Effect of Semaglutide on the Inflammatory Biomarker High-Sensitivity CRP in Patients With Established Cardiovascular Disease and Overweight or Obesity in SELECT: A Prespecified Secondary Analysis
- Design
- Prespecified secondary (sub-study) analysis of a randomized, double-blind, placebo-controlled, parallel-group phase 3 trial (select) · 17604 participants · Mixed outcome
- Match to healthy normal-weight adults aged 55–75
- Different populationObese secondary-prevention population; but the early, weight-loss-independent hs-CRP fall is the most relevant human inflammation signal available. Biomarker, not outcome.
- Could weight loss explain it?
- Specifically testedmethod: Combination of (1) temporal analysis — hsCRP reduction with semaglutide was 'evident by 4 and 8 weeks,' i.e., preceding the period of major weight loss; (2) subgroup analysis — hsCRP reduction 'occurred among those without weight loss' (non-losers); and (3) statistical modeling described only as 'modeling suggests decreased inflammation as contributing in part to the benefits seen with semaglutide' linking hsCRP change to MACE risk reduction. The abstract does not name a specific mediation-analysis method (e.g., causal mediation, formal weight-adjusted Cox model) beyond 'multiple approaches, including Cox modeling.'; prespecified: NI; key limitation: None of these specific weight-independence analyses (early 4-/8-week hsCRP timepoints, non-weight-loser subgroup, or the mediation/modeling statement) could be located in the protocol or SAP text provided — only the standard week 0-to-week 104 hsCRP ratio-to-baseline endpoint is documented there — so whether they were prespecified, and their full methodology (denominators, handling of missing early timepoints, adjustment set), is not verifiable from the sources available to this pass; full text and any dedicated biomarker-substudy protocol/SAP would be needed to confirm.
Adjusted for: weight loss (time-varying), LDL cholesterol, statin use - Study tier
- Study tier 2Prespecified analysis of a large RCT; biomarker endpoint with modelled mediation.
- Assessment
- Version 5 · Checked against the source by AI · Oct 19, 2026
Study facts come from the paper. Population match, weight-loss explanation and study tier are our judgments, made against the reference group of healthy normal-weight adults aged 55–75.
In SELECT, semaglutide lowered hs-CRP by 38% and the fall began within weeks, before major weight loss and even in people who did not lose weight, although larger hs-CRP falls went with greater weight loss. This suggests an anti-inflammatory effect not fully explained by weight loss, but it is a blood marker in people with obesity and heart disease, in a trial funded by Novo Nordisk with Novo Nordisk employees among the authors.
01Findings
What the study reported
- Drugs
- Semaglutide
- Dose
- 2.4 mg weekly
- Route
- subcutaneous
- Comparator
- placebo
- Primary outcome
- Prespecified: baseline hs-CRP as MACE predictor; hs-CRP change vs weight loss and MACE
- Effect
- hs-CRP -37.8% at 104 weeks; reductions evident by 4-8 weeks, before major weight loss, and in those without weight loss; hs-CRP reduction prognostic of lower MACE
- 95% confidence interval
- Not extracted
- Follow-up
- 104-208 weeks
- Adverse events
- Not extracted
- Limitations
- Not extracted
Who was studied
- Condition
- Established/known atherosclerotic cardiovascular disease with overweight or obesity, without diabetes
- Mean age
- not reported; eligibility required age 45 years or older
- Mean bmi
- not reported; eligibility required BMI 27 kg/m2 or higher, established ASCVD and no diabetes
- Diabetes status
- No diabetes (history of type 1 or type 2 diabetes was an exclusion criterion)
- Sample size
- 17604
Study quality details
- Study design
- Prespecified secondary (sub-study) analysis of a randomized, double-blind, placebo-controlled, parallel-group Phase 3 trial (SELECT)
- Sample size
- 17604
- Randomization
- yes (parent trial)
- Blinding
- not stated
- Comparator
- placebo
- Follow up duration
- 104-208 weeks
- Outcome type
- biomarker plus association with MACE
- Replication
- not assessed (auto)
- Consistency with other evidence
- not assessed (auto)
- Population applicability
- INDIRECT
- Statistical precision
- not extracted
- Funding conflicts
- unclear
- Peer review status
- yes
- Rob2
- O1: {"D1":{"judgment":"Low","rationale":"Large industry-sponsored Phase 3 trial (17,604 randomized) using a centralized IWRS for randomization and allocation, double-blind with visually identical active/placebo product; baseline hsCRP was near-identical between arms (1.96 vs 1.91 mg/L geometric mean), consistent with adequate concealment/randomization. Per the domain rule this pattern warrants PY on allocation concealment absent contradi"},"D2":{"judgment":"Low","rationale":"Per the domain rule, D2 for this lab biomarker outcome is judged on discontinuation imbalance and ITT handling rather than unblinding. Trial-completion imbalance between arms was modest (NOT COMPLETED 259/8803 semaglutide vs 284/8801 placebo), and the SAP specifies an intention-to-treat estimand (FAS, irrespective of adherence) for all objectives, with the hsCRP imputation model fitted on data 'ir"},"D3":{"judgment":"Low","rationale":"7472/8803 (84.9%) semaglutide and 7400/8801 (84.1%) placebo participants contributed to the week-104 hsCRP result: about 15% of randomised participants lack the outcome and the between-arm difference in availability is 0.8 percentage points, i.e. below both of the guide's thresholds (>20% missing, or >5-point between-arm gap), so \"data available for nearly all\" is PY. Missing values were handled b"},"D4":{"judgment":"Low","rationale":"hsCRP is a central-lab, objective biomarker measurement. The protocol specifies that laboratory analyses (including the panel containing hsCRP) are performed by a central laboratory throughout the trial, with no evidence in the available sources of an assay change partway through or of differential measurement procedures between arms. Per the domain rule, this defaults to Low."},"D5":{"judgment":"Low","rationale":"The outcome, timepoint, transformation, model and missing-data handling were all prespecified in the protocol (v7.0, 09 Feb 2022) and SAP (v3.0, 22 Apr 2022), both finalised well before primary completion (21 Jun 2023) and before any unblinded data; the emphasised timepoint (104 weeks) is exactly the prespecified one, and the corresponding result is posted in the registry, so there is no sign of s"},"overall":{"judgment":"Low"},"result":"Change in hsCRP (ratio to baseline), semaglutide vs placebo, at week 104; registered secondary continuous endpoint 'Change in High Sensitivity C-Reactive Protein (hsCRP) - Ratio to Baseline' (FAS, n=7472 semaglutide / 7400 placebo; geometric mean ratio 0.61 vs 1.00); abstract reports this as '-37.8% [104 weeks]' in the RESULTS section, no table/figure identifiable (full text not available).","passes":[{"pass":"A","model":"claude-sonnet"},{"pass":"B","model":"claude-opus"}],"guide_version":"rob2-guide v1 + v1.1 calibration rulings (2026-09-13)","label":"Human-reviewed","resolution":"agreed domains accepted; disagreements decided by the owner 2026-09-13 (IN-009)"}
O4: {"D1":{"judgment":"Low","rationale":"Same trial-level randomization/allocation infrastructure as O1 (central IWRS, double-blind, identical-appearing product); no result-specific randomization concern for the weight-independence analyses."},"D2":{"judgment":"Some concerns","rationale":"The weight-independence claims rest on a non-loser subgroup and early (4-/8-week) hsCRP trajectories; no protocol/SAP text was found defining an intercurrent-event/estimand strategy for these specific sub-analyses, so it is unclear how deviations (discontinuation, non-adherence, dose interruption during up-titration through week 4-8) were handled for them specifically, beyond the trial's general I"},"D3":{"judgment":"Some concerns","rationale":"No information in the abstract, registry, protocol, or SAP on the number of participants contributing to the week 4/week 8 hsCRP measurements or to the 'without weight loss' subgroup used for O4; denominators and missingness for these specific analyses are not stated anywhere in the available sources. Given O1's overall hsCRP missingness is already ~15%/arm, and week 4/8 draws would be additional/"},"D4":{"judgment":"Low","rationale":"hsCRP measurement itself is via the same central laboratory used for the week-104 endpoint (per protocol, all specified lab tests are run centrally); no evidence of a different/changed assay for the earlier (4-/8-week) draws. Objective biomarker measurement defaults to Low per the domain rule."},"D5":{"judgment":"High","rationale":"The paper is titled a \"prespecified secondary analysis\", but the only hsCRP analysis prespecified anywhere in the protocol or SAP is change from randomisation to year 2 (week 104), on the log scale, by MI + ANCOVA. Neither document contains any week-4 or week-8 hsCRP analysis, any subgroup defined by weight change, or any mediation analysis, and the SAP's subgroup list is entirely baseline-defined"},"overall":{"judgment":"High"},"result":"Weight-independence of the hsCRP effect: (a) temporal precedence of hsCRP reduction over weight loss, 'evident by 4 and 8 weeks'; (b) hsCRP reduction 'occurred among those without weight loss' (implied non-weight-loser subgroup); (c) mediation-type modeling statement that 'decreased inflammation [is] contributing in part to the benefits seen with semaglutide in SELECT' on MACE. All from the abstract METHODS/RESULTS; no table/figure identifiable (full text not available), and none of these specific analyses (4-/8-week hsCRP timepoints, non-loser subgroup, mediation model) appear in the available protocol or SAP text.","passes":[{"pass":"A","model":"claude-sonnet"},{"pass":"B","model":"claude-opus"}],"guide_version":"rob2-guide v1 + v1.1 calibration rulings (2026-09-13)","label":"Human-reviewed","resolution":"agreed domains accepted; disagreements decided by the owner 2026-09-13 (IN-009)"}
Methodological notes
Calibration two-pass assessment 2026-09-13 (drafts in data/assessments/42610271/); sources: abstract, registry, public protocol/SAP where available; no paper full text.
01bRisk of bias
How much the result can be relied on
Risk of bias is judged per result, not per paper: the same study can report one marker at low risk and another at high. Two reviewers assess each result independently against a written guide, and every domain judgment below carries its reasoning and the sentences it rests on.
Risk of bias is not the same as whether the numbers work. A trial can be well conducted and still print a standard deviation that is arithmetically impossible, or a table whose sign contradicts its own text. Risk of bias has no question for that, so where it was found it is recorded separately below, under Source integrity.
These words are not quality scores. On the RoB 2 scale, High risk of bias is the worst rating a result can receive and Low risk of bias is the best — the opposite of how the same words read on some other scales. Levels are always written out in full here for that reason.
O1
Low risk of bias · RoB 2 · Two reviewers differed; settled by a recorded ruling
How the overall was reached: All five domains Low (RoB 2 algorithm: overall equals the worst domain); the endpoint is a prespecified, registry-posted, centrally assayed ITT comparison with ~15% balanced missingness handled by a prespecified imputation plan.
| Domain | Judgment and reasoning |
|---|---|
| D1 · Randomisation process Low risk of bias | Large industry-sponsored Phase 3 trial (17,604 randomized) using a centralized IWRS for randomization and allocation, double-blind with visually identical active/placebo product; baseline hsCRP was near-identical between arms (1.96 vs 1.91 mg/L geometric mean), consistent with adequate concealment/randomization. Per the domain rule this pattern warrants PY on allocation concealment absent contradicting evidence.All subjects will be centrally randomised using an IWRS and assigned to the next available This is a randomised, double-blind, parallel group, placebo-controlled trial comparing semaglutide |
| D2 · Deviations from intended interventions Low risk of bias | Per the domain rule, D2 for this lab biomarker outcome is judged on discontinuation imbalance and ITT handling rather than unblinding. Trial-completion imbalance between arms was modest (NOT COMPLETED 259/8803 semaglutide vs 284/8801 placebo), and the SAP specifies an intention-to-treat estimand (FAS, irrespective of adherence) for all objectives, with the hsCRP imputation model fitted on data 'irrespective of adherence to randomised treatment'. No arm-specific GI-adverse-event rates were available in the provided sources to assess unblinding directly (NI on that sub-question), but the guide states awareness alone rarely biases a lab biomarker result."type": "NOT COMPLETED" The estimand for all objectives is an intention-to-treat estimand, evaluating the effect of the |
| D3 · Missing outcome data Low risk of bias | 7472/8803 (84.9%) semaglutide and 7400/8801 (84.1%) placebo participants contributed to the week-104 hsCRP result: about 15% of randomised participants lack the outcome and the between-arm difference in availability is 0.8 percentage points, i.e. below both of the guide's thresholds (>20% missing, or >5-point between-arm gap), so "data available for nearly all" is PY. Missing values were handled by a prespecified arm-specific multiple-imputation model. Noted but not judgment-changing: no jump-to-reference or tipping-point sensitivity analysis was prespecified for hsCRP (only for body weight), so an informative-missingness assumption is untested."counts": [{"groupId": "OG000", "value": "7472"}, {"groupId": "OG001", "value": "7400"}] The continuous supportive secondary endpoints (change from baseline to week 104) are analysed using multiple imputation for missing values. |
| D4 · Measurement of the outcome Low risk of bias | hsCRP is a central-lab, objective biomarker measurement. The protocol specifies that laboratory analyses (including the panel containing hsCRP) are performed by a central laboratory throughout the trial, with no evidence in the available sources of an assay change partway through or of differential measurement procedures between arms. Per the domain rule, this defaults to Low.The laboratory analyses will be performed by a central laboratory, unless otherwise High sensitivity C-Reactive Protein (hsCRP) (mg/L) |
| D5 · Selection of the reported result Low risk of bias | The outcome, timepoint, transformation, model and missing-data handling were all prespecified in the protocol (v7.0, 09 Feb 2022) and SAP (v3.0, 22 Apr 2022), both finalised well before primary completion (21 Jun 2023) and before any unblinded data; the emphasised timepoint (104 weeks) is exactly the prespecified one, and the corresponding result is posted in the registry, so there is no sign of selection among timepoints or transformations for this particular estimate. The paper also presents hsCRP categorically (<2, 2-<10, >=10 mg/L), but that is used for the baseline-prognosis result, not for the treatment effect assessed here; the residual caveat is that without the full text one cannot confirm the -37.8% is the ANCOVA treatment ratio rather than a descriptive summary (the abstract gives no CI).Change from randomisation to year 2 (visit 14) in: ... High sensitivity C-Reactive Protein (hsCRP) (mg/L) Analyses of lipids and hsCRP are on logarithmic scale. |
O4
High risk of bias · RoB 2 · Two reviewers differed; settled by a recorded ruling
How the overall was reached: RoB 2 algorithm: overall takes the worst domain - D5 is High (no verifiable prespecification of the timepoints, subgroup or model behind the weight-independence claim), compounded by Some concerns in D1, D2 and D3 arising from a post-randomization, treatment-affected stratification and complete absence of information on early-timepoint data availability.
| Domain | Judgment and reasoning |
|---|---|
| D1 · Randomisation process Low risk of bias | Same trial-level randomization/allocation infrastructure as O1 (central IWRS, double-blind, identical-appearing product); no result-specific randomization concern for the weight-independence analyses.All subjects will be centrally randomised using an IWRS and assigned to the next available The IWRS is used for blind-breaking instructions. |
| D2 · Deviations from intended interventions Some concerns about risk of bias | The weight-independence claims rest on a non-loser subgroup and early (4-/8-week) hsCRP trajectories; no protocol/SAP text was found defining an intercurrent-event/estimand strategy for these specific sub-analyses, so it is unclear how deviations (discontinuation, non-adherence, dose interruption during up-titration through week 4-8) were handled for them specifically, beyond the trial's general ITT/FAS approach. Per the guide, D2 is judged on discontinuation/ITT handling rather than unblinding; the general trial-level ITT approach is documented, but its specific application to these substudy analyses is NI.The estimand for all objectives is an intention-to-treat estimand, evaluating the effect of the |
| D3 · Missing outcome data Some concerns about risk of bias | No information in the abstract, registry, protocol, or SAP on the number of participants contributing to the week 4/week 8 hsCRP measurements or to the 'without weight loss' subgroup used for O4; denominators and missingness for these specific analyses are not stated anywhere in the available sources. Given O1's overall hsCRP missingness is already ~15%/arm, and week 4/8 draws would be additional/earlier timepoints not listed among the registered outcome measures at all, missingness for O4 could plausibly be as large or larger; absent any data, the guide's principle that NI on a key question usually means Some concerns (not Low) applies. |
| D4 · Measurement of the outcome Low risk of bias | hsCRP measurement itself is via the same central laboratory used for the week-104 endpoint (per protocol, all specified lab tests are run centrally); no evidence of a different/changed assay for the earlier (4-/8-week) draws. Objective biomarker measurement defaults to Low per the domain rule.The laboratory analyses will be performed by a central laboratory, unless otherwise |
| D5 · Selection of the reported result High risk of bias | The paper is titled a "prespecified secondary analysis", but the only hsCRP analysis prespecified anywhere in the protocol or SAP is change from randomisation to year 2 (week 104), on the log scale, by MI + ANCOVA. Neither document contains any week-4 or week-8 hsCRP analysis, any subgroup defined by weight change, or any mediation analysis, and the SAP's subgroup list is entirely baseline-defined. The reported claim is therefore drawn from a large space of eligible timepoints, transformations (ratio-to-baseline, percent change, categories) and subgroup definitions ("without weight loss" is nowhere defined), with no verifiable analysis plan - PY to both "multiple eligible outcome measurements" and "multiple eligible analyses" under the RoB 2 algorithm.Change from randomisation to year 2 (visit 14) in: ... High sensitivity C-Reactive Protein (hsCRP) (mg/L) In this prespecified SELECT substudy, we evaluated ... the relationships between changes in hsCRP levels and time to first MACE, baseline body weight, weight loss, and other clinical measures among treatment groups over time (104-208 weeks) using multiple approaches, including Cox modeling. |
02Funding
Funding and conflicts
- Funding
- Novo Nordisk A/S (trial sponsor)
- Industry funded
- Yes
- Manufacturer
- Novo Nordisk A/S
- Sponsor role
- Trial funded by Novo Nordisk A/S; several authors are Novo Nordisk employees; the sponsor's role in the analysis is not stated (the first author vouches for the analyses).
- Author conflicts
- Authors report Novo Nordisk relationships; sponsor co-authors.
- Independent replication
- no (single trial dataset)
Funding is shown on every study and never used to score it.
03Claims
Claims this study bears on
- GLP-1 receptor agonist treatment reduces systemic inflammatory markers beyond what can be explained by treatment-associated weight loss.Mixed
SELECT prespecified hsCRP analysis in adults with CVD and obesity: hsCRP fell 37.8% at 104 weeks, and the reduction began by weeks 4 and 8 and was seen in people without weight loss. Greater reductions went with greater weight loss, the non-loser reduction was smaller (ETR 0.86 vs 0.60 with at least 2% weight loss), and the early-timepoint and non-loser analyses are high risk of bias (not locatable in the protocol or SAP). Novo Nordisk funded.
- The cardiovascular benefit of GLP-1 receptor agonists is not explained by weight loss.Consistent, but doesn't test the claim
Prespecified SELECT analysis (ASCVD with overweight or obesity, no diabetes): semaglutide lowered hsCRP, including before major weight loss and in people without weight loss, but greater hsCRP reductions went with greater weight loss; modelling attributes part of the MACE benefit to inflammation, but hsCRP is a biomarker and the pathway is not separated from weight loss.
04Source
The source, as retrieved
Abstract
[BACKGROUND] In SELECT (Semaglutide Effects on Heart Disease and Stroke in Patients With Overweight or Obesity), among 17 604 patients with known atherosclerotic cardiovascular disease and overweight or obesity, but not diabetes, randomization to the glucagon-like peptide-1 receptor antagonist semaglutide significantly reduced the primary outcome of major adverse cardiovascular events (MACEs; cardiovascular death, nonfatal myocardial infarction, or nonfatal stroke) compared with placebo (mean follow-up, 39.8 months). Inflammation, as indicated by plasma hsCRP (high-sensitivity C-reactive protein) level, is implicated as a biomarker predicting cardiovascular risk in obesity and atherosclerotic cardiovascular disease. SELECT provides a unique opportunity to study the relationship among hsCRP, obesity, weight loss, and MACE outcomes in semaglutide versus placebo groups. [METHODS] In this prespecified SELECT substudy, we evaluated whether baseline hsCRP levels predicted MACE risk and examined the relationships between changes in hsCRP levels and time to first MACE, baseline body weight, weight loss, and other clinical measures among treatment groups over time (104-208 weeks) using multiple approaches, including Cox modeling. [RESULTS] Baseline hsCRP level, which was similar in the semaglutide (geometric mean 1.96 mg/L) and placebo (geometric mean 1.91 mg/L) groups, was prognostic of future MACEs. The risk of MACEs increased across baseline hsCRP level <2, 2-<10, and ≥10 mg/L subgroups, including significant associations with cardiovascular and all-cause death. Semaglutide reduced hsCRP levels (-37.8% [104 weeks]) and risk of MACEs across all hsCRP subgroups. Greater reductions in ratio-to-baseline hsCRP with semaglutide were associated with greater weight loss, but preceded major weight loss, evident by 4 and 8 weeks, and occurred among those without weight loss. Semaglutide-associated changes in hsCRP were independent of low-density lipoprotein cholesterol levels, statin use, and atherosclerotic cardiovascular disease entry criteria. hsCRP reductions were found to be prognostic of decreased risk of MACEs. Modeling suggests decreased inflammation as contributing in part to the benefits seen with semaglutide in SELECT. [CONCLUSIONS] In SELECT, hsCRP data at baseline and in response to treatment with semaglutide support inflammation as a potential prognostic factor associated with cardiovascular risk in these generally well-treated patients with atherosclerotic cardiovascular disease and overweight or obesity but not diabetes. These findings suggest that the MACE reduction observed with semaglutide versus placebo in SELECT may have partially involved a decrease in inflammation. [REGISTRATION] URL: https://clinicaltrials.gov; Unique identifier: NCT03574597.