# Turning device telemetry into real-world evidence
A single continuous glucose monitor (CGM) generates a reading every five minutes. That is roughly 288 data points per patient per day. Multiply that across millions of users, and one device category produces more clinical data in a week than most drug trials collect in years.
That firehose of data is not just a clinical convenience. It has become a regulatory asset. In this lesson we trace how device telemetry (the automatic, continuous stream of measurements a device sends back to manufacturers and clinicians) turns into real-world evidence (RWE) that regulators like the FDA accept to support new claims.
Telemetry is raw. Evidence is curated. The gap between them is where the work happens.
Here is the journey a CGM reading takes:
1. Sensor measures interstitial glucose and transmits to a phone app.
3. Cloud platform stores, cleans, and aggregates readings across the user base.
4. Analytics layer detects patterns: sensor failures, out-of-range readings, adverse events.
5. Regulatory and safety teams use the aggregated output for post-market surveillance and, sometimes, new regulatory claims.
Two very different uses sit at the end of that pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.Voir la définition complète →. Let us define both.
Once a device is approved and sold, manufacturers are legally required to keep watching how it performs in the real world. This is post-market surveillance. In the US, serious problems flow into the FDA's MAUDE database (Manufacturer and User Facility Device Experience), a public record of adverse event reports.
Telemetry supercharges PMS. Instead of waiting for a clinician to mail in a paper report, a pacemaker can flag a lead malfunction automatically.
RWE is clinical evidence about a product derived from real-world datareal-world dataRWD, données collectées en dehors des essais cliniques contrôlés : dossiers médicaux, claims d'assurance, données de dispositifs connectés, base des Real-World Evidence (RWE). (data collected outside traditional randomized trials). Under the 21st Century Cures Act, the FDA has a formal framework for using RWE to support regulatory decisions, including new indications and label changes.
The key shift: telemetry that was once only a safety monitoring tool can now, if collected rigorously, help expand what a device is officially allowed to claim.
Modern CGMs from manufacturers like Dexcom and Abbott stream glucose data continuously. That stream feeds three data products at once.
Patient-facing. The app shows trends and alerts for highs and lows.
Clinician-facing. Aggregated reports summarize time in range (the percentage of the day a patient's glucose stays within target), a metric now widely used in diabetes care.
Manufacturer-facing. De-identified, pooled data across the user base.
That third stream is the RWE engine. A manufacturer can analyze how the device performs across age groups, skin types, and real-world wear conditions that a controlled trial never captures.
A concrete example of RWE impact: CGM makers have used real-world and trial data to expand labeling toward broader populations and, in some cases, to support claims that let patients dose insulin directly from the sensor reading without a confirmatory finger-stick. Each expansion requires evidence that the device is accurate enough at scale. Telemetry provides the volume.
Raw telemetry is messy. Before it counts as evidence, teams must handle:
A quick illustration of computing time in range from a raw glucose stream:
import pandas as pd
# readings: timestamped glucose values in mg/dL
readings = pd.read_csv("cgm_stream.csv", parse_dates=["timestamp"])
target_low, target_high = 70, 180
in_range = readings["glucose"].between(target_low, target_high)
time_in_range_pct = 100 * in_range.mean()
print(f"Time in range: {time_in_range_pct:.1f}%")Simple to write. The hard part is trusting that cgm_stream.csv is complete, correctly attributed, and representative.
A pacemaker's telemetry is lower volume but higher stakes. Modern cardiac devices transmit data through remote monitoring systems, often overnight while the patient sleeps.
What flows back:
Consider one lead fracture alert. On its own, it is a PMS event: the manufacturer investigates, may notify the clinician, and if a pattern emerges, may issue a recall.
Now aggregate across an entire implanted population. If telemetry shows that a specific lead model fails at a higher-than-expected rate after a certain number of years, that signal can trigger:
This is exactly the kind of long-horizon safety question that randomized trials cannot answer. Trials end. Implanted devices keep transmitting for a decade or more. Telemetry is the only practical way to watch a device age inside real patients.
🎬 [VIDEO: "What is Real-World Evidence?" — youtube.com — a short, plain-language explainer on how RWE differs from clinical trial data and where regulators accept it]
Not all data is created equal. Regulators distinguish between data you happen to have and data you can defend.
The FDA's RWE framework emphasizes two things:
Relevance. Does the data actually capture the outcome you care about? A CGM measures glucose well. It does not directly measure heart attacks, so it cannot alone support a cardiovascular claim.
Reliability. Was the data collected and processed with quality controls? This means documented data curation, validated algorithms, and an audit trail.
For device makers, this reframes data infrastructure as a regulatory function, not just an IT function. The team that designs your cloud schemaschemaA schema is the formal blueprint that defines how data is structured, named, typed, and related within a database, file, or message.Voir la définition complète → is, indirectly, shaping what claims you can make later.
A device company that treats telemetry as exhaust (a byproduct to be discarded) leaves value on the table. A company that treats telemetry as a designed evidence pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.Voir la définition complète → can:
The business case is not abstract. RWE-backed claims can shorten the path to new indications, and real-world outcome data increasingly matters to payers deciding what to reimburse.
Vérification des acquis
1. The lesson describes the relationship between telemetry and real-world evidence by saying 'Telemetry is raw. Evidence is curated.' What is the primary conceptual implication of this distinction?
2. Why does the lesson emphasize that one CGM produces roughly 288 readings per patient per day across millions of users?
3. How does continuous telemetry fundamentally change post-market surveillance (PMS) compared to traditional methods?
4. Select ALL correct answers. Based on the telemetry-to-evidence pipeline described, which statements are accurate?
Sélectionnez toutes les réponses correctes.
5. Select ALL correct answers about the distinction between post-market surveillance (PMS) and real-world evidence (RWE) as presented in the lesson.
Sélectionnez toutes les réponses correctes.
RWE is powerful, not magic. A few honest constraints:
Confounding. In real-world datareal-world dataRWD, données collectées en dehors des essais cliniques contrôlés : dossiers médicaux, claims d'assurance, données de dispositifs connectés, base des Real-World Evidence (RWE)., patients are not randomized. If sicker patients get a particular device, raw outcomes look worse even if the device is fine. Analysts use statistical methods to adjust, but adjustment is never perfect.
Selection effects. Telemetry only covers patients whose devices connect and who stay engaged. The patients you fail to observe may be exactly the ones at highest risk.
Interoperability. A hospital may use devices from several manufacturers, each with its own dataown dataData collected directly from your own customers and prospects through your own channels: your most reliable and privacy-compliant source.Voir la définition complète → format. Combining them into one evidence base is a genuine engineering challenge, and standards like HL7 FHIR aim to help but are unevenly adopted.
Privacy and consent. Using patient telemetry for evidence beyond direct care raises consent questions. What patients agreed to at implant may not clearly cover every downstream analysis.
None of these kill the value. They define the guardrails. The credible device companies are the ones that acknowledge these limits in their submissions rather than hiding them.