This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
SPIRE-1 randomized high-risk participants to bococizumab, given to lower low-density lipoprotein cholesterol (LDL-C), or to placebo, and followed them for first occurrence of adjudicated major cardiovascular events. The trial was terminated when the bococizumab clinical development program was discontinued, so its outcome analyses reflect a maximum follow-up of up to 3.4 years rather than a planned event-driven completion.
| Feature | SPIRE-1 |
|---|---|
| Official title | The Evaluation of Bococizumab (PF-04950615; RN316) in Reducing the Occurrence of Major Cardiovascular Events in High Risk Subjects |
| Phase | Phase 3 |
| Condition | Cardiovascular disease |
| Primary purpose | Prevention |
| Design | Randomized, parallel-group, quadruple-masked, placebo-controlled |
| Arms | 2 (bococizumab; placebo) |
| Enrollment | 16,784 |
| Primary endpoint | Event rate per 100 participant-years for first occurrence of adjudicated major cardiovascular event |
| Status | Terminated |
| Dates | Start 2013-10-29; primary completion 2017-03-22 |
| ClinicalTrials.gov | NCT01975376 |
| Funding | Pfizer (industry) |
2. Clinical Question
SPIRE-1 asked whether lowering LDL-C with bococizumab would reduce the rate of first major cardiovascular events compared with placebo in people at high cardiovascular risk. Statistically, this is a superiority question about a time-to-first-event composite endpoint, supported by a large set of component outcomes and lipid biomarker endpoints.
Population
High-risk participants with, or at risk of, cardiovascular disease, enrolled for cardiovascular event prevention.
Intervention
Bococizumab (PF-04950615), a drug intervention administered in a masked fashion.
Comparator
Matching placebo, with the same masking of participants, care providers, investigators and outcomes assessors.
Primary question
Does bococizumab reduce the hazard of a first adjudicated major cardiovascular event (CV death, non-fatal MI, non-fatal stroke, or hospitalization for unstable angina needing urgent revascularization) relative to placebo?
3. Trial Design
Bococizumab (PF-04950615)
- Drug intervention (PF-04950615; RN316)
- Masked to participant, care provider, investigator and outcomes assessor
- Followed for adjudicated CV events and lipid response
Placebo
- Matching placebo drug intervention
- Identical masking
- Followed with the same schedule and adjudication process
4. Randomization, Stratification, and Analysis Populations
The posted analyses describe two stratification variables that were carried into every model: geographic region and LDL-C at pre-screening (<100 mg/dL versus ≥100 mg/dL). In the time-to-event analyses these define the strata of the Cox model and log-rank test; in the lipid analyses they enter the MMRM and ANCOVA models as fixed effects. Analysing according to the factors used to balance randomization keeps the analysis aligned with the design and typically improves precision.
| Analysis population | Definition / role |
|---|---|
| Full analysis set (FAS) | All participants who were randomized, excluding those who attempted to be randomized more than once into a bococizumab CV outcomes trial (B1481022/B1481038) or into more than one CV outcomes trial, and all participants enrolled at study Site 3027, where a quality-related event was identified. Used for all cardiovascular outcome analyses. |
| FAS, evaluable participants | For lipid and biomarker endpoints, FAS participants who were evaluable for the specific outcome measure. |
| Participants at risk for adverse events | 8,386 in the bococizumab arm and 8,374 in the placebo arm for serious adverse event reporting. |
The FAS is a modified intention-to-treat population. Its exclusions are defined by events that are unrelated to the randomized treatment (duplicate randomization attempts, a site with a quality problem), which is the kind of exclusion that generally preserves the comparability created by randomization. It is still worth recognizing that it is not literally "all randomized participants".
5. Endpoints
Primary endpoint
| Endpoint | Definition (registry) | Time frame |
|---|---|---|
| Event rate per 100 participant-years for first occurrence of major cardiovascular (CV) event | First occurrence of a major CV event adjudicated by the Adjudication Committee: CV death (sudden cardiac death, fatal MI, death due to heart failure, death due to stroke [fatal ischemic stroke or fatal stroke of undetermined etiology], or death due to other cardiovascular causes), non-fatal MI, non-fatal stroke, or hospitalization for unstable angina needing urgent revascularization. Event rate calculated as the number of events per 100 participant-years at risk. | From baseline until the date of first adjudicated and confirmed major CV event (maximum duration: up to 3.4 years) |
Secondary endpoints with posted analyses
| Group | Endpoints | Time frame |
|---|---|---|
| Composite CV endpoints | CV death, non-fatal MI or non-fatal stroke; all-cause death, non-fatal MI, non-fatal stroke or hospitalization for unstable angina needing urgent revascularization; all-cause death, non-fatal MI or non-fatal stroke; CV death, non-fatal MI, non-fatal stroke and hospitalization for unstable angina | Up to 3.4 years |
| Death | CV death; all-cause death | Up to 3.4 years |
| Myocardial infarction | Any MI (fatal or non-fatal); fatal MI; non-fatal MI | Up to 3.4 years |
| Stroke | Any stroke (fatal or non-fatal); any stroke of any etiology; fatal stroke; non-fatal stroke | Up to 3.4 years |
| Hospitalization and revascularization | Hospitalization for unstable angina needing urgent revascularization; hospitalization for unstable angina; hospitalization for congestive heart failure; coronary revascularization; CABG; PCI; any arterial revascularization | Up to 3.4 years |
| LDL-C | Percent and nominal change from baseline at Week 14; percent change at last post-baseline measurement | Baseline, Week 14; baseline to last measurement (any time up to Week 140) |
| Other lipids and biomarkers | Percent change in non-HDL-C, total cholesterol, VLDL-C, RLP-C, Apo B, HDL-C, Apo A-I, log-transformed Lp(a) and triglycerides, and log-transformed hs-CRP | Baseline, Week 14 |
All cardiovascular outcomes are reported in the registry in units of events per 100 participant-years, but every between-group comparison is expressed as a hazard ratio from a stratified Cox model. The event rate describes each arm's absolute incidence; the hazard ratio is the relative comparison.
6. Primary Endpoint Results
The primary analysis compared the time to first adjudicated major CV event between bococizumab and placebo in the FAS. The registry labels the comparison "Placebo vs Bococizumab"; the hazard ratios are read here as bococizumab relative to placebo, the same orientation implied by the negative LDL-C differences reported for the lipid endpoints.
Hazard ratio for first major cardiovascular event
95% CI: 0.80–1.22 · log-rank P = 0.930905
Stratified Cox proportional-hazards model (geographic region; pre-screening LDL-C <100 vs ≥100 mg/dL); superiority hypothesis
| Endpoint | Method | HR | 95% CI (two-sided) | P-value |
|---|---|---|---|---|
| First major CV event | Stratified Cox model; log-rank test | 0.99 | 0.80–1.22 | 0.930905 |
What the estimate means. An HR of 0.99 means the estimated instantaneous rate of a first major CV event in the bococizumab group was 99% of that in the placebo group, a 1% lower estimated hazard, averaged over the follow-up actually observed. For practical purposes the point estimate sits at the null value of 1.
What it does not mean. It does not show that bococizumab has no effect on cardiovascular events. A trial that stops early, with limited follow-up, can produce an estimate near 1 either because there is truly little effect or because the effect of LDL-C lowering takes time to emerge and the trial did not run long enough to observe it. The HR also says nothing about absolute risk; that requires the per-arm event rates per 100 participant-years.
Precision. The 95% CI of 0.80 to 1.22 is compatible with anything from a 20% lower hazard to a 22% higher hazard. That range includes effects that would be clinically meaningful in either direction, so the result is better described as inconclusive than as evidence of equivalence. SPIRE-1 was a superiority trial with no equivalence or non-inferiority margin, and a wide interval around 1 cannot be converted into a claim of "no difference".
The p-value. P = 0.930905 indicates that data like these are entirely unremarkable if the true HR were 1. It does not measure the size of the effect, and a large p-value is not the probability that the null hypothesis is true. The information about magnitude lives in the HR and its interval, not in the p-value.
Cautions. The log-rank test and Cox HR are most interpretable when hazards are roughly proportional; a treatment whose benefit accrues gradually would violate that assumption and would be summarized poorly by a single average HR over a short follow-up. The analysis population is the FAS rather than all randomized participants, and participants who did not have an event were censored at the end of their follow-up, which in a terminated trial is driven largely by the stopping date.
7. Secondary Cardiovascular Outcomes
Every secondary cardiovascular outcome used the same framework as the primary analysis: time to first adjudicated event, log-rank test, and an HR with 95% CI from a Cox model stratified by geographic region and pre-screening LDL-C, all in the FAS under a superiority hypothesis.
| Outcome (first occurrence unless stated) | HR | 95% CI | P-value |
|---|---|---|---|
| CV death, non-fatal MI or non-fatal stroke | 1.03 | 0.82–1.30 | 0.784265 |
| All-cause death, non-fatal MI, non-fatal stroke or hospitalization for unstable angina needing urgent revascularization | 0.99 | 0.81–1.20 | 0.892441 |
| All-cause death, non-fatal MI or non-fatal stroke | 1.02 | 0.83–1.26 | 0.845797 |
| CV death, non-fatal MI, non-fatal stroke and hospitalization for unstable angina | 0.98 | 0.80–1.21 | 0.883797 |
| Hospitalization for unstable angina needing urgent revascularization | 0.82 | 0.49–1.36 | 0.431903 |
| Hospitalization for unstable angina | 0.82 | 0.53–1.27 | 0.367069 |
| Cardiovascular death | 1.20 | 0.74–1.95 | 0.455690 |
| All-cause death | 1.12 | 0.79–1.60 | 0.526269 |
| Any MI (fatal or non-fatal) | 1.11 | 0.84–1.48 | 0.469496 |
| Fatal MI | 1.54 | 0.26–9.23 | 0.633022 |
| Non-fatal MI | 1.11 | 0.83–1.48 | 0.467650 |
| Any stroke (fatal or non-fatal) | 0.53 | 0.32–0.89 | 0.015462 |
| Any stroke (fatal or non-fatal), of any etiology | 0.54 | 0.33–0.88 | 0.011863 |
| Fatal stroke | 0.50 | 0.12–2.00 | 0.316066 |
| Non-fatal stroke | 0.52 | 0.30–0.91 | 0.020328 |
| Hospitalization for congestive heart failure | 0.85 | 0.56–1.29 | 0.443081 |
| Coronary revascularization | 0.90 | 0.71–1.12 | 0.343817 |
| CABG | 1.04 | 0.59–1.85 | 0.889065 |
| PCI | 0.88 | 0.69–1.13 | 0.308388 |
| Any arterial revascularization | 1.03 | 0.74–1.42 | 0.874835 |
Reading the pattern
The four composite endpoints tell a consistent story: point estimates of 0.98 to 1.03, with intervals roughly spanning 0.80 to 1.30. Changing the composite definition (CV death versus all-cause death, with or without unstable angina) did not move the estimate materially, which suggests the primary result is not an artefact of a particular composite choice.
Component endpoints are much less precise. The fatal MI interval of 0.26 to 9.23 and the fatal stroke interval of 0.12 to 2.00 are so wide that they are essentially uninformative. Intervals of this width are the signature of very few events: when a Cox model is fitted to a handful of events, the standard error of the log HR becomes large and the interval spans more than an order of magnitude.
8. Lipid and Biomarker Results
The lipid endpoints confirm the pharmacological effect that the trial was built around. Unlike the outcome analyses, these continuous endpoints were analysed with a mixed model for repeated measures (MMRM) or, for the last post-baseline measurement, analysis of covariance (ANCOVA). Differences are least-squares (LS) mean differences between arms.
LDL-C percent change from baseline at Week 14
LS-mean difference, percentage points · 95% CI: −61.43 to −59.71 · P < 0.001
MMRM including observations through Week 70
| Endpoint | Model | LS-mean difference | 95% CI | P-value |
|---|---|---|---|---|
| LDL-C, % change at Week 14 | MMRM (through Week 70) | −60.57 | −61.43 to −59.71 | <0.001 |
| LDL-C, nominal change at Week 14 (mg/dL) | MMRM (through Week 70) | −54.70 | −55.48 to −53.92 | <0.001 |
| LDL-C, % change at last post-baseline measurement (up to Week 140) | ANCOVA | −46.72 | −47.68 to −45.76 | <0.001 |
| Non-HDL-C, % change at Week 14 | MMRM (through Week 70) | −54.88 | −55.66 to −54.09 | <0.001 |
| Total cholesterol, % change at Week 14 | MMRM (through Week 70) | −36.51 | −37.07 to −35.95 | <0.001 |
| VLDL-C, % change at Week 14 | MMRM (through Week 70) | −20.17 | −21.42 to −18.93 | <0.001 |
| RLP-C, % change at Week 14 | MMRM (through Week 70) | −30.25 | −32.44 to −28.07 | <0.001 |
| Apo B, % change at Week 14 | MMRM (through Week 52) | −58.55 | −59.42 to −57.69 | <0.001 |
| HDL-C, % change at Week 14 | MMRM (through Week 70) | 6.14 | 5.69 to 6.59 | <0.001 |
| Apo A-I, % change at Week 14 | MMRM (through Week 52) | 3.53 | 3.02 to 4.03 | <0.001 |
| Lp(a), log-transformed, Week 14 | MMRM on log scale (through Week 52) | 0.67 | 0.66 to 0.68 | <0.001 |
| Triglycerides, log-transformed, Week 14 | MMRM on log scale (through Week 70) | 0.81 | 0.80 to 0.81 | <0.001 |
| hs-CRP, log-transformed, Week 14 | MMRM on log scale (through Week 52) | 1.06 | 1.03 to 1.10 | <0.001 |
An LS-mean difference of −60.57 means that, after adjusting for baseline LDL-C, visit, geographic region and pre-screening LDL-C category, the average percent change from baseline at Week 14 was about 60.57 percentage points lower in the bococizumab group than in the placebo group. It is a difference in percent changes, not a statement that each participant's LDL-C fell by 60.57%.
The very narrow intervals reflect the large sample and the low noise of a laboratory measurement. They are a reminder that precision and clinical relevance are different things: the lipid effect is estimated extremely precisely, while the clinical outcome effect, estimated from comparatively few events, is not.
The later LDL-C estimate (−46.72 at the last post-baseline measurement, up to Week 140) is smaller than the Week 14 estimate (−60.57). The two come from different models and time points, so they are not a direct test of attenuation, but the gap is consistent with a lipid-lowering effect that was less pronounced by the participants' final measurement than at Week 14.
For Lp(a), triglycerides and hs-CRP, the models were fitted to log-transformed values, and the posted estimates (0.67, 0.81, 1.06) are on a ratio-like scale rather than in percentage points. If read as ratios of geometric means relative to placebo, 0.67 corresponds to 33% lower Lp(a), 0.81 to 19% lower triglycerides, and 1.06 to 6% higher hs-CRP. Because the registry labels these as LS-mean differences in percent change, the exact back-transformation should be confirmed against the full statistical report before quoting them.
9. Safety: Serious Adverse Events
The registry reports serious adverse events by arm as the number of participants affected over the number at risk.
| Arm | Participants with serious adverse events / at risk |
|---|---|
| Bococizumab (PF-04950615) | 1060 / 8386 |
| Placebo | 986 / 8374 |
These counts are descriptive. No formal statistical comparison of serious adverse events is posted, and adverse-event tables are not designed as hypothesis tests: they aggregate many different event types, and differences between arms should be interpreted in light of exposure time, event categories and clinical plausibility rather than a single p-value. The denominators here (participants at risk) also differ from the FAS used for efficacy, which is typical: efficacy follows the randomized comparison, whereas safety summarizes participants exposed to study treatment.
10. Statistical Methodology
Stratified log-rank test
The p-values for all cardiovascular outcomes come from a log-rank test, which compares observed and expected numbers of events in each arm at every event time. Stratifying by geographic region and pre-screening LDL-C means the observed-minus-expected comparison is made within strata and then combined, so differences in baseline risk between regions or LDL-C categories do not contaminate the treatment comparison.
where Ot and Et are the observed and expected events in one arm at event time t and Vt is the hypergeometric variance. The test is most powerful when hazards are proportional.
Stratified Cox proportional-hazards model
Hazard ratios and 95% CIs came from a Cox model stratified by geographic region and pre-screening LDL-C, with treatment as the only covariate. Stratification lets each stratum have its own baseline hazard while estimating a single common treatment HR, which is less restrictive than entering the stratification factors as covariates.
HR < 1 indicates a lower estimated instantaneous event rate with bococizumab. The HR is a model-based relative measure averaged over follow-up; it is not a relative risk, an absolute risk difference, or the proportion of participants who benefit.
Event rates per 100 participant-years
Each cardiovascular outcome is also summarized as events per 100 participant-years at risk. This incidence rate divides the number of first events by total follow-up time, which accommodates the variable follow-up created by staggered enrollment and early termination. It implicitly assumes a roughly constant rate over time; the Cox model does not require that assumption.
Mixed model for repeated measures (MMRM)
Week 14 lipid and biomarker endpoints were analysed with MMRM models including observations through Week 70 (Week 52 for Apo B, Apo A-I, Lp(a) and hs-CRP). Fixed effects were treatment group, visit, treatment-by-visit interaction, baseline value, baseline-by-visit interaction, geographic region and pre-screening LDL-C category. Using all visits, rather than Week 14 alone, lets the model borrow information across time through the within-participant correlation and provides valid estimates under a missing-at-random assumption without imputing values.
Analysis of covariance (ANCOVA)
Percent change in LDL-C at the last post-baseline measurement was analysed with ANCOVA, with fixed effects for treatment, baseline LDL-C, geographic region and pre-screening LDL-C category. Because there is one value per participant, a single-time-point regression model is appropriate, and baseline adjustment removes variation explained by the starting level.
Log transformation for skewed biomarkers
Lp(a), triglycerides and hs-CRP were modelled on the log scale. These biomarkers are typically right-skewed, and modelling logs turns multiplicative treatment effects into additive ones; back-transformed results are then ratios of geometric means rather than differences in arithmetic means.
11. Early Termination and Multiplicity
Why termination limits power
Why external stopping matters for bias
Because termination was driven by a program decision rather than by the accumulating results, the stopping itself does not bias the treatment comparison in the way stopping for efficacy or harm can. The main cost is lost information, not distorted estimates.
Many secondary comparisons
The registry reports 20 secondary time-to-event comparisons and 13 lipid and biomarker comparisons. Without a described hierarchical or adjusted testing procedure, secondary p-values are nominal.
Primary endpoint first
In a conventional gatekeeping framework, secondary endpoints are tested confirmatorily only after the primary endpoint is significant. With a primary P of 0.930905, no secondary efficacy claim would pass such a gate.
| Analysis | Role | Interpretation |
|---|---|---|
| First major CV event | Primary endpoint | Confirmatory superiority comparison; not statistically significant |
| Composite CV secondary endpoints | Secondary | Supportive; consistent with the primary result |
| Individual components (MI, stroke, death, revascularization) | Secondary | Lower precision; nominal p-values; stroke findings hypothesis-generating |
| Lipids and biomarkers | Secondary | Pharmacodynamic confirmation of LDL-C lowering |
| Health care resource utilization | Planned | Not evaluated after program discontinuation |
12. Statistical Methods Explained
Why is an HR of 0.99 with a 95% CI of 0.80 to 1.22 not proof that bococizumab has no effect?
The trial tested superiority. Failing to reject the null hypothesis means the data are compatible with no difference, but the interval is equally compatible with a 20% lower or a 22% higher hazard. To claim "no meaningful effect", a trial would need a prespecified equivalence margin and an interval lying entirely within it. SPIRE-1 had neither, and early termination limited the events available to narrow the interval.
Why were the outcome analyses stratified by region and pre-screening LDL-C?
These factors are plausibly related to baseline cardiovascular risk and were used to organize the randomization. A stratified Cox model lets each stratum have its own baseline hazard, so the treatment HR is estimated from within-stratum comparisons. This respects the design and can improve efficiency without imposing a parametric form on how region or LDL-C category affects risk.
Why is the lipid effect so precise while the outcome effect is so uncertain?
Why use MMRM for Week 14 LDL-C but ANCOVA for the last measurement?
The Week 14 endpoint is one visit within a repeated-measures series, so an MMRM that models all scheduled visits jointly uses the correlation between visits and handles missing visits under a missing-at-random assumption. "Last post-baseline measurement" produces exactly one value per participant at varying times, so there is no visit structure to model, and ANCOVA with baseline adjustment is the natural choice.
Should the stroke results (HR 0.53, P = 0.015462) be taken as a finding?
They should be treated as a signal worth examining in other data, not as a demonstrated effect. The primary endpoint was not significant, the stroke outcomes are among many secondary comparisons with no described multiplicity control, and several stroke endpoints overlap, so their small p-values are not independent pieces of evidence. The corresponding interval (0.32 to 0.89) is also wide, reflecting a modest number of stroke events.
Why do the fatal MI and fatal stroke intervals look so extreme?
The standard error of a log HR is driven by the number of events. With very few fatal events, the fatal MI interval spans 0.26 to 9.23. An estimate of 1.54 from such an interval carries almost no information about direction, and it would be a mistake to describe it as showing an increase in fatal MI.
13. Limitations
- Early termination: the trial stopped when the development program was discontinued, limiting follow-up to a maximum of up to 3.4 years and reducing the number of events available for the primary comparison.
- Limited precision for the primary endpoint: the 95% CI of 0.80 to 1.22 includes both clinically important benefit and harm.
- Proportional hazards over a short horizon: if the benefit of LDL-C lowering grows over time, a single HR over a short follow-up may understate later effects; the registry does not report time-varying analyses.
- Multiplicity: numerous overlapping secondary endpoints were analysed without a described adjustment procedure, so nominally significant results (the stroke outcomes) are exploratory.
- Analysis population: the FAS excludes duplicate-randomization participants and all participants at Site 3027; the exclusions are unrelated to treatment, but the population is not all randomized participants.
- Attenuation of lipid effect: the LDL-C difference at the last measurement (−46.72) was smaller than at Week 14 (−60.57), which matters for interpreting the achieved exposure behind the outcome results.
- Log-scale biomarker estimates: the scale of the posted Lp(a), triglyceride and hs-CRP estimates needs confirmation before quantitative use.
- Unevaluated endpoints: health care resource utilization endpoints were not evaluated.
- Reporting scope: the ClinicalTrials.gov record does not report subgroup analyses or Kaplan-Meier curves, so heterogeneity and the timing of any effect cannot be assessed from the registry alone.
14. Why This Trial Matters Statistically
SPIRE-1 is a valuable teaching case precisely because its primary result is null. It shows how a very large randomized trial can deliver a precise and unmistakable biomarker effect alongside an imprecise clinical-outcome estimate, and why "not significant" and "no effect" are different statements.
| Concept | How it appears in SPIRE-1 |
|---|---|
| Superiority testing | Primary HR 0.99 with P = 0.930905: failure to demonstrate superiority, not a demonstration of equivalence |
| Confidence intervals | 0.80–1.22 conveys the range of effects compatible with the data |
| Hazard ratio | Relative effect for all CV outcomes from a stratified Cox model |
| Log-rank test | Stratified time-to-event comparison for every CV outcome |
| Stratification | Geographic region and pre-screening LDL-C used in both survival and lipid models |
| Events drive precision | Narrow intervals for composites, very wide intervals for fatal MI and fatal stroke |
| MMRM | Repeated-measures analysis of Week 14 lipid endpoints |
| ANCOVA | Baseline-adjusted analysis of LDL-C at last measurement |
| Multiplicity | Nominal stroke significance amid a non-significant primary endpoint |
| Early termination | External stopping reduces information without biasing the comparison |
| Surrogate vs clinical endpoint | Large LDL-C reduction without a demonstrated reduction in major CV events over this follow-up |
Statistically, SPIRE-1 did not show that bococizumab reduced the hazard of a first major cardiovascular event, and the interval around the estimate is too wide to rule out modest benefit or harm. Clinically, the trial confirms substantial LDL-C lowering at Week 14 but, because it was cut short, it does not answer how that lipid reduction translates into outcomes over longer exposure. The two conclusions are compatible; they describe different endpoints measured with very different amounts of information.
15. Related Tutorials
Learn more about the methods used in this trial:
16. Related Calculators
17. Sources
- ClinicalTrials.gov: NCT01975376 — The Evaluation of Bococizumab (PF-04950615; RN316) in Reducing the Occurrence of Major Cardiovascular Events in High Risk Subjects.
- PubMed: PMID 28304242
- PubMed: PMID 29685591
- PubMed: PMID 29716940
- PubMed: PMID 35277540
Keep exploring clinical trial statistics
Follow the methods behind SPIRE-1 into step-by-step tutorials, try them in a calculator, or compare this trial with other published results.
Record Summary
SPIRE-1 randomized 16,784 high-risk participants to bococizumab or placebo and was terminated when the bococizumab development program was discontinued. Bococizumab produced a large, precisely estimated reduction in LDL-C at Week 14 (LS-mean difference −60.57 percentage points), but the hazard ratio for a first major cardiovascular event was 0.99 (95% CI 0.80–1.22; P = 0.930905). The most useful reading combines the relative effect, the width of its interval, the information lost to early termination, and the distinction between confirmatory and nominal results among the many secondary endpoints.