CGM vs finger-stick meter: accuracy, lag and when to check

How do a meter and a CGM measure glucose?

A meter measures glucose directly in a drop of blood, while a CGM estimates it in the interstitial fluid, the fluid in the spaces around your cells ^{3, 10}.

For a meter check, you prick the side of a fingertip with a small needle called a lancet and put the drop of blood on a test strip in the meter ^{1, 3}. The international meter standard describes these as capillary blood samples, tested by lay people for self-monitoring ¹¹. Each result is a snapshot of “one moment in time” ³.

A CGM sensor sits under the skin, often on the belly or arm, and estimates glucose in the fluid between your cells, which the US National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) describes as very similar to the level in your blood ². Sensors typically last 7 to 15 days, depending on the brand ¹⁰. Because readings arrive all day, a CGM can sound an alarm when glucose is too low or too high ².

How accurate are blood glucose meters?

A meter that meets the international standard ISO 15197:2013 gives at least 95% of results within 0.83 mmol/L (15 mg/dL) of a laboratory reference when glucose is below 5.55 mmol/L (100 mg/dL), and within 15% at 5.55 mmol/L and above ⁴. It must also put 99% of results in zones A or B of the consensus error grid, the clinically acceptable zones ⁴.

For home-use meters, the US Food and Drug Administration (FDA) asks for 95% of results within 15% across the meter’s usable range and 99% within 20% ⁴. The ADA notes that meters on the market only have to meet the standard they were approved under, which may not be the current one, and that checking accuracy after approval is left to the manufacturer ⁴.

Independent testing has found gaps. A US surveillance program tested 18 cleared meters in 1,035 people across three blinded studies, against laboratory plasma glucose ⁵. A result counted as accurate if it was within 15%, or within 0.83 mmol/L (15 mg/dL) below 5.55 mmol/L (100 mg/dL). Against a standard the authors describe as more lenient than regulatory standards, 6 of the 18 meters passed in all three studies and 4 passed in none ⁵.

Diabetes Canada recommends comparing your meter with a fasting laboratory glucose taken at the same time at least once a year, and whenever your A1c (a blood test that reflects average glucose; see A1c and average glucose) doesn’t match your meter readings. Above 4.2 mmol/L (about 76 mg/dL), a difference under 15% is acceptable ¹³.

What can throw a meter reading off

Mayo Clinic advises washing and drying your hands, because food on your fingers can give a wrong reading, and using only unexpired strips made for your meter ¹. Some medicines can interfere with meters ⁴. Forearm or palm readings may be less accurate than fingertip readings, especially after a meal or during exercise ¹.

How accurate is a CGM compared with a finger-stick meter?

In their main published studies, the Dexcom G7 and FreeStyle Libre 3 read an average of 7.8% to 9.1% away from laboratory blood glucose, a measure called MARD ^{6, 7}. MARD stands for mean absolute relative difference: each sensor reading is paired with a comparison value taken at the same time, and the gaps, as percentages, are averaged ¹⁴. Lower means closer agreement on average.

• Dexcom G7. 316 adults with type 1 or type 2 diabetes at 12 US sites wore sensors for 10.5 days, with frequent comparisons against a laboratory analyzer ⁶. MARD was 8.2% on the arm and 9.1% on the abdomen. On the arm, 89.6% of readings were within 15% or 0.83 mmol/L (15 mg/dL) of the reference. Agreement was best when glucose changed by no more than 0.06 mmol/L (1 mg/dL) a minute ⁶.
• FreeStyle Libre 3. 100 people aged 4 and older with type 1 or type 2 diabetes were evaluated over 14 days ⁷. Against a laboratory analyzer in those aged 6 and older, MARD was 7.8%, and 93.4% of readings were within 20% or 1.1 mmol/L (20 mg/dL) ⁷.

MARD is easy to calculate but harder to interpret than it looks. A 2020 review notes that it also reflects the error of the comparison method and changes with how many readings are low, how fast glucose was moving, the study population and the number of comparisons; studies in type 1 diabetes most probably give higher values ¹⁴. The authors concluded that MARD “should not be used as the sole parameter” to judge a CGM ¹⁴.

What the US iCGM standard requires

Most CGMs are cleared by the FDA as integrated CGMs (iCGMs), class II devices that can work with other connected devices ⁴, such as automated insulin dosing systems ¹². Adult study results must show, at the lower bound of a one-sided 95% confidence interval ¹²:

• below 3.9 mmol/L (70 mg/dL): more than 85% of readings within 0.83 mmol/L (15 mg/dL) of blood glucose, and more than 98% within 2.2 mmol/L (40 mg/dL);
• 3.9 to 10.0 mmol/L (70 to 180 mg/dL): more than 70% within 15%, and more than 99% within 40%;
• above 10.0 mmol/L (180 mg/dL): more than 80% within 15%, and more than 99% within 40%;
• across the whole range: more than 87% within 20% ¹².

No sensor reading below 3.9 mmol/L may match a blood glucose above 10.0 mmol/L, or the reverse. And no more than 1% of readings may show glucose rising faster than 0.06 mmol/L (1 mg/dL) a minute when it is actually falling faster than 0.11 mmol/L (2 mg/dL) a minute, or the reverse ¹².

Why do CGM and meter readings differ?

Glucose reaches the fluid under the skin a few minutes after it reaches the blood, so the two readings drift apart most when glucose is rising or falling fast ^{10, 4}.

Cleveland Clinic explains that glucose “hits your bloodstream first and then leaks into your interstitial fluid” ¹⁰. Mayo Clinic researchers measured the delay by injecting traceable glucose into a vein and sampling the fluid under the skin of the abdomen. In eight healthy adults it took 5.3 to 6.2 minutes on average ⁹. In six adults with type 1 diabetes the median was 6.8 minutes, ranging from 4.8 to 9.8 ⁸.

In plain terms: When glucose is steady, a CGM and a meter tend to read closer together. After a meal, during exercise or while treating a low, blood glucose moves first and the sensor catches up minutes later, so for a while the two numbers can look quite different.

Other reasons the numbers can disagree

• Meters have a margin of error too. A meter can read 15% away from a laboratory value and still meet its standard ⁴.
• Pressure on the sensor. Healthy volunteers filmed while they slept wore four sensors each. Single sensors sometimes read more than 1.4 mmol/L (25 mg/dL) away from the middle value, strongly linked to the person lying on that sensor, and the authors attributed the sudden drops to reduced blood flow from compression ¹⁵. These dips are sometimes called compression lows.
• Medicines and supplements. The ADA lists substances that make some sensors read higher than actual glucose, including acetaminophen above 4 g a day for the Dexcom G6 and G7, vitamin C above 500 mg a day for the FreeStyle Libre 2 and 3, and hydroxyurea for the Dexcom G6 and G7 ⁴.

When should you check with a meter if you use a CGM?

The ADA says CGM users must have access to a meter at all times, and should use it whenever the CGM is suspected to be inaccurate, while a new sensor warms up, when transmission is disrupted, for calibration if needed, when a warning message appears and when CGM supplies are delayed ⁴.

Cleveland Clinic puts it plainly: “don’t completely rely on CGMs without finger stick checks, especially if your CGM readings don’t match the symptoms of low or high blood sugar you’re experiencing, or if your CGM gives you an error message” ³.

Device labelling differs. Diabetes Canada notes that some real-time CGMs require a meter check before a treatment change or before treating a suspected low, while others only require one if alerts and readings don’t match symptoms ¹³. NIDDK adds that you may need a finger stick if you are changing your insulin dose ². Your device’s instructions and your care team tell you which applies. For treating lows, see low blood sugar.

Do CGMs still need calibration?

Some do. With some CGMs you still need finger-stick tests that “set the CGM,” and the user guide says how often ¹. Others are factory calibrated: Diabetes Canada notes that flash glucose monitors need no finger-stick calibration ¹³, and calibration is optional on the Dexcom G7 ⁶.

What do trials and guidelines say about CGM?

In randomized trials of adults who take insulin, CGM lowered A1c by 0.4 to 0.6 percentage points more than the comparison approach, and guidelines now recommend CGM for people on insulin ^{16, 17, 18, 4}.

• DIAMOND randomly assigned 158 adults with type 1 diabetes on insulin injections, with an A1c of 7.5% to 9.9%, to CGM or usual care ¹⁶. Over 24 weeks, A1c fell 1.0 point with CGM and 0.4 with usual care, an adjusted difference of 0.6. Median time below 3.9 mmol/L (70 mg/dL) was 43 minutes a day against 80 ¹⁶.
• GOLD enrolled 161 adults in Sweden with type 1 diabetes on injections, who used CGM and conventional therapy for 26 weeks each, in random order ¹⁷. In the 142 with data from both periods, A1c averaged 7.92% (63 mmol/mol) with CGM and 8.35% (68 mmol/mol) with conventional therapy ¹⁷.
• MOBILE randomly assigned 175 adults with type 2 diabetes, cared for in primary care and taking one or two daily injections of basal insulin without mealtime insulin, to CGM or a meter ¹⁸. Over 8 months, A1c fell from 9.1% to 8.0% with CGM and from 9.0% to 8.4% with a meter. Time in range, 3.9 to 10.0 mmol/L (70 to 180 mg/dL), was 59% against 43% ¹⁸.

The ADA’s 2026 Standards of Care recommend CGM at diabetes onset and anytime after for children, adolescents and adults on insulin, for people on other medicines that can cause lows, and on any treatment where CGM helps in management. For people on insulin, CGM should be used as close to daily as possible ⁴. Diabetes Canada’s 2025 type 1 guideline states that “CGM is recommended for people using insulin injections or pumps” ¹⁹, and its 2018 guideline says flash glucose monitoring may be offered to people with diabetes to decrease time spent low ¹³.

What are the practical trade-offs?

A CGM gives more information with fewer finger pricks, but it is worn on the body, needs a warm-up after each new sensor and costs more than a meter ^{2, 10}.

• Skin. Contact dermatitis, irritant or allergic, has been reported with all devices that attach to the skin, and the ADA says skin reactions should be assessed and addressed ⁴.
• Warm-up. A new sensor typically needs 30 minutes to two hours before it shares data ¹⁰; the G7 study reports 27 minutes ⁶.
• Nights. Lying on the sensor can cause false drops ¹⁵, so a short, sharp dip during sleep deserves a second look.
• Cost. A CGM costs more than a meter but may be covered by insurance ², and the ADA notes that coverage can lag behind device availability ⁴. Test strips can be expensive too ³.