What a High Glucose Number Means

A single glucose value is high in the diagnostic sense in three cases. A venous plasma draw reads 126 mg/dL or more after an eight-hour fast. It reads 200 mg/dL or more two hours into a 75 g oral glucose tolerance test, or 200 mg/dL or more at a random moment in someone with classic symptoms such as thirst, frequent urination and unexplained weight loss. Between those and normal sits an intermediate band: 100–125 mg/dL fasting, 140–199 mg/dL at two hours. Those numbers come from lab chemistry on venous plasma, and they are not the same thing as the number your continuous glucose monitor (CGM) shows on your phone. If you are reading 200 on a sensor after lunch, you have one data point from an instrument with roughly 9–14% error and a five-to-ten-minute lag. The correct response is to collect more data rather than draw a conclusion. If you are reading 200 repeatedly on a fingerstick meter, or you feel unwell, that is a conversation with a clinician rather than a data project.

Where the thresholds come from

The diagnostic cutoffs were chosen because they mark where the risk of diabetic retinopathy begins to rise steeply in population studies. Physiology does not change state at those points. The cutoffs are defined on a specific specimen, which is plasma separated from venous blood, ideally drawn into a tube containing a glycolysis inhibitor. Whole blood glucose runs about 10–12% lower than plasma, which is why meters and sensors report plasma-equivalent values after an internal correction. Capillary blood during a rising postprandial excursion can read meaningfully higher than a simultaneous venous draw, because the tissue is extracting glucose between the arterial and venous ends of the bed.

The hemoglobin A1c thresholds (5.7% and 6.5%) describe the same construct through a different window: the fraction of hemoglobin that has been non-enzymatically glycated over the red cell lifespan, roughly 90–120 days. A1c is weighted toward the most recent month and is distorted by anything that changes red cell turnover, including iron deficiency, hemolysis, recent transfusion, and several hemoglobin variants. Two people with identical mean glucose can differ by half a point of A1c on red cell biology alone.

A CGM number is not a lab number

CGMs measure glucose in interstitial fluid using a subcutaneous electrode, typically glucose oxidase generating a current proportional to concentration, then convert that current to a displayed value through a calibration model ¹². Three consequences follow, and all three change how you should read a high number.

First, interstitial glucose lags plasma glucose by roughly five to ten minutes because glucose has to diffuse out of the capillary. During a steep rise after a meal, the sensor understates the true peak and reports it late. During a rapid fall it overstates. Rate of change and error are correlated, so the readings you most want to trust (the sharp peaks) are the ones with the widest error bars ³.

Second, accuracy is described by mean absolute relative difference (MARD), the average percentage gap between sensor readings and paired reference values. Modern systems with improved calibration algorithms have reported overall MARD near 9% against laboratory reference in clinical testing ⁴, with older or less refined systems landing higher. A 9% MARD at a displayed 200 mg/dL means a typical error of about 18 mg/dL, and the distribution has a tail: individual readings can be off by 30% or more, especially on day one of a sensor or during compression.

Third, the sensor has failure modes that produce numbers unconnected to your physiology. Lying on the sensor compresses the tissue and produces a false low, sometimes a dramatic one, that recovers the instant you roll over. Acetaminophen interferes with glucose oxidase chemistry in some sensor designs and inflates readings. The first twelve to twenty-four hours after insertion are noisier as the wound settles. These artifacts have a signature in the raw trace: they are abrupt, not physiologic in shape, and they lack a matching context.

The practical rule we use: never act on a single CGM value. Confirm with a fingerstick if the number matters, and treat the trend and the aggregate statistics as the real signal ⁵.

What to compute from your own export

Both major systems export CSV. Dexcom Clarity gives you Timestamp (YYYY-MM-DDThh:mm:ss), Event Type, and Glucose Value (mg/dL). The cadence is five minutes, or 288 readings per complete day. Values outside the reportable range appear as the strings Low and High rather than numbers, which will silently break a naive astype(float). FreeStyle Libre exports Device Timestamp, Record Type, Historic Glucose mg/dL, and Scan Glucose mg/dL. Record type 0 is the fifteen-minute historic series (five-minute on Libre 3), type 1 is a manual scan, type 6 is a note. Mixing scan and historic rows double-counts moments and biases your mean.

A minimal pass that gets you the numbers that matter:

import pandas as pd

df = pd.read_csv("clarity_export.csv", skiprows=range(1, 11))
df = df[df["Event Type"] == "EGV"].copy()
df["ts"] = pd.to_datetime(df["Timestamp (YYYY-MM-DDThh:mm:ss)"])
df["g"] = pd.to_numeric(df["Glucose Value (mg/dL)"], errors="coerce")
df.loc[df["Glucose Value (mg/dL)"] == "Low", "g"] = 40
df.loc[df["Glucose Value (mg/dL)"] == "High", "g"] = 400
df = df.dropna(subset=["g"]).set_index("ts").sort_index()

# require >=70% sensor coverage per day before trusting daily stats
cov = df["g"].resample("D").count() / 288
days = cov[cov >= 0.7].index

mean_g = df["g"].mean()
cv = df["g"].std() / mean_g * 100
tir = ((df["g"] >= 70) & (df["g"] <= 180)).mean() * 100
gmi = 3.31 + 0.02392 * mean_g            # Bergenstal et al. 2018

The glucose management indicator (GMI) is the linear conversion from CGM mean glucose to an estimated A1c-equivalent, and it exists precisely because measured A1c and CGM mean often disagree in the same person ⁶. When your GMI and your lab A1c diverge by more than about 0.3 percentage points, the interesting question is which one is wrong and why, and that usually points at red cell biology rather than at your glucose.

Coefficient of variation (CV) is the variability metric worth watching. A CV above roughly 36% indicates a glycemic profile dominated by large swings rather than a shifted baseline, and it is a different phenotype from a uniformly elevated curve with a flat shape. Fourteen days of data with at least 70% sensor coverage is the usual minimum before period-level statistics such as mean, time in range, and GMI are stable enough to compare across time ⁷.

Beyond the standard metrics, the analysis that repays the effort is per-meal. Align every logged meal to t=0 and compute the two-hour incremental area under the curve and the peak delta from the fifteen-minute pre-meal baseline. Then look at the distribution across repeats of the same meal. Repeatability is poor for a single instance and good across five, which is the entire argument for wearing a sensor rather than doing one OGTT. Sleep, prior exercise and time of day all move the same meal by a wide margin, as does the sequence of macronutrients within the meal.

What makes a number high that has nothing to do with metabolism

Before interpreting a high value, rule out the trivial explanations. Sensor compression and first-day noise, discussed above. Acute illness and fever, which raise counter-regulatory hormones. A hand contaminated with fruit sugar before a fingerstick, which can produce absurd readings. Glucocorticoids, which raise glucose substantially and dose-dependently. Sleep deprivation and acute psychological stress, both of which shift the whole curve upward for a day or two. Intense resistance or interval exercise, which transiently raises glucose through catecholamine-driven hepatic output, a rise that is normal and not a warning sign.

When this stops being a data question

We are describing measurement and interpretation. We are not describing care. Sustained readings above 250 mg/dL need urgent medical attention rather than further analysis. That is especially so with nausea, vomiting, deep rapid breathing, abdominal pain, or confusion. A confirmed fasting value at or above 126 mg/dL, or a two-hour value at or above 200 mg/dL, is a diagnostic finding that belongs with a physician who can order the confirmatory testing and rule out the many things that mimic it. Nothing you compute from a CSV replaces that, and no article should tell you what to do about a number. Bring the analysis to the appointment: a fourteen-day trace with meal annotations is far more useful to a clinician than the memory of one alarming reading.

Questions people also ask

Is 200 too high for your blood sugar? A fasting or random venous plasma value of 200 mg/dL is above the diagnostic threshold and needs clinical evaluation. A CGM showing 200 an hour after a large meal, once, tells you much less given 9% instrument error and sensor lag. It warrants confirming with a fingerstick and looking at the pattern over two weeks.

Should I go to the ER if my blood sugar is over 200? That is a clinical judgment, and 200 alone is not the deciding factor. Symptoms are: vomiting, abdominal pain, rapid breathing, drowsiness, or confusion alongside high readings are reasons to seek emergency care immediately rather than continue monitoring.

At what level does glucose start to do damage? There is no clean threshold. The diagnostic cutoffs were set where retinopathy risk begins to rise steeply in populations, but risk is continuous across the range and depends on exposure over years, which is why mean glucose, time in range, and A1c matter more than any one peak ⁶.

What is a normal blood glucose for a non-diabetic? Fasting venous plasma below 100 mg/dL and a two-hour post-load value below 140 mg/dL. On a sensor, people without diabetes spend the large majority of the day between about 70 and 140 mg/dL, with brief post-meal excursions above that which the sensor’s lag tends to blunt ⁵.

What brings a high glucose number down quickly? That is a treatment question, and it depends entirely on why the number is high and on what medications you take. Ask a clinician. From a measurement standpoint, the useful move is to keep the sensor running and capture the full shape of the excursion and its recovery, since that trace is the evidence anyone treating you will want ⁸.

Footnotes

1. Sandeep Vashist. Continuous Glucose Monitoring Systems: A Review. Diagnostics, 2013. https://doi.org/10.3390/diagnostics3040385 ↩

2. John J. Mastrototaro. The MiniMed Continuous Glucose Monitoring System. Diabetes Technology & Therapeutics, 2000. https://doi.org/10.1089/15209150050214078 ↩

3. Benjamin Jasha van Enter, Elizabeth von Hauff. Challenges and perspectives in continuous glucose monitoring. Chemical Communications, 2018. https://doi.org/10.1039/c8cc01678j ↩

4. Timothy S. Bailey, Anna Chang, Mark Christiansen. Clinical Accuracy of a Continuous Glucose Monitoring System With an Advanced Algorithm. Journal of Diabetes Science and Technology, 2014. https://doi.org/10.1177/1932296814559746 ↩

5. Ramzi Ajjan, David Slattery, Eugene Wright. Continuous Glucose Monitoring: A Brief Review for Primary Care Practitioners. Advances in Therapy, 2019. https://doi.org/10.1007/s12325-019-0870-x ↩ ↩²

6. Richard M. Bergenstal, Roy W. Beck, Kelly L. Close, et al. Glucose Management Indicator (GMI): A New Term for Estimating A1C From Continuous Glucose Monitoring. Diabetes Care, 2018. https://doi.org/10.2337/dc18-1581 ↩ ↩²

7. David Rodbard. Continuous Glucose Monitoring: A Review of Successes, Challenges, and Opportunities. Diabetes Technology & Therapeutics, 2016. https://doi.org/10.1089/dia.2015.0417 ↩

8. Bruce W. Bode. Clinical Utility of the Continuous Glucose Monitoring System. Diabetes Technology & Therapeutics, 2000. https://doi.org/10.1089/15209150050214104 ↩