This phase 2 study (ClinicalTrials.gov identifier NCT03452228) was conducted at 17 sites across four countries. The first patient was enrolled on 7 June 2018 and the last patient was enrolled on 8 July 2019. The study protocol was approved by institutional review boards (IRBs) and/or ethics committees (Quorum Review, Comitato Etico dell Universita, Policlinico Umberto I di Roma, North West – Greater Manchester South Research Ethics Committee, The University of Pennsylvania IRB, The University of Texas Institutional Review Board, Western IRB, Human Research Protection Program, The University of Kansas Medical Center and Copernicus IRB). The study was conducted in accordance with ethical principles originating from the Declaration of Helsinki and was consistent with International Conference on Harmonization/Good Clinical Practices and applicable regulatory requirements. All participants provided written informed consent. The principal investigators and sponsor designed the study protocol and selected participating sites. Monitoring and site supervision were performed by a contract research organization with oversight by the sponsor. The first author wrote all drafts of the manuscript. All authors had access to the data, participated in revisions and vouch for the accuracy and completeness of data and adherence to the protocol.

Adults aged 18 to 75 years with sHTG (fasting serum triglycerides >500 mg dl−1 at screening on two separate occasions; documented medical history of fasting triglycerides ≥1,000 mg dl−1) with a history of hospitalization for AP were enrolled based on genotype according to the presence of LOF mutations in LPL pathway genes. Cohort 1 consisted of patients with FCS (with known bi-allelic LOF mutations in APOA5, APOC2, GPIHBP1, LMF1 or LPL); cohort 2 consisted of patients with MCS (with known heterozygous LOF mutations in APOA5, APOC2, GPIHBP1, LMF1 or LPL); and cohort 3 consisted of patients with MCS and without LPL pathway mutations. Initially patients were enrolled into the aforementioned cohorts based on available genotype information from the patient’s medical history at screening. All patients were subsequently exome sequenced and analyzed by the Regeneron Genetics Center (Regeneron Pharmaceuticals). At screening, 11 patients were missing genotype information; based on the exome sequencing, three of these were subsequently assigned to cohort 1 and eight were assigned to cohort 2. In addition, one patient from original cohort 3 was withdrawn from the study before dosing as they did not meet the eligibility criteria. Thus, the final ‘actual cohort’ assignments for the purpose of analysis were cohort 1, n = 17; cohort 2, n = 15; and cohort 3, n = 19. The full list of patient eligibility criteria is provided in the Supplementary Information.

Patients in each cohort were randomized 2:1 to receive either i.v. evinacumab 15 mg kg−1 every 4 weeks or matching placebo. The study comprised a screening phase, a 4-week single-blind placebo run-in, a 12-week DBTP, a 12-week SBTP and a 20-week off-drug observation phase (Fig. 2). During the SBTP, all patients received i.v. evinacumab 15 mg kg−1 every 4 weeks.

The primary end point of the study was to determine the intra-patient percent change in mean serum triglycerides from baseline following 12 weeks of evinacumab treatment in cohort 3 patients (the 12 weeks of treatment encompassed a combination of the DBTP and SBTP). Cohort 3 was prespecified as the analysis population for the primary end point analysis, as this group of patients had an intact LPL pathway and would thus be expected to respond optimally to evinacumab treatment. To reduce the variability of baseline triglyceride measurements for patients randomized to evinacumab in the DBTP, baseline was defined as the geometric mean of all available triglyceride results at days –28, –14 and 1. Similarly, for those switching from placebo to evinacumab in the SBTP, baseline was defined as the geometric mean of all available triglyceride results at weeks 6, 8 and 12. The percent change in other lipid/lipoprotein parameters from baseline to weeks 12 and 24 were also evaluated.

All blood sampling for the determination of lipid parameters were determined under fasting conditions (at least 8 h of fasting). Triglycerides and total cholesterol were assessed by an enzymatic colorimetric assay run on a Beckman–Coulter analyzer. HDL-C was determined by precipitation, which involved precipitating all non-HDL-C using 50 kDa dextran sulfate with magnesium ions as the precipitating agent, followed by the determination of HDL-C in the supernatant using an adapted method for determining total cholesterol on a Beckman–Coulter analyzer. LDL-C was determined by ultracentrifugation; after separation of the very-low-density lipoprotein/chylomicron sub-fraction by ultracentrifugation, LDL-C was determined as the cholesterol in the infranatant (performed by an enzymatic colorimetric assay) minus HDL-C. ApoB and ApoA1 were assessed by nephelometry using a Siemens BNII nephelometer. In addition, serial ultracentrifugation was performed to separate lipoprotein subfractions (chylomicrons, very-low-density lipoprotein, intermediate-density lipoprotein, LDL and HDL) and lipids (triglycerides, cholesterol and phospholipids) and proteins (for example, ApoB, ApoA1, ApoC2, ApoC3 and ApoC5) were measured in the fractions by established methods.

Genomic DNA was extracted from peripheral blood samples and submitted for whole exome sequencing at the Regeneron Genetics Center (RGC). Briefly, 1 μg of genomic DNA was fragmented and prepared for exome capture with a custom reagent kit from Kapa Biosystems. Samples were captured using the NimbleGen SeqCap VCRome 2.1 exome target design and sequenced using 75-bp paired-end sequencing on an Illumina HiSeq 2500 with v.4 chemistry. Following sequencing, data were processed using a cloud-based pipeline developed at the RGC that uses DNAnexus and AWS to run standard tools for sample-level data production and analysis. Sequence reads were mapped and aligned to the GRCh37/hg19 human genome reference assembly using BWA-mem. Single-nucleotide polymorphisms and INDEL variants and genotypes were called using GATK’s HaplotypeCaller. Standard quality-control filters were applied to called variants. Passing variants were classified, annotated and analyzed using an RGC-implemented Mendelian analysis pipeline to evaluate their potential functional effects. Variants were annotated for their observed frequencies in population control databases such as dbSNP, the 1000 Genomes Project, the Exome Aggregation Consortium Database and internal RGC databases to filter out common polymorphisms and high frequency, likely benign variants. Algorithms for bioinformatic prediction of functional effects of variants (LRT, Poly-phen2, SIFT, CADD and Mutation Taster), along with conservation scores, were incorporated as part of the annotation process of variants and used to inform on the potential deleteriousness of identified candidate variants. Individuals in this study were screened for variants in a list of 28 genes compiled for their reported associations with triglyceride or lipid levels.

Screened genes were LPL, ANGPTL3, ANGPTL4, ANGPTL8, APOA1, APOA4, APOA5, APOC2, APOC3, APOD, COL18A1, CREB3L3, GALNT2, GPIHBP1, LMF1, PCSK7, MLXIPL, LIPI, USF1, ABCA1, GPD1, GCKR, TRIB1, BTN2A1, LRP8, TIMD4, ABCG5 and ABCG8.

18F-FDG-PET/CT and MRI imaging were performed to assess subclinical signs of pancreatic inflammation and injury in study populations. 18F-FDG-PET/CT was performed at baseline and after 12 weeks of treatment during the DBTP to examine the impact of treatment on pancreatic inflammation using standardized uptake values, SUVmax and SUVmean.

MRI was performed at baseline to assess pancreatic injury/inflammation through measurement of apparent diffusion coefficient and levels of hepatic fat fraction. Patients underwent repeat MRI scans at 12 and 24 weeks of treatment to assess changes in pancreatic injury/inflammation and changes in liver hepatic fat fraction.

The primary end point was prespecified as the least squares mean percent reduction in triglycerides from baseline after 12 weeks of evinacumab exposure (combination of DBTP and SBTP) in cohort 3. This was assessed in cohort 3 using asynchronous study periods that were dependent upon treatment group to assess changes in triglycerides after 12 weeks of evinacumab treatment; for patients randomized to evinacumab, the primary analysis included the 12-week DBTP, whereas for patients randomized to placebo the primary analysis included the subsequent 12-week SBTP of active treatment with evinacumab. Point estimates of mean percent changes in triglycerides between the placebo run-in period and each observation were calculated using a mixed-effect model for repeated measures (MMRM) approach.

Before MMRM analysis of the primary end point, fasting serum triglycerides were log-transformed with the aim to provide a normal data distribution. A log-scale s.d. of 0.5 was used based on available evinacumab phase 1 data. To reduce variability of baseline triglyceride measurements, baseline was the mean of the three log-transformed measurements (day –28, day –14 and week 0 (DBTP); weeks 6, 8 and 12 (SBTP)).

The primary end point analysis was based on the percent change in triglycerides from baseline following 12 weeks of treatment with evinacumab in cohort 3. Point estimates of mean percent changes in triglycerides between the placebo run-in period and each observation were calculated using an MMRM approach. The MMRM model assessed within-patient treatment comparisons using an unstructured covariance matrix while accounting for baseline triglyceride values, study visit and baseline triglyceride values by study visit interaction, but not trough levels of evinacumab, which varied and were often below the targeted threshold of 100 mg l−1. Study visits were adjusted to the start of evinacumab treatment to pool data from the DBTP and the SBTP. Least squares means with CIs and least squares mean ratios with CIs were used to assess treatment effects.

Secondary end points included the percent triglyceride lowering from baseline following 2–24 weeks repeated i.v. doses of evinacumab; the proportion of patients who achieved at least a 40%, 50%, 60%, 70%, 80% or 90% reduction in triglycerides from baseline and the proportion of patients who achieved a reduction in triglycerides below 500 mg dl−1 after 2–24 weeks evinacumab treatment (not reported); the percent change in post-heparin LPL activity from baseline (not reported); changes in patient-reported abdominal and gastrointestinal symptoms, dietary habits and symptom/dietary impact measures, assessed via questionnaires (not reported); the degree of pancreatic injury/inflammation at baseline and change from baseline following 12 weeks of evinacumab treatment (see further details above); the evaluation of evinacumab pharmacokinetics, total ANGPTL3 levels and anti-drug antibodies during the treatment and follow-up periods (not reported); and the incidence and severity of TEAEs, serious adverse events, laboratory abnormalities and other safety variables.

Post hoc analyses were undertaken to evaluate whether (1) mean triglyceride values at baseline and week 12; (2) percent change in mean triglyceride values at week 12; (3) mean log-transformed triglyceride values from baseline and week 12; and (4) change in mean log-transformed triglyceride values from baseline at week 12 achieved normal distribution using the MMRM model. It was essential to determine whether triglyceride values that do not follow a normal distribution are log-normal, as applying MMRM analysis without confirming the data obey a log-normal distribution, could lead to misinterpretation of data. Accordingly, tests of normality, including Anderson–Darling, Cramer–von Mises, Kolmogorov–Smirnov and Shapiro–Wilk, were performed. These tests demonstrated that log-transformed triglyceride values were not normally distributed, thus median percent changes were additionally determined and analyzed.

For the DBTP and SBTP, both percent change and absolute triglyceride values, as well as safety and other efficacy end points, were summarized descriptively. For the DBTP, post hoc nominal P values are provided for descriptive purposes. All efficacy analyses were performed in the full analysis population, which consisted of all randomized patients who received study drug. The safety analysis set included all randomized individuals who received at least one dose, or part of a dose, of study drug. Clinical data were analyzed using SAS v.9.4.

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.