The first widely noticed outbreak is not necessarily the first outbreak. A new analysis uses viral genetic sequences to reconstruct EV-D68 history, then asks whether its relationship with a rare neurological condition remains stable.
- Genetic sequences reflect both circulation and the effort spent looking for it.
- An inferred epidemic trajectory is not a register of counted infections.
- A temporal association can weaken even when the virus continues circulating.
- Paper date
- 23 October 2026
- Regions
- Europe and North America
- Question
- How robust is an association across time?
Original publication: 23 Oct 2026 · The date above refers to this brief.
A surveillance claim has three layers
Observe
Document which sequences and reported cases were available.
Infer
Keep model assumptions and uncertainty beside the reconstructed history.
Compare
Test whether the association survives a later period and its exceptions.
A longer history than the first headline
Jorgensen and colleagues use publicly available EV-D68 sequences and a phylodynamic model, which relates viral family trees to transmission and sampling. Their reconstruction supports seasonal transmission before the widely recognised 2014 outbreak. The study then compares inferred viral activity with reported US acute flaccid myelitis, or AFM, cases.
Source 1 ↗The analysis uses two grouped lineages
The main model groups related clades as A+D and B+C. This is an analytical grouping, not a claim that each pair is biologically identical. A separate sensitivity analysis considers clade B alone. The distinction matters when a headline shortens a result about a grouped lineage into a statement about one clade.
Source 1 ↗The changing association is part of the finding
The inferred B+C trajectory is strongly associated with AFM case timing during earlier years of the surveillance series. In two-year sliding windows, that relationship weakens around the COVID-19 period and is not strongly restored in the 2021–2022 outbreaks. The paper also reports a consistent association when B is analysed alone. This does not establish why the relationship changed.
Source 1 ↗Three counts that should not be collapsed
Our reading sheet separates deposited sequences, infections inferred by a model and reported neurological cases. Sequencing can increase after an outbreak attracts attention. A model attempts to account for sampling, but its infection-scale trajectory remains sensitive to assumptions. The authors explicitly caution against treating the reconstructed population numbers as true infection counts.
The data overview contains 3,955 quality-filtered sequences overall; the Methods identify 3,216 from the two main regions before further exclusions and subsampling. Those are sequence counts, not numbers of AFM patients. The figure’s sampling series runs through the end of 2022.
Source 1 ↗What would make a future comparison more persuasive?
Our proposed check uses a fixed future observation window, a documented surveillance denominator and prespecified criteria for a successful forecast. It would evaluate both periods in which viral activity and neurological cases rise together and periods in which they diverge. Reproducing only the earlier association would leave the most informative exception unexplained.
A correlation at zero monthly lag does not prove that every infection causes an immediate neurological event. Aggregated time series cannot determine an individual’s risk or diagnose a child. This article explains surveillance inference; it does not offer symptom triage or clinical advice.
Connect this to other predictive claims
The companion genetic-score guide asks whether a model transfers to another setting. This EV-D68 analysis asks a parallel question across time: does a previously observed relationship persist after surveillance and epidemic context change? Both benefit from recording the denominator and testing the exception rather than treating past fit as guaranteed future performance.
A strong historical fit is not an unconditional forecast
Selected accepted PDF: data overview/Fig. 1 p.2, sampling and trajectory Results pp.5–6, Fig. 5 caption p.6, model limitations and sequence Methods pp.8–9. Fig. 5 and caption visually inspected. Publisher HTML metadata and abstract also checked. Supplements, source-data files, code and clinical records were not independently audited. No publisher figure reproduced. The accepted version may receive further edits.
01Is the trajectory an infection census?
- What was observed
- The model reconstructs population trajectories from genetic data and assumptions.
- Where the conclusion stops
- The authors warn against reading the scale as true infection counts.
Source 1 · PDF Discussion p.8; trajectory sensitivity discussion
02Does the AFM association stay stable?
- What was observed
- Sliding-window analysis shows an earlier B+C association that weakens around the COVID-19 period.
- Where the conclusion stops
- The analysis does not establish why it changed or predict an individual outcome.
Source 1 · PDF p.6; Association with AFM cases and Fig. 5
Numbers you can inspect
| Measure | Value & unit | Origin & method |
|---|---|---|
| Initial quality-filtered sequences | 3955 sequences | Reported Overall data overview before geographic/model restrictions. Source 1 · PDF p.2; Fig. 1 caption |
| Sequences in main regions | 3216 sequences | Reported Europe and North America before additional date exclusions/subsampling. Source 1 · PDF p.9; sequence Methods |
| Transmission-model time intervals | 6 months | Reported Piecewise modelling interval, not a clinical follow-up period. Source 1 · PDF p.4 Fig. 3 caption; Methods |
| Sliding association window | 2 years | Reported Time-series comparison window, not an individual risk horizon. Source 1 · PDF p.6; Fig. 5 caption |
Compare the actual experiments
These studies answer different questions. Read the unit and endpoint before comparing results.
| Study | Unit & setting | Readout | Interpretation boundary |
|---|---|---|---|
| Phylodynamic reconstruction Source 1 · PDF pp.2–5; Results | Public viral sequences | Timing and relative epidemic patterns | Sampling effort and modelling choices affect inference. |
| AFM time-series comparison Source 1 · PDF p.6; Fig. 5 | Inferred trajectories and reported US cases | Association across lag and sliding windows | Ecological comparison does not identify individual causation. |
The export includes claims, available numbers, methods and source locations. It contains our reading notes and published summaries; it is not raw participant data or an independent reanalysis.
Evidence update · 23 Oct 2026
First publication. Observations, our arithmetic and proposed follow-up questions are separated. No raw-data reanalysis. AI source check; no human editorial or clinical review.
Three records, three different meanings
| Reading question | Evidence to retain | What remains open |
|---|---|---|
| What was deposited? | Sequence, date and geographic sampling coverage | How many infections were missed |
| What was reconstructed? | Modelled trajectory and uncertainty | True absolute infection counts |
| What was reported? | US AFM surveillance series and its coverage | Comparable neurological ascertainment elsewhere |
| Does the link persist? | Earlier and later periods, including mismatches | The biological or surveillance explanation for a change |
Original Press-News assessment framework. Follow-up questions are our interpretation, not additional experiments performed by the authors.
Your questions, answered
Did EV-D68 first appear in 2014?
No. The reconstruction supports earlier seasonal transmission; 2014 marks the first widely recognised outbreak discussed here.
Were 3,955 AFM patients analysed?
No. That number describes quality-filtered viral sequences in the data overview.
Does clade B always predict an AFM wave?
The study reports a changing temporal relationship, including a weaker association after the COVID-19 interruption.
Can the model count every infection?
No. The authors caution that reconstructed population scale is sensitive to assumptions and should not be read as true infection counts.
Limits of this interpretation
- Genomic surveillance is uneven across time, locations and viral groups.
- The model groups related clades and relies on assumptions about sampling and movement.
- Reconstructed population scale is not an independently counted infection burden.
- Population-level association does not determine individual risk or the reason for its change.
Sources & transparency
- Jorgensen, Lefrancq, Grassly et al. (2026): Clade-specific transmission dynamics of enterovirus D68 and their association with acute flaccid myelitis
Selected accepted PDF: data overview/Fig. 1 p.2, sampling and trajectory Results pp.5–6, Fig. 5 caption p.6, model limitations and sequence Methods pp.8–9. Fig. 5 and caption visually inspected. Publisher HTML metadata and abstract also checked. Supplements, source-data files, code and clinical records were not independently audited. No publisher figure reproduced. The accepted version may receive further edits. · Accessed 23 Oct 2026
DOI: 10.1038/s41467-026-78448-2
Prepared and source-checked with AI. Press-news Team is the collective publication byline, not a medical reviewer. No human editorial or clinical review has taken place. This educational article examines research methods, model evaluation and laboratory or observational findings, not individual diagnosis or treatment. We did not conduct these experiments or reanalyse raw data. Reported results, our interpretation and proposed future tests are separated. Source-access limits are recorded below. Photographs are illustrative.
Source check: AI source check — selected primary PDF results, captions and study boundaries
Clinical review: Not applicable to this educational guide
Suggest a correction