CGM accuracy and reliability compared to point of care testing in older inpatients with comorbid type 2 diabetes and cognitive impairment

This study demonstrates that continuous glucose monitoring (CGM) is a safe, clinically accurate, and viable alternative to point-of-care testing for managing blood glucose in older hospitalized patients with type 2 diabetes and cognitive impairment, offering comparable metrics while better capturing nocturnal hypoglycemia.

The Gist

The Big Picture: A Safety Net for a Vulnerable Group

Imagine you are managing a busy kitchen. Usually, you check the temperature of the oven every hour with a thermometer (this is like the standard Point-of-Care Test or POCT that nurses do with a finger prick). But what if the chef is tired, confused, or has trouble remembering to check the oven? They might miss a moment when the oven gets too hot or too cold, leading to a burnt meal or a raw one.

This study looked at a specific group of "chefs": older hospital patients who have Type 2 Diabetes and cognitive impairment (like dementia or memory loss). Because of their memory issues, they might forget to report how they feel, or nurses might miss a quick drop in blood sugar between their scheduled checks.

The researchers wanted to see if a new tool, a Continuous Glucose Monitor (CGM), could act like a smart, 24/7 security camera for their blood sugar, catching problems that the "hourly thermometer" misses.

---

The Experiment: The "Blind" Test

The researchers put a small sensor (the Dexcom G7) on the upper arm of 30 older patients in a UK hospital.

• The Catch: The sensors were "blinded." This means neither the patients nor the doctors could see the numbers on the screen. The doctors kept doing their usual finger-prick tests (POCT) to manage care, while the sensor quietly recorded data in the background.
• Why do this? To see if the sensor's data matched the finger-prick data without anyone changing their behavior based on what the sensor showed. It was a pure comparison of accuracy.

The Results: Two Ways of Looking at the Same Thing

1. The "Big Picture" Match

When the researchers looked at the overall average blood sugar levels over the whole hospital stay, the CGM (the camera) and the POCT (the hourly thermometer) agreed very well.

• The Analogy: If you take a photo of a landscape every hour, and a drone takes a photo every 5 minutes, the overall view of the landscape looks the same. Both methods showed that these patients spent a lot of time with high blood sugar and a little time with low blood sugar.

2. The "Hidden Danger" Discovery

This is where the study got exciting. The CGM found many more low blood sugar events (hypoglycemia) than the finger-prick tests did.

• The Stat: The CGM found 63% more low blood sugar episodes than the nurses' standard checks.
• The Analogy: Imagine a security guard who checks the front door every 4 hours. A burglar (low blood sugar) sneaks in at 2:00 AM, steals something, and leaves by 2:15 AM. The guard checks at 4:00 AM and sees nothing. The CGM is like a motion-sensor camera that recorded the burglar the whole time.
• Why it matters: Many of these low-sugar events happened at night or were very brief. Because the patients had cognitive issues, they couldn't tell the nurses they felt shaky or confused. The CGM saw it; the standard checks missed it.

3. Is the Camera Accurate?

The researchers asked: "Is the CGM reading the numbers correctly?"

• The Verdict: Yes, it was clinically safe and accurate.
• The Details: The numbers from the CGM were very close to the finger-prick numbers (99% of the time, the readings were in a "safe zone" where a doctor would make the right decision).
• The "Bias": The CGM tended to read slightly higher than the finger prick (by about 0.9 mmol/L). Think of it like a slightly optimistic weather forecast that says "sunny" when it's actually "partly cloudy." It's not dangerous; it just means the CGM is a bit cautious.

The Takeaway: Why This Matters

1. The "Safety Net" Effect
For older patients with memory loss, relying on them to say "I feel low" or waiting for a nurse to check every 4 hours is risky. The CGM acts as a constant safety net, catching dips in blood sugar that happen in the dark or between checks.

2. It's Not a Replacement (Yet)
The study concludes that while the CGM is amazing at detecting problems, we still need the finger-prick test to confirm them before giving medicine.

• The Analogy: The CGM is the smoke alarm. It screams "Fire!" immediately. But before you call the fire department (give insulin or sugar), you still need to look and see if there's actually a fire (do a finger prick).

Summary in One Sentence

This study proves that for older, confused hospital patients with diabetes, wearing a continuous glucose monitor is like upgrading from a periodic security check to a 24/7 surveillance system, catching dangerous drops in blood sugar that would otherwise go unnoticed and untreated.

Technical Summary

1. Problem Statement

Older inpatients with comorbid Type 2 Diabetes Mellitus (T2DM) and cognitive impairment represent a high-risk, vulnerable population. Cognitive decline often leads to missed glucose measurements, failure to report symptoms, and an inability to manage complex diabetes regimens.

• Clinical Gap: While Continuous Glucose Monitoring (CGM) is known to reduce hypoglycemia in community settings, there is a lack of evidence regarding its analytical and clinical accuracy specifically within the acute hospital setting for patients with both T2DM and cognitive decline.
• Current Standard: Point-of-Care Testing (POCT) via finger-prick blood glucose is the standard of care but is intermittent (typically 4 times daily), potentially missing asymptomatic or nocturnal hypoglycemia, which is particularly dangerous in cognitively impaired patients.
• Objective: To evaluate whether blinded CGM (Dexcom G7) provides glucose metrics comparable to POCT and to assess its ability to detect hypoglycemic events that POCT might miss in this specific demographic.

2. Methodology

Study Design:

• Type: Single-centre, prospective, observational, blinded feasibility study.
• Setting: Norfolk and Norwich University Hospital, UK (Non-ICU wards).
• Timeline: Data collection occurred between October 2024 and July 2025.

Participants:

• Sample Size: 32 recruited; 30 included in final analysis (2 excluded: 1 for >50% missing data, 1 for lack of antidiabetic medication).
• Inclusion Criteria: Age ≥65, T2DM duration ≥6 months, on medication (not diet-only), and cognitive impairment defined by:
  • Abbreviated Mini-Mental Test (AMT) score ≤8/10, OR
  • Mini-Addenbrooke's Cognitive Examination (Mini-ACE) score ≤21/30.
• Exclusion Criteria: End-of-life care, lack of mental capacity, current CGM users, dialysis, skin irritation at sensor site, or other major neurological/psychiatric conditions (e.g., autism, learning difficulties).

Intervention & Data Collection:

• Device: Blinded Dexcom G7 real-time CGM (rt-CGM) sensors applied to the upper arm.
• Duration: Up to 10 days during hospital stay.
• Blinding: Both participants and clinical staff were blinded to CGM readings to ensure "usual care" was maintained.
• Reference Standard: Standard of care POCT (minimum 4 times daily).
• Data Matching: POCT readings were paired with CGM readings taken within ±2 minutes. Data was filtered to the range of 2.2–22.2 mmol/L to avoid ceiling/floor effects.

Statistical Analysis:

• Metrics: Mean Absolute Relative Difference (MARD), Median Absolute Relative Difference (Median ARD), Clarke Error Grid (CEG), Correlation ($R^2$), and Bland-Altman analysis.
• Accuracy Criteria: Proportion of readings within ±20% (for glucose >5.6 mmol/L) or ±1.1 mmol/L (for glucose ≤5.6 mmol/L).
• Software: RStudio (v4.3.2).

3. Key Contributions

• First-of-its-Kind: This is the first study to specifically investigate the accuracy of next-generation CGM (Dexcom G7) in a non-ICU hospital setting for patients with comorbid T2DM and cognitive impairment.
• Blinded Feasibility: By using blinded sensors, the study eliminated the "observer effect," providing an objective comparison between CGM data and routine clinical POCT without influencing clinical decision-making during the study period.
• Hypoglycemia Detection: The study quantifies the specific gap in hypoglycemia detection between intermittent POCT and continuous monitoring in a population unable to self-report symptoms.

4. Results

Participant Demographics:

• Mean age: 78.7 years.
• Mean HbA1c: 75.1 mmol/mol.
• 80% had ≥4 long-term conditions.
• 30% admitted specifically for diabetes-related complications (DKA, HHS, hypoglycemia).

Glucose Metrics (CGM vs. POCT):

• Time in Range (TIR 4–10 mmol/L): CGM = 36.2% vs. POCT = 41.0% (No significant group-level difference, $p=0.97$).
• Time Above Range (TAR ≥10 mmol/L): CGM = 62.8% vs. POCT = 57.2%.
• Time Below Range (TBR ≤3.9 mmol/L): CGM = 1.0% vs. POCT = 1.8%.
• Note: While group averages were similar, individual discrepancies were noted where POCT missed TBR events detected by CGM.

Analytical Accuracy:

• Correlation: Strong correlation between CGM and POCT ($R^2 = 0.82$).
• MARD: 17.43% (Median ARD: 12.24%).
• Agreement: 72.32% of paired readings fell within the ISO 15197 accuracy criteria (±20%/±1.1 mmol/L).
• Bias: Systematic mean bias of +0.88 mmol/L (CGM tended to overestimate slightly).
• Limits of Agreement (Bland-Altman): +5.09 mmol/L to -3.33 mmol/L.

Clinical Accuracy (Clarke Error Grid):

• Zone A (Clinically Accurate): 83.1%
• Zone B (Benign Errors): 16.2%
• Total Clinically Acceptable (A+B): 99.3% of readings.
• Interpretation: The vast majority of readings would lead to correct or benign treatment decisions.

Hypoglycemia Detection:

• Incidence: 30% of participants (9/30) experienced CGM-detected hypoglycemia (≤3.9 mmol/L).
• Missed Events: Of 27 hypoglycemic events detected by CGM, 17 (62.9%) were missed entirely by POCT.
• Delay: For events where POCT eventually detected the low, there was a significant delay (34 minutes to 4 hours) compared to the initial CGM reading.
• Severity: Most events were Level 2 (<3.0 mmol/L), requiring therapeutic intervention.

5. Significance and Conclusion

Clinical Implications:

• Safety: CGM (Dexcom G7) demonstrates clinically acceptable accuracy (99.3% in Zones A/B) for older inpatients with cognitive impairment, supporting its use as a monitoring tool in this setting.
• Superior Hypoglycemia Detection: The study confirms that intermittent POCT fails to detect the majority (63%) of hypoglycemic episodes in this population, particularly nocturnal or brief events. This is critical as asymptomatic hypoglycemia is a major cause of morbidity and mortality in cognitively impaired patients.
• Workflow Integration: CGM should be viewed as a complement to, not a total replacement for, POCT. The authors recommend using CGM for continuous surveillance and POCT for verifying abnormal readings before clinical intervention.

Limitations:

• Sample Size: Small cohort (n=30) from a single center.
• Observational Nature: Lack of a control group or randomized intervention arm.
• Reference Standard: Used POCT rather than laboratory venous glucose for accuracy validation, though this reflects real-world hospital practice.

Conclusion:
The study concludes that CGM is a safe, reliable, and superior tool for detecting dysglycemia, specifically hypoglycemia, in older hospitalized patients with T2DM and cognitive impairment. Its adoption could significantly reduce the risk of unrecognised hypoglycemia and improve glycemic management in this vulnerable demographic.