Skip to main content
Tessa NewsLink
Paste a health news link, or browse
Latest checks

Source study found

Story checked

Blood may reveal signs of gut-related disease that stool samples miss (opens in a new tab)

news-medical.net · 2026-11-03

Short answerEvidenceSource

Short answer

Mixed

Mixed.

The claims we could check match the study, but some claims were not covered by the evidence reviewed.

  • 3 supported
  • 2 not covered

Checked against the study summary. The full text wasn't available, so some details couldn't be settled either way.

Share this check

Follow the evidence trail
1
2

NewsLink checks it

Mixed

Every claim we could check holds up. Three of five claims match the study. This overall rating is based only on the claims we could check. Two claims the study doesn't address.

  • 3 supported
  • 2 not covered
Open claim evidence
3
Then inspect each claim

Evidence layer

Claim by claim

Each claim gets a verdict. Expand it to see the evidence directly below.

5 claims in this story

Showing all 5 claimsChoose a verdict to focus the list.

Claim 1 of 5Not coveredThe fecal SCFA model had an overall classification accuracy of 59% in the held-out test set, while the serum SCFA model achieved 95% accuracy; the combined serum-fecal model improved accuracy to 97%.View evidenceHide evidence

As stated59%, 95%, and 97% accuracy

Why this verdict

The reported performance numbers are broadly consistent with the abstract-level profile: fecal accuracy about 59%, serum accuracy 95%, and combined accuracy about 97.4%. However, the claim specifically says these were measured in a held-out test set. The profile at abstract depth reports internal validation methods such as bootstrap validation and permutation testing but does not verify a held-out test-set design or the exact test-set basis for the estimates.

1

Study evidence

Serum SCFA–based multiclass models achieved high classification performance (95% accuracy) with a reported multiclass ROC-AUC of 1.00.95% accuracy; multiclass ROC-AUC 1.00

“We developed an interpretable machine-learning framework combining stacking ensembles, bootstrap validation, permutation testing, SHAP and LIME analyses.”
Claim 2 of 5Not coveredThe story reports that the study was small, that substantial differences in age, body mass index, and other characteristics may have influenced the results, and that larger independently recruited cohorts are needed before serum SCFA profiling can be considered a clinical diagnostic test.View evidenceHide evidence

Why this verdict

Several caveats are supported at abstract depth: the cohort was modest with small disease subgroups, validation was internal rather than external, and further independent validation would be needed before clinical diagnostic use. But the specific assertion that substantial differences in age, BMI, and other characteristics influenced results is not verifiable from the abstract-level profile, which says detailed demographics and covariate adjustment are not provided.

1

Study evidence

Circulating (serum) SCFA profiles provided substantially stronger disease-associated signatures and classification of disease groups than fecal SCFA profiles.Serum model accuracy ≈95%; multiclass ROC-AUC reported as 1.00 (abstract).

“Seven fecal and serum SCFAs were quantified by gas chromatography-mass spectrometry in 193 individuals, including patients with colorectal cancer (CRC, n = 30), celiac disease (CD, n = 35), obesity (OB, n = 33), and healthy controls (HCs, n = 95).”
2

Study evidence

Serum SCFA–based multiclass models achieved high classification performance (95% accuracy) with a reported multiclass ROC-AUC of 1.00.95% accuracy; multiclass ROC-AUC 1.00

“We developed an interpretable machine-learning framework combining stacking ensembles, bootstrap validation, permutation testing, SHAP and LIME analyses.”
Claim 3 of 5SupportedResearchers compared microbial metabolites in blood and feces to investigate which measurements provide clearer clues about colorectal cancer, celiac disease, and obesity.View evidenceHide evidence

Why this verdict

The abstract-level profile supports that seven SCFAs were quantified in paired serum and fecal samples from participants with colorectal cancer, celiac disease, obesity, and healthy controls, with the analytic goal of comparing disease-associated signatures and classification performance across compartments. The story frames this as investigating which measurements provide clearer clues, which is appropriately associational and hedged.

1

Study evidence

Circulating (serum) SCFA profiles provided substantially stronger disease-associated signatures and classification of disease groups than fecal SCFA profiles.Serum model accuracy ≈95%; multiclass ROC-AUC reported as 1.00 (abstract).

“Seven fecal and serum SCFAs were quantified by gas chromatography-mass spectrometry in 193 individuals, including patients with colorectal cancer (CRC, n = 30), celiac disease (CD, n = 35), obesity (OB, n = 33), and healthy controls (HCs, n = 95).”
2

Study evidence

Serum SCFA–based multiclass models achieved high classification performance (95% accuracy) with a reported multiclass ROC-AUC of 1.00.95% accuracy; multiclass ROC-AUC 1.00

“We developed an interpretable machine-learning framework combining stacking ensembles, bootstrap validation, permutation testing, SHAP and LIME analyses.”
Claim 4 of 5SupportedIn a study published in QJM, researchers compared the diagnostic performance of serum and fecal short-chain fatty acid profiles for distinguishing colorectal cancer, celiac disease, obesity, and healthy controls.View evidenceHide evidence

Why this verdict

The profile supports the scientific substance of the claim: serum and fecal SCFA profiles were compared for multiclass discrimination among CRC, celiac disease, obesity, and healthy controls using machine-learning classification models. The supplied profile does not independently document the journal venue, but the diagnostic-performance comparison itself is supported at abstract depth.

1

Study evidence

Circulating (serum) SCFA profiles provided substantially stronger disease-associated signatures and classification of disease groups than fecal SCFA profiles.Serum model accuracy ≈95%; multiclass ROC-AUC reported as 1.00 (abstract).

“Seven fecal and serum SCFAs were quantified by gas chromatography-mass spectrometry in 193 individuals, including patients with colorectal cancer (CRC, n = 30), celiac disease (CD, n = 35), obesity (OB, n = 33), and healthy controls (HCs, n = 95).”
2

Study evidence

Serum SCFA–based multiclass models achieved high classification performance (95% accuracy) with a reported multiclass ROC-AUC of 1.00.95% accuracy; multiclass ROC-AUC 1.00

“We developed an interpretable machine-learning framework combining stacking ensembles, bootstrap validation, permutation testing, SHAP and LIME analyses.”
Claim 5 of 5SupportedThe article says serum SCFA profiles outperformed fecal SCFAs in discriminating healthy controls and patients with celiac disease, obesity, and colorectal cancer, suggesting serum SCFAs may better reflect disease-related changes than fecal SCFAs.View evidenceHide evidence

Why this verdict

The profile supports that serum/circulating SCFAs showed markedly stronger disease-associated signatures and better classification performance than fecal SCFAs across the evaluated groups. The story’s wording is hedged—'suggesting' serum SCFAs may better reflect disease-related changes—so it does not overstate the observational evidence as causal.

1

Study evidence

Circulating (serum) SCFA profiles provided substantially stronger disease-associated signatures and classification of disease groups than fecal SCFA profiles.Serum model accuracy ≈95%; multiclass ROC-AUC reported as 1.00 (abstract).

“Seven fecal and serum SCFAs were quantified by gas chromatography-mass spectrometry in 193 individuals, including patients with colorectal cancer (CRC, n = 30), celiac disease (CD, n = 35), obesity (OB, n = 33), and healthy controls (HCs, n = 95).”
2

Study evidence

Serum SCFA–based multiclass models achieved high classification performance (95% accuracy) with a reported multiclass ROC-AUC of 1.00.95% accuracy; multiclass ROC-AUC 1.00

“We developed an interpretable machine-learning framework combining stacking ensembles, bootstrap validation, permutation testing, SHAP and LIME analyses.”
Then look for missing context

Context layer

What the story left out

Important study details the story did not include.

  • Feature-attribution analyses identified valeric, propionic, 2-methylbutyric, and isovaleric acids as main contributors to disease discrimination.

    This is a reported key finding in the paper profile, but the story presentation does not mention the specific SCFAs identified by SHAP/LIME as major contributors.

    From Secondary-data observational diagnostic/prognostic modeling: stacking ensembles with internal validation and interpretab

  • Cross-compartment predictive modeling found weak predictive relationships between fecal and serum SCFA pools, suggesting they are largely non-redundant and may reflect distinct biological processes.

    The paper profile includes this as a secondary contribution, but the story presentation focuses on comparative disease-classification performance and does not report the reciprocal fecal-serum informativeness analysis.

    From cross-compartment predictive modeling (secondary_data)

7 things the story did carry across
  • The study measured seven SCFAs in paired serum and fecal samples from 193 human participants across CRC, celiac disease, obesity, and healthy-control groups using GC–MS.
  • Serum/circulating SCFA profiles showed stronger disease-associated signatures and better disease-group classification than fecal SCFA profiles.
  • Model performance was reported as approximately 95% accuracy for serum SCFAs, 59% for fecal SCFAs, and 97.4% for combined serum-fecal data, with only marginal gain from combining compartments.
  • Validation was internal only at the abstract level; no external or independently recruited validation cohort is reported.
  • The cohort was moderate in size overall and had modest, uneven disease subgroup sizes, which may affect model stability and generalizability.
  • The abstract-level profile does not provide participant-level demographic, medication, diet, fasting, sampling-timing, preprocessing, normalization, or covariate-adjustment details, limiting assessment of confounding and generalizability.
  • The study is observational and cross-sectional, so SCFA differences and model feature attributions should not be interpreted as causal effects or temporal predictors of disease development.
Then read the study layer

Study layer

Study at a glance

Scan the study first. Expand only the parts you want to inspect.

Pieces of work

3

Evidence read

study summary

Lead result

secondary data

1Lead resultsecondary dataBuild and internally validate interpretable machine-learning models for multiclass disease discrimination using serum vs fecal SCFAs (and combined), including feature attribution to identify key SCFAs driving predictions.Secondary-data observational diagnostic/prognostic modeling: stacking ensembles with internal validation and interpretable feature attributionExpand

In plain English

Using serum and fecal measurements of seven short-chain fatty acids (SCFAs) from 193 participants across colorectal cancer, celiac disease, obesity, and healthy controls, the authors developed an interpretable machine-learning framework (stacking ensembles with bootstrap validation, permutation testing, and SHAP/LIME interpretability) to build and internally validate multiclass disease classifiers. Serum SCFAs produced substantially stronger discrimination (95% accuracy; multiclass ROC-AUC 1.00) than fecal SCFAs (59% accuracy); combining serum and fecal data gave only marginal additional benefit (97.4% accuracy). Valeric, propionic, 2-methylbutyric, and isovaleric acids were repeatedly identified as the main contributors to classification. Cross-compartment modeling indicated weak predictive relationships between fecal and serum SCFAs, consistent with non-redundant biological information in the two pools.

Key findings

  • Serum SCFA–based multiclass models achieved high classification performance (95% accuracy) with a reported multiclass ROC-AUC of 1.00.95% accuracy; multiclass ROC-AUC 1.00
  • Fecal SCFA–based multiclass models showed substantially lower predictive power (59% accuracy) compared with serum models.59% accuracy
“We developed an interpretable machine-learning framework combining stacking ensembles, bootstrap validation, permutation testing, SHAP and LIME analyses.”
What this piece can’t prove
  • Total cohort size is 193 with subgroup sizes reported (CRC n=30, CD n=35, OB n=33, HC n=95); subgroup sample sizes are modest and uneven across classes as reported.
  • Abstract does not provide detailed modeling hyperparameters, exact ensemble composition, or full cross-validation procedures in the text provided.
  • Abstract does not report confidence intervals or uncertainty measures for the primary performance metrics beyond point estimates.

1 further detail could not be confirmed from the summary.

2human in vivoTest whether circulating (serum) short-chain fatty acids (SCFAs) provide stronger disease-associated signatures and better classification of gut-related diseases than fecal SCFAs.observational case-control/cohort-style comparison (paired biospecimens)Expand

In plain English

In a cross-sectional observational study of 193 human participants (CRC n=30, celiac disease n=35, obesity n=33, healthy controls n=95), seven short-chain fatty acids (SCFAs) were quantified in paired serum and fecal samples by gas chromatography–mass spectrometry (GC–MS). Interpretable machine-learning models showed that circulating (serum) SCFA profiles provided substantially stronger disease-associated signatures and classification of disease groups than fecal SCFAs; serum models achieved ~95% accuracy (multiclass ROC-AUC reported as 1.00) versus ~59% accuracy for fecal SCFAs. Combining serum and fecal data yielded only marginal improvement (~97.4% accuracy). Valeric, propionic, 2-methylbutyric, and isovaleric acids were repeatedly identified as key contributors to discrimination, and cross-compartment modeling indicated weak predictive relationships between fecal and serum SCFA pools.

Key findings

  • Circulating (serum) SCFA profiles provided substantially stronger disease-associated signatures and classification of disease groups than fecal SCFA profiles.Serum model accuracy ≈95%; multiclass ROC-AUC reported as 1.00 (abstract).
  • Fecal SCFA–based models had substantially lower predictive performance compared with serum-based models.Fecal model accuracy ≈59% (abstract).
“Seven fecal and serum SCFAs were quantified by gas chromatography-mass spectrometry in 193 individuals, including patients with colorectal cancer (CRC, n = 30), celiac disease (CD, n = 35), obesity (OB, n = 33), and healthy controls (HCs, n = 95).”
What this piece can’t prove
  • Moderate overall sample size with unequal group sizes (controls larger than individual disease groups), which may influence model stability and generalizability.
  • Cross-sectional observational design precludes assessment of temporal or causal relationships between SCFA levels and disease status.

2 further details could not be confirmed from the summary.

3secondary dataAssess cross-compartment ‘reciprocal informativeness’ between fecal and serum SCFAs via cross-compartment predictive modeling to evaluate redundancy vs non-redundancy of the two pools.cross-compartment predictive modeling (secondary data)Expand

In plain English

The study performed cross-compartment predictive modeling to test reciprocal informativeness between fecal and serum SCFA profiles measured by GC–MS in 193 individuals (CRC, celiac disease, obesity, healthy controls). Cross-compartment modeling reported weak predictive relationships between fecal and serum SCFAs, interpreted as these pools being largely non-redundant and reflecting distinct biological processes. The abstract does not provide detailed modeling specifications, directionality of prediction, or quantitative cross-prediction metrics for these analyses.

Key findings

  • Cross-compartment modeling revealed weak predictive relationships between fecal and serum SCFAs, indicating these pools are largely non-redundant and reflect distinct biological processes.weak predictive relationship
“Cross-compartment predictive modeling was used to assess the reciprocal informativeness of fecal and circulating SCFAs.”
What this piece can’t prove
  • Abstract lacks methodological detail for the cross-compartment predictive modeling (algorithms, cross-validation, performance metrics).
  • Abstract does not specify whether fecal→serum, serum→fecal, or both directions were evaluated, nor does it report numeric cross-prediction results.

1 further detail could not be confirmed from the summary.

Finally, the search trail

Method layer

NewsLink found the paper. Tessa takes you deeper.

NewsLink checks the story. Tessa is where you inspect the paper, authors, evidence, and research context.

Papers considered

The selected paper, plus nearby candidates.

PubMed, Crossref, Europe PMC · 15 candidate papers

And 9 more candidates considered.