolling the largest sample size of patients and thus offering the most robust power, and detected largely no significant difference in major adverse events rates between antibody administration and control treatment, and no difference between different dosages of evolocumab. Notably, evolocumab reduced the rate of abnormal liver function, and alirocumab was associated with reduced rates of death and abnormal kidney function. Meanwhile, we determined the extent of LDL-C lowering of anti-PCSK9 antibodies: LDL-C level was reduced by over 50 % even though patients were on stable statin treatment. The extents of other favorable lipids changes were also documented in our meta-analyses.
It is worth noting that the favorable effects of anti-PCSK9 antibodies were largely achieved in populations who were already on stable statin treatments, indicating an additive, or even synergistic effect to statin in lowering LDL-C levels. This is not a surprise because statin therapy has been well documented to increase PCSK9 levels [40]; thus, inhibiting PCSK9 might enhance the LDL-C lowering effects of statins. Indeed, missense mutations in PCSK9 increased response to statin therapy in unrelated hypocholesterolemic subjects and familial hypercholesterolemia patients [41]. Likewise, in our meta-analysis, the combination of anti-PCSK9 antibody and statin resulted in a very high-intensity LDL-lowering effect, which is recommended by the 2013 American College of Cardiology (ACC)/AHA guideline suggesting no LDL-C goals. Meta-analyses of RCTs on statins also demonstrated that further reductions in LDL-C produce definite further reductions in CVD events [42], even in people at low risk of CVDs [43].
Two fundamental elements could lead to safety concerns: firstly those due to monoclonal antibody administration and secondly due to achieving very low LDL-C levels. We were unable to obtain the mean LDL-C level at the end of antibody administration due to lack of patient-level data in our study. However, estimated lowest LDL-C levels from observations of each study were less than 50 mg/dl, which was more remarkable than those achieved in the recently presented IMPROVE-IT trial (Improved reduction of outcomes: vytorin efficacy international trial) [44]. A combination of simvastatin and ezetimibe led to a mean LDL-C level as low as 53.2 mg/dl at one-year follow-up in high-risk patients with acute coronary syndrome, but showed good safety outcomes, indicating that an even lower level of LDL-C might not result in safety concerns. This notion was further confirmed by two recently published trials with regard to both evolocumab and alirocumab, with a longer follow-up of approximately 12 months [23,45]. The OSLER trial, an extension trial of several phase 2 and phase 3 parent trials (most of which were included in our meta-analysis), showed similar rates of adverse events in patients with LDL-C levels less than 40 mg/dl or less than 25 mg/dl as in those with higher LDL-C levels following evolocumab treatment [45]. The full-term follow-up of the ODYSSEY LONG TERM trial also revealed similar frequency of adverse events among patients who had a LDL-C level less than 25 mg/dl and those who did not [23]. More straightforwardly in our meta-analysis, administration of both anti-PCSK9 antibodies showed promising safety profiles, except that administration of Whether anti-PCSK9 antibody treatments could translate into improved cardiovascular outcomes remains to be confirmed. The ongoing FOURIER (NCT01764633) and ODYSSEY OUTCOMES (NCT01663402) trials will answer this question by assessing the effect of evolocumab and alirocumab on major CVD events with about five years follow-up. However, the probable clinical benefits could be preliminarily inferred based on current evidence: 1) anti-PCSK9 antibodies substantially reduced LDL-C, non-HDL-C and ApoB levels, all of which are positively associated with CVD events [46], and 'a lower LDL-C, a better outcome' has been indicated not only in the era of statins but also following combined use of statins and ezetimibe [44]; 2) anti-PCSK9 antibodies significantly increased HDL-C and ApoA1 levels, which are strongly associated with reduced CVD risk, even in patients achieving very low LDL-C [47]; 3) in the ARIC study, loss-of-function PCSK9 mutations resulted in 28 % (15 %) reduction in LDL-C, and 88 % (47 %) reduction in CHD risk in African-Americans (white people) [48]; combined analyses in other cohort studies also generated 30 % reduction in ischemic heart disease risk [49]; and 4) more direct evidence from the longer-term follow-up results of the OSLER and ODYSSEY LONG TERM trials, although with a limitation of exploratory analysis, both of these trials suggested that patients receiving anti-PCSK9 antibodies had a significantly lower risk of major adverse cardiovascular events, which is consistent with our study showing alirocumab reduced the rates of death. Notably, both trials demonstrated that the cumulative incidence curves diverged progressively over time; therefore, a more remarkable benefit might be expected given a longer-term follow-up. Provided the exploratory nature of these trials, the limited follow-up length and the small number of cardiovascular events, results from ongoing FOURIER (over 27,500 high-risk patients with cardiovascular disease) and ODYSSEY OUTCOMES trials (over 18,000 patients who have experienced an acute coronary syndrome event 4 to 52 weeks prior to randomization) are urgently needed to provide definite answers.

Study limitations
First, the meta-analysis was based on study-level instead of patient-level data. Second, a high level of heterogeneity exists in several analyses. Heterogeneities in patient profile (unrelated or familial hypercholesterolemia) and background lipid-lowering therapy (maximum-tolerated statin, statin-intolerance, or no background anti-lipid therapy) are likely to account for part of this heterogeneity. We performed subgroup analyses based on the type of study population and heterogeneity still existed (data not show). Therefore, we pooled these data with randomeffects models. Third, additional ongoing trials evaluating the efficacy and safety of alirocumab are to be published in a few years. However, with respect to primary efficacy endpoint, dramatic upregulating-LDL effects needed to be reported to balance the lowering-LDL effects demonstrated in our study given the number of patients known to have participated in these ongoing trials (ODYSSEY CHOICE I, ODYSSEY OLE, and so on), which is unlikely.
Fourth, with respect to analysis on safety profiles, widerange 95 % CIs were observed in several endpoints, which made precise estimation of the incidences of these endpoints impossible. Meanwhile, several composite endpoints were included in our study, such as adjudicated cardiovascular events, which might lower the ability of detecting each individual endpoint. Fifth, most trials included in our study had a relatively short-term follow-up (12 and 52 weeks for evolocumab, and mostly 24 weeks for alirocumab), thus rare events could not be fully revealed. Sixth, we could not rule out bias of selective reporting on several safety outcomes; to minimize this bias, we reviewed all the materials (including supplementary materials and relevant publications in other papers) provided by these studies and extracted and analyzed all these data. Notably, no obvious selective reporting bias was detected in major safety endpoints, such as any TEAEs, serious TEAEs, abnormal liver function, abnormal kidney function, injection-site reactions, musculoskeletal

Conclusions
Evolocumab and alirocumab were safe and well-tolerated, largely showing no significant differences in rates of common adverse events with placebo or ezetimibe controls. No difference was detected following different dosages of evolocumab treatments regarding safety profiles. Both anti-PCSK9 antibodies substantially reduced LDL-C by over 50 %, increased HDL-C levels, and resulted in favorable changes in other lipids. We await the results of ongoing trials evaluating their effects on CVD events.
Additional file 2: Table S1. Additional baseline characteristics of included randomized trials. Table S2. Safety endpoints and monitoring methods (mainly visiting periods) of included randomized trials. Table S3.