Residual Inflammatory Risk On Treatment with PCSK9 Inhibition and Statin Therapy

Abstract

Background—

The combination of statin therapy and proprotein convertase subtilisin-kexin type 9 (PCSK9) inhibition markedly lowers low-density lipoprotein cholesterol (LDL-C) and reduces cardiovascular event rates. Whether residual inflammatory risk as measured by on-treatment high sensitivity C-reactive protein (hsCRP) remains an important clinical issue in such patients is uncertain.

Methods—

We evaluated residual inflammatory risk among 9,738 patients participating in the Studies of PCSK9 Inhibition and the Reduction of Vascular Events (SPIRE)-1 and −2 cardiovascular outcomes trials who were receiving both statin therapy and bococizumab, according to on-treatment levels of hsCRP (hsCRP_OT) and LDL-C_OT measured 14 weeks after drug initiation. The primary endpoint was nonfatal myocardial infarction, nonfatal stroke, hospitalization for unstable angina requiring urgent revascularization, or cardiovascular death.

Results—

At 14 weeks, the mean percent change in LDL-C among statin treated patients who additionally received bococizumab was −60.5% (95% CI −61.2 to −59.8; p<0.001; median change −65.4%) as compared to 6.6% (95% CI −1.0 to 14.1; p=0.09; median change 0.0%) for hsCRP. Incidence rates for future cardiovascular events for patients treated with both statin therapy and bococizumab according to on treatment levels of hsCRP <1, 1-3, and >3 mg/L were 1.96, 2.50, and 3.59 events per 100 person-years, respectively, corresponding to multivariable adjusted hazard ratios of 1.0, 1.16 (95% CI 0.81 to 1.66), 1.62 (95% CI 1.14 to 2.30) (p-trend=0.001) after adjustment for traditional cardiovascular risk factors and LDL-C_OT. Comparable adjusted hazard ratios for LDL-C_OT (<30, 30-50, >50 mg/dL) were 1.0, 0.87, and 1.21, respectively (p-trend=0.16). Relative risk reductions with bococizumab were similar across hsCRP_OT groups (p-interaction=0.87).

Conclusions—

In this post-hoc analysis of the SPIRE trials of bococizumab in a stable outpatient population, evidence of residual inflammatory risk persisted among patients treated with both statin therapy and PCSK9 inhibition.

Clinical Trial Registration—

URL: https://clinicaltrials.gov Unique Identifiers: NCT01975376, NCT01975389

Introduction

Patients with residual inflammatory risk have high rates of recurrent cardiovascular events in association with persistently elevated levels of high sensitivity C-reactive protein (hsCRP) despite aggressive use of statin therapy.^1-7 Such patients, commonly defined as those taking statin therapy who have hsCRP ≥ 2 mg/L and LDL cholesterol (LDL-C) < 70 mg/dl,⁸ comprise nearly 30 percent of patients in contemporary practice and are twice as common as those with residual cholesterol risk (defined by LDL-C levels ≥ 70 mg/dL and hsCRP < 2 mg/L).⁹ Recently, the Canakinumab Anti-inflammatory Thrombosis Outcomes Study (CANTOS) demonstrated that IL-1 inhibition with canakinumab significantly reduces both hsCRP and cardiovascular events in stable patients with prior myocardial infarction (MI) and elevated hsCRP,¹⁰ data providing the first potential treatment for patients with residual inflammatory risk. Indeed, the magnitude of risk reduction with canakinumab in CANTOS, despite no change in LDL-C, was virtually identical to that achieved in the FOURIER and SPIRE proprotein convertase subtilisin-kexin type 9 (PCSK9) inhibitor trials^{11, 12} of evolocumab and bococizumab, respectively, in stable high-risk populations. Table 1 provides a brief description of design elements of these trials as well as the recently completed ODYSSEY Outcomes Trial,¹³ which tested the PCSK9 inhibitor alirocumab. Importantly, the absolute event rates of 5.3% and 9.1% at 1-year and 2-years on treatment with evolocumab in FOURIER inform us that many patients achieving very low LDL-C levels will continue to experience cardiovascular events. Whether residual inflammatory risk, that cardiovascular risk attributable to residual subclinical inflammation, remains an important clinical issue among statin treated patients who additionally receive PCSK9 inhibition is unknown. We addressed this issue in the recently completed SPIRE-1 and SPIRE-2 trials.

Table 1.

Comparison of the CANTOS, SPIRE-1, SPIRE-2, FOURIER, and ODYSSEY Outcomes Clinical Trials

	CANTOS10	SPIRE-112, 14	SPIRE-212, 14	FOURIER11	ODYSSEY Outcomes13
Monoclonal Antibody	Canakinumab (human)	Bococizumab (humanized)	Bococizumab (humanized)	Evolocumab (human)	Alirocumab (human)
Entry LDL-C (mg/dL)	No Entry Threshold	≥ 70	≥ 100	≥ 70	≥ 70
Statin Requirement	No requirement; 91.1% taking statins Median (IQR) LDL-C: 82.0 (63.0–106.7)	Atorvastatin 40 or 80 mg, rosuvastatin 20 or 40 mg, simvastatin 40 mg (or 80 mg if > 1 year) or documented intolerance to high intensity statin (SPIRE-1 and SPIRE 2) or documented complete statin intolerance (SPIRE-2)		High-intensity statin preferred, minimum dose atorvastatin 20 mg or equivalent	Atorvastatin 40 or 80 mg, rosuvastatin 20 or 40 mg or maximum tolerated dose of one of these agents
High-Risk Secondary Prevention	Yes	Yes	Yes	Yes	Yes
High-Risk Primary Prevention	No	Yes	Yes	No	No
Status	Completed	Prematurely Terminated Due to Bococizumab Immunogenicity	Prematurely Terminated Due to Bococizumab Immunogenicity	Completed	Completed

Methods

Data Availability

The data will not be made available to other researchers for purposes of reproducing the results. However, the methods used in the analysis are available upon request.

Study Population and Procedures

The SPIRE bococizumab development program consisted of two parts: the six SPIRE lipid-lowering studies and the SPIRE-1 and SPIRE-2 event-driven cardiovascular trials. The design and primary findings of SPIRE-1 and SPIRE-2 have been previously published.^{12, 14} The virtually identical designs of the two trials permitted them to be combined according to an integrated statistical analysis plan. In brief, patients were eligible for enrollment if they had either a prior cardiovascular event (secondary prevention cohort) or a history of diabetes, chronic kidney disease, or peripheral vascular disease with additional cardiovascular risk conditions or a history of familial hypercholesterolemia (high-risk primary prevention cohort). All patients were required to have received at least 4 weeks of stable statin therapy (atorvastatin ≥ 40 mg/day, rosuvastatin ≥ 20 mg/day, or simvastatin ≥ 40 mg/d) unless they could not take those doses without side effects and were thus on lower intensity statin therapy or had complete statin intolerance (eligible for SPIRE-2 only). Patients were required to have a directly measured LDL-C level of ≥ 70 mg/dL in SPIRE-1 and of ≥ 100 mg/dL in SPIRE-2. Patients were also eligible according to their non-HDL cholesterol level at entry (≥ 100 mg/dL for SPIRE-1 and ≥ 130 mg/dL for SPIRE-2). In a double-blinded fashion, patients were randomized in a 1:1 ratio to treatment with subcutaneous bococizumab 150 mg every 2 weeks or matching placebo. The SPIRE program was sponsored by Pfizer.

The study population for the current analysis comprises the subgroup of SPIRE-1 and −2 patients who were receiving moderate- or high-intensity statin therapy, were allocated to active bococizumab and had available baseline and 14 week hsCRP measures available for analysis (n=9,738). Adherence to randomized treatment was high with 89.9% and 88.8% of patients assigned to bococizumab having ≥ 80% compliance (≥ 80% of doses administered of doses planned) and a mean compliance (number of doses taken of number of doses planned) of 93.1% and 93.0% in SPIRE-1 and SPIRE-2, respectively. All patients provided written informed consent. Ethics committees at each center approved the protocol.

Endpoints

The pre-specified primary endpoint of the two trials was a composite of adjudicated and confirmed nonfatal myocardial infarction (MI), nonfatal stroke, hospitalization for unstable angina requiring urgent revascularization, or cardiovascular death. All incident events that were components of these endpoints were adjudicated by a committee in which the members were unaware of treatment assignments.

Statistical Analyses

Of 13,675 patients randomized to the active treatment arm, 12,711 (93.0%) were receiving moderate- or high-intensity statin therapy, and 9,738 (71.2%) also had hsCRP_OT levels available at the 14 week timepoint. The corresponding proportion of patients randomized to placebo, receiving statin therapy and having follow-up biomarker levels was 9,785 (71.6%). Baseline characteristics of included versus excluded patients are shown in Supplementary Table 1.

The study population was then restricted to individuals allocated to bococizumab and divided into three groups according to hsCRP_OT level <1, 1-3, and >3 mg/dL comprising 30.4%, 34.8%, and 34.9% of patients, respectively. These cut points are consistent with the previously proposed Center for Disease Control/American Heart Association classification scheme¹⁵ and correspond to approximate tertiles of the population distribution. When cut points of <2 and ≥2 mg/dl were used, these percentages were 52.8% and 47.2%. Baseline characteristics according the three primary hsCRP_OT groups were summarized using percentages for categorical values and medians (interquartile ranges) for continuous variables. Trends in these characteristics across ordered hsCRP_OT categories were assessed using the Cochran-Armitage trend test for differences in proportions and the Jonckheere-Terpstra test for differences in medians.

To evaluate the treatment effect of bococizumab on lipid levels and on hsCRP, median on-treatment levels were determined at baseline and 14 weeks of therapy. Linear mixed model repeated measure analyses conditioning on the baseline value were used with the independent value being the biomarker of interest after log transformation as deemed appropriate for non-normal distributions. The mean percent change and bococizumab treatment effect was estimated by fitting terms corresponding to the study drug assignment.

Percent change in lipid levels in each hsCRP_OT group among patients allocated to bococizumab was then estimated using mixed models as before, conditioning on the baseline value and fitting a term corresponding to the hsCRP_OT group.

Cox proportional hazards models were used to estimate hazard ratios (HRs) according to hsCRP_OT group. As the LDL-C lowering effects of bococizumab emerge as early as 4 weeks after drug initiation,¹² all endpoints including those occurring before 14 weeks were used in the primary analysis. Sensitivity analyses were conducted after removing these events (Supplemental Table 2). The three models presented are adjusted for the following: 1) age and sex, 2) age, sex, traditional cardiovascular risk factors (including current smoking, diabetes, hypertension, and body-mass index) plus statin intensity at enrollment (moderate-intensity or high-intensity), and 3) model 2 variables plus on-treatment LDL-C (LDL-C_OT). For each model, a test for trend across hsCRP_OT categories was performed after assigning the median value to each group. All analyses were stratified by study (SPIRE-1 or SPIRE-2), region, and screening LDL-C threshold (< 70 or <100 mg/dL). We also assessed for heterogeneity in treatment effects of bococizumab versus placebo according to hsCRP_OT groups by use of an interaction term (bococizumab x hsCRP_OT group).

To permit comparison to associations for on-treatment LDL-C measured at 14 weeks, the study population was additionally divided into LDL-C_OT groups (approximate tertiles) using the categories of <30, 30-50, and >50 mg/dL and comparable Cox models were used to estimate adjusted HRs in each of these groups. Cut points of < or ≥2 mg/L for hsCRP and < or ≥40 mg/dl for LDL-C were also used. Finally, to examine the risk association throughout the range of hsCRP_OT, we plotted the relationship between hsCRP_OT and cardiovascular event rates using a smoothing function to the average of estimated event rates at each hsCRP_OT level based on adjusted Cox models.

Results

Study Population by On-Treatment hsCRP Levels

The study population comprised 2958 (30.4%) with hsCRP_OT < 1 mg/L, 3385 (34.8%) with hsCRP_OT 1-3 mg/L, and 3395 (34.9%) with hsCRP_OT > 3 mg/L. Baseline characteristics according to hsCRP_OT are shown in Table 2. Patients with higher hsCRP_OT were more likely to be women, to be obese, have diabetes or diagnosed hypertension, and to be current smokers but less likely to have prior cardiovascular disease. Several baseline lipid parameters were also significantly different across increasing hsCRP groups, including higher levels of LDL-C, total cholesterol (TC), non-HDL cholesterol (non-HDL-C), triglycerides, TC:HDL-C ratio, and apolipoprotein B (apoB) and lower levels of HDL-C.

Table 2.

Baseline Characteristics According to hsCRPOT at 14 Weeks

	hsCRPOT Group
Baseline Characteristic	<1 mg/L N=2958 (30.4%)	1-3 mg/L N=3385 (34.8%)	>3 mg/L N=3395 (34.9%)	P-value
Age, years	63 (56, 69)	64 (57, 70)	63 (57, 69)	0.19
Female Sex, %	28.0	29.4	31.4	<0.001
Body-Mass Index, kg/m2	27.9 (25.5, 30.9)	29.4 (26.6, 32.7)	31.4 (27.9, 35.9)	<0.001
Diabetes, %	37.4	47.2	60.2	<0.001
Hypertension, %	75.1	82.4	87.4	<0.001
Current Smoking, %	19.0	23.8	30.0	<0.001
High-Risk Primary Prevention, %	8.1	14.0	19.2	<0.001
US/Canada, %	21.3	27.4	35.6	<0.001
Statin Regimen, %
Moderate Intensity	8.3	9.0	9.4	0.10
High Intensity	91.8	91.0	90.6
LDL Cholesterol, mg/dL	92.4 (80.5, 110.4)	96.5 (82.4, 118.0)	101.0 (85.1, 125.5)	<0.001
Total Cholesterol, mg/dL	161.8 (144.8, 184.2)	166.2 (147.5, 193.1)	171.5 (151.2, 200.5)	<0.001
Non-HDL Cholesterol, mg/dL	112.0 (97.1, 132.2)	117.7 (101.2, 143.8)	124.9 (105.5, 153.9)	<0.001
HDL Cholesterol, mg/dL	47.0 (40.0, 56.0)	45.2 (38.4, 53.5)	43.8 (37.0, 52.5)	<0.001
Triglycerides, mg/dL	116.5 (87.6, 160.2)	137.5 (101.3, 192.5)	149.6 (108.0, 210.6)	<0.001
Total:HDL Cholesterol Ratio	3.4 (2.9, 4.1)	3.7 (3.1, 4.4)	3.9 (3.2, 4.7)	<0.001
Apolipoprotein B, mg/dL	78 (68, 91)	82 (71, 98)	87 (74, 105)	<0.001
High-Sensitivity CRP, mg/L	0.7 (0.4, 1.2)	1.8 (1.1, 2.9)	4.7 (2.7, 7.6)	<0.001

Table entries are medians (IQRs).

Percentages may not add up to 100% due to rounding.

Bococizumab Treatment Effects on Lipid Levels and hsCRP

When compared to placebo, bococizumab was associated with statistically significant reductions in LDL-C (−60.5%), TC (−37.6%), non-HDL-C (−54.9%), TC:HDL-C ratio (−41.1%), apoB (−56.0%), and triglycerides (−19.9%) as well as an increase in HDL-C (+6.4%) (Table 3; all p<0.001). By contrast, there was no significant effect on hsCRP; mean percent change was +6.6% (95% CI: −1.0 to 14.1; p=0.09; median change 0.0%) at 14 weeks and +6.7% (−9.3 to 16.9%; p=0.57; median change 0.0%) at 52 weeks (n=3267). Percent changes in lipid fractions were slightly lower in magnitude in higher hsCRP_OT groups (Figure 1). Nonetheless, even among those with hsCRP > 3 mg/L, the median LDL-C_OT at 14 weeks was 41.7 (IQR 25.9, 67.0) mg/L. Bococizumab treatment effects by hsCRP_OT were similar in magnitude, and there was no evidence of heterogeneity across hsCRP_OT groups (p-interaction=0.87).

Table 3.

Median Lipid Levels and hsCRP at Baseline and 14 Weeks and Treatment Effect (Percent Change) with Bococizumab

	Bococizumab		Placebo		Treatment Effect*
Parameter	No.	Median (IQR)	No.	Median (IQR)	%Change	95% CI	P-Value
LDL-C (mg/dL)
Baseline	9662	96.5 (82.5, 118.0)	9716	96.5 (82.6, 117.5)	−60.5	(−61.2 to −59.8)	<0.001
14 Weeks	9662	34.7 (22.4, 56.4)	9716	97.7 (82.0, 120.3)
Total Cholesterol (mg/dL)
Baseline	9670	166.5 (147.9, 192.3)	9711	166.4 (148.0, 191.9)	−37.6	(−38.1 to −37.1)	<0.001
14 Weeks	9670	102.7 (84.2, 128.2)	9711	167.6 (146.7, 195.0)
Non-HDL Cholesterol (mg/dL)
Baseline	9648	118.0 (101.0, 143.8)	9690	117.8 (101.4, 142.7)	−54.9	(−55.1 to −53.7)	<0.001
14 Weeks	9648	50.0 (34.7, 76.0)	9690	118.9 (100.0, 146.3)
HDL Cholesterol (mg/dL)
Baseline	9649	45.2 (38.2, 54.0)	9694	45.6 (38.6, 54.3)	6.4	(6.1 to 6.8)	<0.001
14 Weeks	9649	48.0 (40.9, 57.9)	9694	45.9 (38.6, 54.8)
Triglycerides (mg/dL)
Baseline	9699	134.5 (98.2, 189.0)	9713	133.6 (98.5, 187.2)	−19.9	(−21.0 to −18.8)	<0.001
14 Weeks	9699	107.0 (76.1, 157.5)	9713	133.0 (96.0, 189.4)
Total: HDL Cholesterol Ratio
Baseline	9648	3.6 (3.1, 4.4)	9690	3.6 (2.1, 4.4)	−41.1	(−41.7 to −40.6)	<0.001
14 Weeks	9648	2.0 (1.7, 2.7)	9690	3.6 (3.0, 4.5)
Apolipoprotein B (mg/dL)
Baseline	9641	82.0 (71.0, 99.0)	9678	82.0 (71.0, 98.0)	−56.0	(−56.7 to −55.2)	<0.001
14 Weeks	9733	37.0 (17.5, 56.0)	9782	82.5 (71.0, 99.0)
hsCRP (mg/L)
Baseline	9738	1.88 (0.87, 4.21)	9756	1.90 (0.85, 4.08)	6.6	(−1.0 to 14.1)	0.09
14 Weeks	9738	1.84 (0.83, 4.19)	9785	1.68 (0.78, 3.88)

^*The percent change is from baseline to 14 weeks for the bococizumab group as compared with the placebo group.

Figure 1. Mean percent change in lipid levels from baseline to 14 weeks according to hsCRP_OT.

Median on-treatment lipid values in each hsCRP_OT group are shown to the right of each plot.

Event Rates According to On-Treatment hsCRP and On-Treatment LDL-C

Overall, a monotonic increase in adjusted event probabilities for the primary CVD endpoint was observed with increasing on-treatment hsCRP levels (Figure 2). Event rates in hsCRP_OT groups were 1.96, 2.50, and 3.59 per 100 person-years for hsCRP <1, 1-2, and >3 mg/L, respectively (Table 4). In multivariable models that adjusted for age and sex (model 1), the corresponding HRs for CVD were 1.0 (ref), 1.23 (95% CI 0.86 to 1.75) and 1.79 (95% CI 1.28 to 2.50); p-trend<0.001. In models additionally adjusting for traditional cardiovascular risk factors and baseline intensity of statin therapy (model 2), the HR comparing highest to lowest hsCRP_OT category (>3 vs. <1 mg/L) was 1.67 (95% CI 1.18 to 2.37; p=0.004). Further adjustment for LDL-C_OT minimally attenuated this risk (model 3, Table 4 and left panel, Figure 3); adjusted HRs were 1.0 (ref), 1.16 (95% CI 0.81 to 1.66) and 1.62 (95% CI 1.14 to 2.30) with p-trend=0.001.

Figure 2. Relationship between hsCRP_OT on a continuous scale and the adjusted event rate for the trial primary endpoint (myocardial infarction, stroke, unstable angina requiring urgent coronary revascularization, and cardiovascular death).

Model adjusts for age, sex, current smoking, diabetes, hypertension, and body-mass index) statin intensity at enrollment (moderate-intensity or high-intensity), and LDL-C_OT. Dots represent individual hsCRP_OT values. The solid red line indicates the estimated event curve from adjusted models.

Table 4.

Hazard Ratios for Cardiovascular Events According to hsCRPOT at 14 weeks

	hsCRPOT Group
	<1 mg/L N=2958 (30.4%)	1-3 mg/L N=3385 (34.8%)	>3 mg/L N=3395 (34.9%)
Primary Endpoint* Events per 100 person-years	52 1.96	76 2.50	109 3.59	P-trend
Model 1	1 (ref)	1.23 (0.86 to 1.75) p=0.3	1.79 (1.28 to 2.50) p=0.001	<0.001
Model 2	1 (ref)	1.17 (0.82 to 1.68) p=0.4	1.67 (1.18 to 2.37) p=0.004	<0.001
Model 3	1 (ref)	1.16 (0.81 to 1.66) p=0.4	1.62 (1.14 to 2.30) p=0.007	0.001
Individual Endpoints (Model 3)
Nonfatal Myocardial Infarction	N=31 1 (ref)	N=36 0.91 (0.56 to 1.49) p=0.7	N=61 1.46 (0.92 to 2.32) p=0.11	0.017
Nonfatal Stroke	N=7 1 (ref)	N=14 1.62 (0.65 to 4.05) p=0.3	N=14 1.47 (0.56 to 3.85) p=0.4	0.4
Hospitalization for Unstable Angina Requiring Urgent Revascularization	N=10 1 (ref)	N=16 1.33 (0.60 to 2.95) p=0.5	N=21 1.65 (0.74 to 3.68) P=0.2	0.2
Cardiovascular Death	N=5 1 (ref)	N=11 1.60 (0.54 to 4.73) p=0.4	N=23 3.76 (1.38 to 10.2) p=0.009	0.002
Any Death	N=10 1 (ref)	N=20 1.58 (0.73 to 3.41) p=0.3	N=38 3.45 (1.68 to 7.08) p=0.001	<0.001

^*The primary endpoint was nonfatal myocardial infarction, nonfatal stroke, hospitalization for unstable angina requiring urgent revascularization, or cardiovascular death.

Model 1: age- and sex-adjusted

Model 2: additionally adjusted for baseline smoking, diabetes, hypertension, body-mass index, baseline statin (moderate- or high-intensity)

Model 3: additionally adjusted for on-treatment LDL-C_OT (no. missing = 76)

All models stratified by study (SPIRE-1 or SPIRE-2), region, and screening LDL-C.

Figure 3. Risk association of hsCRP_OT and LDL-C_OT with incident cardiovascular events according to categories of each biomarker.

Adjusted for age, sex, current smoking, diabetes, hypertension, and body-mass index) statin intensity at enrollment (moderate-intensity or high-intensity), and hsCRP_OT and LDL-C_OT as appropriate. Models for hsCRP_OT are on the left and models for LDL-C_OT is on the right.

Adjustments made for other potential confounding or mediating factors had minimal impact on these results. In models adjusting for on-treatment TC:HDL-C ratio (model 3 plus TC:HDL-C), adjusted HRs were 1.0 (ref), 1.13, and 1.58 (p-trend=0.002). In models adjusting for prior history of CVD (including peripheral vascular disease), adjusted HRs (model 3 plus prior CVD) were 1.0 (ref), 1.18, and 1.62 (p-trend= 0.001). When adjusted for prior history of chronic kidney disease HRs (model 3 plus prior CKD) were 1.0 (ref), 1.16, and 1.60 (p-trend=0.001).

We found similar hazard ratios when analyses were restricted to the placebo group; adjusted HRs (model 3) were 1.0 (ref), 1.11, and 1.72 (p-trend<0.001) according to hsCRP values at 14 week categorized as < 1, 1-3, and >3 mg/L, respectively. When individual components of the composite endpoint were examined (model 3 adjustments), hsCRP_OT category was significantly associated with non-fatal MI (adjusted HRs 1.0, 0.91, 1.46, p-trend=0.017), cardiovascular mortality (adjusted HRs 1.0, 1.60, 3.76, p-trend=0.002), and total mortality (adjusted HRs 1.0, 1.58, 3.45, p-trend<0.001) (Table 4). Similar but non-significant trends were noted for stroke and unstable angina requiring urgent coronary revascularization.

In parallel analyses in which patients were categorized according to LDL-C_OT (<30, 30-50, > 50 mg/dl), the HRs for the primary CVD endpoint were 1.0 (ref), 0.87 (95% CI 0.62 to 1.22) and 1.21 (0.87 to 1.68) with p-trend=0.16 in analyses adjusting for model 3 covariates and hsCRP_OT instead of LDL-C_OT (right panel Figure 2 and Supplemental Table 3). Similar findings were observed when the alternate cut points of ≥ 2 mg/L for hsCRP_OT and ≥ 40 mg/dl for LDL-C_OT were used (Supplemental Tables 4 and 5).

Discussion

In this population of 9,738 high-risk patients concomitantly treated with statins and PSCK9 inhibition, 47.2% had residual inflammatory risk defined by on-treatment hsCRP level ≥ 2 mg/L, with 34.9% having values > 3 mg/L. Individuals with persistent hsCRP elevation tended to be those with multiple risk factors including diabetes, obesity, hypertension, and mixed dyslipidemia, conditions known to correlate with, if not be driven by, a pro-inflammatory state. PCSK9 inhibition with bococizumab had no effect on hsCRP over time. Despite aggressive reduction of LDL-C, there was a continuous gradient in risk for future cardiovascular events according to on-treatment hsCRP. Compared to those without evidence of subclinical inflammation, those with on-treatment hsCRP > 3 mg/L (median on-treatment LDL-C 41.7 mg/dl) had a 62% increase in risk of future cardiovascular events. Elevated hsCRP was significantly associated with increased rates of non-fatal MI, cardiovascular death, and all-cause mortality.

There is broad consensus that atherosclerosis is both a disorder of lipid accumulation and inflammation. From a clinical perspective, extensive prior work has found hsCRP to be an independent predictor of cardiovascular events both in primary prevention and high-risk secondary prevention settings. Further, among patients with residual inflammatory risk, randomized clinical trials have proven the efficacy of statin therapy in primary prevention¹⁶ and anti-inflammatory therapy in secondary prevention.¹⁰ It has been uncertain, however, whether residual inflammatory risk persists after the extremely aggressive reduction in LDL-C that can be achieved with the combination of statin therapy and PCSK9 inhibition. Importantly, in an era when ever more specialized therapies in cardiovascular medicine continue to emerge, the call for biomarkers which inform clinicians about risk stratification, drug choice and dose, therapeutic responses, and ultimately personalized interventions will only be amplified.

In this context, these data have several important implications. First, these data clarify that PCSK9 inhibition has no effect on plasma measures of hsCRP despite large effects on atherogenic lipids. Second, the current data demonstrate that, despite inter-relationships of LDL oxidation and inflammation, the combination of high intensity statin therapy and PCSK9 inhibition does not fully address inflammatory mechanisms of atherothrombosis that may be detected by elevated levels of hsCRP. In isolation, our post-hoc findings are associative and could still be explained by underlying conditions that promote subclinical inflammation. As such, as we have argued elsewhere¹⁷, combination therapy with PCSK9 inhibition and anti-inflammatory therapy may provide the optimal method to address residual cardiovascular risk, a hypothesis that requires a prospective 2X2 factorial trial for adequate testing. While canakinumab is currently the only anti-inflammatory agent proven to reduce cardiovascular events, clinical trials are currently in progress using colchicine and low-dose methotrexate.^{18, 19} Novel agents that inhibit the upstream NLRP3 inflammasome and downstream activation of IL-6 are also under consideration. Importantly, our data do not pertain to the setting of acute coronary syndromes, where anti-inflammatory therapies have thus far failed to impart cardiovascular benefit.

The SPIRE cardiovascular outcomes trials were stopped early due to high rates of development of neutralizing anti-drug antibodies.²⁰ While bococizumab immunogenicity is associated with a less durable LDL reduction, treatment with bococizumab in the longer duration SPIRE-2 outcomes trial was nonetheless associated with a 21% (95% CI 3 to 35%; p=0.02) relative risk reduction in major cardiovascular events overall and a 14% (95%CI −2 to −25%) relative risk reduction per 1 mmol/l LDL-C. These data are fully in line with benefits observed in FOURIER^{12, 21} and preliminary data from the ODESSY Outcomes trial.²² Thus, we believe our findings are unlikely to be explained by diminished bococizumab LDL-C lowering efficacy and likely to apply more broadly to biologic agents in this therapeutic class. As in any post-hoc analysis, our findings may be susceptible to residual confounding. In particular, patients with persistent inflammatory risk were more likely to have cardiovascular risk factors and higher median on-treatment LDL-C. However, our multivariable analyses adjusted for achieved LDL-C levels and showed minimal, if any, attenuation in risk. Furthermore, as shown in CANTOS which enrolled on the basis of elevated hsCRP, this risk group is likely to benefit from anti-inflammatory therapy.¹⁰ Consistent associations were noted for the individual trial endpoints of nonfatal MI, cardiovascular mortality and all-cause mortality. However, it should be noted that the number of events was small. Thus, these findings must be interpreted with caution.

In sum, these contemporary randomized trial data demonstrate that elevated levels of on-treatment hsCRP remain a significant predictor of future cardiovascular risk among stable atherosclerosis patients concomitantly treated with statins and PCSK9 inhibition. This evidence of residual inflammatory risk despite maximal LDL-C lowering, if replicated in other cohorts, suggests that inflammation modulation may offer additional opportunities for cardiovascular risk reduction.

Clinical Perspective.

What is new?

• Among high-risk stable outpatients treated with moderate- or high-intensity statins and PSCK9 inhibition, roughly 1 in 2 had residual inflammatory risk defined by on-treatment hsCRP level ≥ 2 mg/L, with roughly 1 in 3 having values > 3 mg/L.

• Treatment with bococizumab was associated with a 60% mean reduction in low-density lipoprotein cholesterol but little change in high sensitivity C-reactive protein.

• Levels of high sensitivity C-reactive protein > 3 mg/L were associated with a 60% greater risk of future cardiovascular events corresponding to a 3.6% annual event rate (3.6 per 100 person-years) even after accounting for on-treatment low-density lipoprotein cholesterol.

What are the clinical implications?

• PCKS9 inhibition added to statin therapy in stable outpatients does not lower high-sensitivity C-reactive protein.

• Persistent elevations of high-sensitivity C-reactive protein is associated with future cardiovascular risk in these patients even after low levels of low-density lipoprotein cholesterol are achieved.

• If corroborated, these data suggest that inflammation modulation may yet have a role in primary and secondary prevention of cardiovascular disease when low-density lipoprotein cholesterol is controlled.