The Drop That Decides: The Science of the Finger-Prick Sample

Watch a finger-prick test in almost any clinic and you will see the same small ritual. An alcohol swab, a click of the lancet, a squeeze, another squeeze, a drop touched to a strip or drawn into a cuvette, and a number on a screen within a minute. The analyser has been evaluated against a reference method, the strips are in date, the quality control passed this morning. Everyone in the room trusts the number because everyone trusts the device.

Almost nobody in that room has thought about the drop itself. It is the specimen most point-of-care testing rests on: glucose meters, haemoglobin photometers, HbA1c, lipid and INR analysers, lactate meters. Yet it is the least standardised specimen in laboratory medicine, collected by the widest range of people, few of whom were ever taught its physiology. Venous phlebotomy has order of draw, tube additives, tourniquet times and decades of pre-analytical research behind it. The finger prick has a lancet and a habit.

I spent years in hospital point-of-care services watching good devices produce numbers that did not fit the patient, and the explanation was rarely the analyser. A capillary result is only as good as the drop it was measured on, and the drop is a variable mixture that technique, circulation and the patient's own hands can shift further than most analytical error budgets allow.

What a capillary sample actually is

The word "capillary" flatters the specimen. A lancet does not tap a single vessel. It cuts through the dermal vascular bed, which in an infant's heel lies roughly 0.35 to 1.6 mm below the surface, and releases whatever is there: blood from arterioles, capillaries and venules in proportions nobody can know, plus interstitial fluid from the surrounding tissue and intracellular fluid from cells the blade has damaged. CLSI GP42, the standard for capillary collection, describes the specimen in these terms, and the Croatian national recommendations, which adapt its sixth edition (GP42-A6, 2008), put the consequence plainly: unknown proportions of blood from venules, arterioles and capillaries, contaminated to unknown extents by interstitial and intracellular fluid^[1][3].

Two consequences explain most of what follows. First, capillary blood sits physiologically between arterial and venous blood. Where tissue consumes a substance, glucose being the obvious case, the arterial side carries more of it, and a warm, well-perfused fingertip leans arterial. Second, anything added from outside the vessels dilutes or contaminates the drop. Interstitial fluid carries almost no cells and less protein than plasma, so it lowers haemoglobin and protein-bound analytes. Damaged cells release potassium. Whatever sits on the skin, sugar, alcohol, lotion, rides along with the first contact.

Kupke and colleagues showed the dilution signature decades ago. In fasting healthy adults, total protein, bilirubin, calcium, sodium and chloride were significantly lower in capillary than venous serum, by up to about 5 percent. In non-haemolysed capillary serum potassium was nearly the same as venous, and glucose tended to be higher^[4]. Note the qualifier on potassium; it matters later.

Systematic differences: capillary is not venous, and the gap moves

Most reference intervals were built on venous plasma; most point-of-care results come from capillary whole blood, often reported as a plasma equivalent. The MHRA reminds users that plasma-calibrated strips read roughly 12 percent higher than whole-blood results, so the calibration basis must be known before a reference range is applied^[21]. Beyond calibration there are true physiological differences, and they are not constant.

Glucose: the post-prandial gradient

Peripheral tissues take up glucose between the arterial and venous ends of the capillary bed, more so when insulin is high after a meal. Fasting capillary and venous samples are close; post-prandial capillary samples run higher. Lee and colleagues sampled finger-stick capillary and antecubital venous blood concurrently in 43 adults with type 1 diabetes, measuring both on glucose meters reporting plasma-equivalent values. Breakfast was eaten on average 15 minutes after the baseline sample. The mean meter-derived capillary minus venous gradient was 0.21 mmol/L at baseline (95% CI 0.08 to 0.34), 0.87 mmol/L at 60 minutes (0.66 to 1.07) and 0.52 mmol/L at 120 minutes (0.33 to 0.71), that is roughly 45 and 105 minutes after the meal; the laboratory-measured gradient at 120 minutes was 0.40 mmol/L (0.20 to 0.60). The gradient correlated with the carbohydrate eaten^[5].

In 103 healthy volunteers, Swaminathan and colleagues found the meter minus laboratory plasma difference moved from minus 0.51 mmol/L fasting to plus 0.81 mmol/L one hour after breakfast for one meter, and from minus 0.13 to plus 1.19 mmol/L for another^[6]. Colagiuri and colleagues found venous plasma higher than capillary blood fasting but lower two hours after a glucose load, so that a two-hour venous 11.1 mmol/L matched a measured capillary 11.7 mmol/L, and warned that published equivalence values could misclassify glucose tolerance^[7]. A fingertip glucose an hour after lunch and a venous glucose drawn twenty minutes later are not measuring the same compartment; a "discrepancy" may simply be physiology.

Lactate: where perfusion changes the answer

Raa and colleagues compared a handheld lactate meter with a blood gas analyser on arterial, venous and capillary samples from 49 intensive care patients and 11 healthy volunteers. On arterial blood the handheld agreed well. But fingertip lactate in the intensive care patients was 47 percent higher than arterial (95% CI 29 to 68) and earlobe lactate 27 percent higher (11 to 45). In the volunteers, who were sampled at rest and repeatedly after a maximal treadmill test, the model-estimated fingertip excess was 14 percent (4 to 24) and the earlobe showed no significant difference^[8]. The device was fine. The site, and the circulation feeding it, changed the number.

Haemoglobin: real differences, inconsistent direction

Haemoglobin is where the evidence is most contested. Neufeld and colleagues reviewed the literature and concluded that significant capillary and venous differences are common, that most but not all studies find capillary higher, and that instrument and human error cannot fully explain the gap^[9]. In one measurement system analysis, single capillary drops read a mean 0.585 g/dL above venous blood after adjustment, while pooled capillary blood did not differ^[24]. Yet in 810 school-age children in Zanzibar, capillary haemoglobin ran 0.91 g/dL lower than venous, and anaemia prevalence was 44.7 percent on capillary samples against 20.5 percent on venous^[10].

Technique, which drop, how hard the finger was worked, how warm the hand was, may contribute to that inconsistency, but biological, population and methodological differences remain possible explanations, and their relative contributions are unresolved^[9]. The WHO's 2024 guideline on haemoglobin cut-offs prefers venous blood on automated analysers in laboratory settings, accepts approved point-of-care devices where that is impractical, and keeps the threshold for non-pregnant women at 120 g/L^[23].

The first drop, the second drop, and the drop after that

Guidance on the first drop is not uniform. WHO tells the collector to wipe it away^[2], as does CLSI GP42 for capillary specimens generally, except where a point-of-care device's manufacturer specifies the first drop, as some self-test glucose meters do^[1]. The rationale is that the first drop carries the most tissue fluid and whatever was on the skin.

Glucose and the contaminated finger

Hortensius, Slingerland and colleagues measured two consecutive drops in 123 insulin-treated patients under several conditions, each compared with a control from washed, dried hands^[12]. Without hand washing, the first drop was 10 percent or more off in 11 percent of patients and the wiped second drop in 4 percent. After cutting an apple or peeling a banana, the first drop was 10 percent or more off in 88 percent: median 15.0 mmol/L against a control of 8.9 mmol/L. The second drop brought that to 11 percent, but did not remove it. Washing the fruit-exposed hands restored agreement. Their advice: wash, dry and use the first drop; if washing is impossible and the hands are not visibly soiled or exposed to sugar, use the second drop after wiping.

Hirose and colleagues added a lesson about alcohol swabs^[13]. Ten fasting volunteers with normal glucose tolerance peeled an orange, grapes or a kiwi fruit. An hour later, unwashed, the median meter glucose after grapes was 360 mg/dL against 87 mg/dL after washing with tap water. One alcohol swab left the grape median at 274 mg/dL, and even five swabs left grape and kiwi readings significantly raised. Alcohol disinfects; water removes sugar.

The competing evidence deserves its place. In 30 fasted non-diabetic adults with ordinary, uncontaminated hands, Laming and colleagues found hand washing, alcohol wipes and drop choice made no substantial difference, with mean differences of minus 0.06 to plus 0.06 mmol/L, though they did not extend this to diabetes or hypoglycaemia^[14]. The findings are compatible. On a clean finger the drop is usually fine. On a contaminated one, and nobody can see glucose on skin, the first drop can be wildly wrong.

Drop-to-drop variation in cells

For cellular analytes the problem is heterogeneity. Bond and Richards-Kortum collected six successive 20 microlitre drops from one finger prick in each of 11 donors, with the first drop wiped and no milking, and ten successive 10 microlitre drops from each of 7 donors for a point-of-care haemoglobinometer^[11]. The average coefficient of variation across drops exceeded venous controls by up to 3.4 times for haemoglobin, 5.7 times for white cells, 3 times for lymphocytes, 7.7 times for granulocytes and 4 times for platelets on an analyser, and up to 5 times for haemoglobin on the haemoglobinometer. In some donors haemoglobin moved by more than 2 g/dL between two successive drops. Results settled within instrument variability only once 60 to 100 microlitres had been accumulated.

That matters for anaemia screening, donor checks and any service that acts on a single capillary haemoglobin near a cut-off. A photometer with a precision of a few percent on well-mixed venous blood can still return a result whose real uncertainty is dominated by which drop filled the cuvette.

Squeezing, dilution and the anaemia cut-off

When the drop is slow, the instinct is to squeeze. WHO says to avoid squeezing the finger too tightly^[2]; the MHRA says excessive massaging should be avoided^[21]; the CLSI-based recommendations explain why: squeezing risks haemolysis, contamination with interstitial and intracellular fluid and obstructed flow, and "milking" can cause both haemolysis and dilution^[1]. Hortensius and colleagues found that pressure around the finger produced 10 percent or greater glucose deviations in 5 to 13 percent of participants^[12]. The alternative is to make squeezing unnecessary: warm the site, choose the right finger and lancet, let the hand hang. Warming with a warm, moist towel at no more than 42 degrees for 3 to 5 minutes is cited as increasing arterial flow up to seven-fold^[1]. A warm finger bleeds freely; a cold one gets milked.

Dilution is worth quantifying because it is invisible and, for cell-based measures, one-directional. Interstitial fluid contains essentially no red cells, so every fraction of the drop that is tissue fluid lowers haemoglobin proportionally.

Dilution and the tendency for capillary haemoglobin to read high can coexist. A warm, freely flowing drop may run high; a squeezed one low. Two collectors on the same device could in principle bias it in opposite directions, which may be one of several reasons the published comparisons disagree.

Haemolysis: the potassium trap

Squeezing, scraping the drop along the skin, residual alcohol and vigorous mixing all lyse red cells, which are rich in potassium. A laboratory flags haemolysis in serum with an index. A whole-blood point-of-care device often cannot, and may consume the whole sample^[1].

The scale is not small. Dietzen and colleagues reviewed 34 months of potassium results in children under two: 20 percent of 56,000 plasma results had significant haemolysis, nearing 50 percent in the neonatal nursery. Of 662 closely paired whole-blood and plasma results, 8 percent showed a raised whole-blood potassium with a normal plasma value, and the bias ranged from minus 1.0 to plus 4.0 mmol/L^[15]. Using true rather than simulated in vitro haemolysis, Mansour and colleagues derived potassium increases of 0.40 to 0.51 mmol/L per 0.1 g/dL of plasma haemoglobin, and advised against "correcting" haemolysed results^[16]. Algeciras-Schimnich and colleagues measured a mean free haemoglobin of 1.62 g/L, or 0.162 g/dL, across 151 neonatal heel-stick samples^[17]. Multiplying one study's mean by the other's coefficients, 1.62 x 0.40 = 0.65 and 1.62 x 0.51 = 0.83 mmol/L. This is an illustrative extrapolation across two studies, not a validated neonatal estimate or a correction factor, but it shows that ordinary heel-stick haemolysis is of a size that could push a normal potassium into apparent hyperkalaemia or mask a low one.

A capillary potassium that would change management should therefore prompt the question of haemolysis, and the procedure should say what happens next. The Croatian recommendations advise that capillary potassium and calcium be confirmed on venous blood when accuracy is critical^[1].

Haematocrit, perfusion and the critically ill patient

Haematocrit is the classic interferent for glucose strips. ISO 15197:2013 treats its influence as acceptable when the mean difference from a 42 percent haematocrit control is no more than 10 mg/dL (0.55 mmol/L) below 100 mg/dL glucose, or 10 percent above it, and requires larger effects to be described in the instructions for use. When Hattemer and Wardat tested six systems within each one's labelled haematocrit range, which varied from 10 to 65 percent to 30 to 55 percent, two exceeded 10 percent at higher glucose concentrations^[18]. Neonates, pregnant women and the acutely bleeding sit near those edges; the MHRA lists extremes of haematocrit, citing neonatal samples and pregnancy, among meter contra-indications^[21].

Perfusion is the deeper problem. A cold, shut-down or shocked periphery sends less arterial blood to the fingertip while metabolism continues. The MHRA lists peripheral circulatory failure, including hypotension, shock, severe dehydration and peripheral vascular disease, and says a venous laboratory glucose is needed where contra-indications are suspected or results unexpected^[21]. The FDA's guidance for prescription point-of-care glucose systems recommends labelling that warns of inaccurate results in severely hypotensive, dehydrated or shocked patients, and sets a tighter accuracy expectation than ISO 15197: 95 percent of results within 12 percent above 75 mg/dL, and 98 percent within 15 percent^[20].

A systematic review of 21 studies in critically ill adults found capillary meter glucose significantly less accurate than arterial measurements on blood gas analysers or meters (odds ratios for error 0.04 and 0.36 for the arterial approaches), more error in the hypoglycaemic range, and greater error with oedema, vasopressors and insulin infusion^[22]. In a sick patient the fingertip is the last place to know what the core is doing.

The patients for whom a fast capillary result feels most valuable, the cold, the shocked, the dehydrated, are the patients in whom the drop is least likely to represent their blood.

Neonates and the heel prick

The heel prick has the narrowest margin for error. The CLSI-based recommendations set a maximum depth of 2.0 mm on the medial or lateral plantar heel, because the minimum skin to perichondrium distance there is 2.4 mm and only 1.2 mm over the posterior curvature, and suggest incision devices of up to 0.85 mm for premature neonates up to 3 kg^[1]. WHO gives heel depth as under 2.4 mm, 0.85 mm for premature neonates, 2.2 mm for the finger in adults, and the middle or ring finger rather than the thumb, index or little finger^[2]. Pressing the device hard makes the cut deeper than the blade length. Capillary sampling is also among the most frequent invasive procedures in neonatal units, with one cited estimate of 56 percent of procedures^[1]. Combine that frequency with fragile neonatal red cells and the haemolysis rates above, and the heel prick deserves a service's most rigorous competency framework.

What a service should put in place on Monday

None of this argues against capillary testing. It argues for treating collection as an audited part of the test. The standard exists; CLSI GP42 is in its seventh edition^[3]. What is usually missing is local evidence that operators follow it. These are questions a POCT committee can ask; the answers must sit within the manufacturer's instructions for use and the validated local procedure, which take precedence wherever they specify the drop or technique.

1. Hand preparation. Does the glucose procedure require hands washed with soap and water and dried, and say what to do when washing is impossible^[12][13]?
2. Drop selection. Does each device's procedure state, from its instructions for use, whether the first drop is used or wiped, and is that consistent across a ward's devices?
3. Site, depth and flow. Are lancets stocked by age group, sites and maximum depths written down, and warming taught as the alternative to milking^[1][2]?
4. Who should not have a capillary test. Are labelled contra-indications listed, including shock, hypotension, peripheral circulatory failure, dehydration and out-of-range haematocrit, with a clear route to a venous sample^[20][21]?
5. Unexpected results and reporting. Is an implausible result repeated on a fresh puncture from a washed site, a management-changing capillary potassium confirmed on venous blood, and every capillary result identifiable as capillary in the record?

A competency checklist that observes the drop

Most assessments watch the device steps. Add these observed items, scored done or not done:

• Hands washed and dried, or the reason recorded.
• Site warm, correct finger or heel zone, alcohol fully dry.
• Lancet right for age and site, applied without excess pressure.
• Drop handled as the instructions for use specify.
• Flow obtained without milking or scraping; fresh puncture if it fails.
• Contra-indications and unexpected-result escalation described correctly.

An audit a service can actually run

• Observed technique audit. A trained observer watches, say, ten routine capillary tests per ward per quarter against the checklist and reports compliance item by item. Washing and milking are the items most likely to fail.
• Paired-hand concordance audit. On consenting patients, take a glucose from each washed, dried hand. Hortensius and colleagues found only 1 percent of such pairs differed by 10 percent or more^[12]. Disagreement far more often than that warrants investigation of collection technique, meter and strip performance and patient factors; paired-hand testing alone cannot identify which.
• Discrepancy audit. Where a capillary and a venous laboratory result exist for the same analyte, prefer near-simultaneous pairs, record meals and treatment between the two, and judge differences against locally validated analytical and clinical criteria; the population gradients in the table explain direction, not individual acceptance limits. Persistent unexplained gaps should prompt review of collection, haemolysis and contra-indicated use.

Feed the results into the same governance loop as QC and EQA; a device can pass every survey while its drops drift, because EQA material never passes through a finger. Our training and consultancy teams work on exactly this, and the templates library and analyte reference carry the method-specific detail.

There is something humbling about how much clinical weight rests on a drop pressed out of a fingertip. The analysers we buy are engineered to remarkable standards and wrapped in regulation. The drop we give them was prepared by whoever was on shift, from a hand that may have peeled an orange, on a finger that may have been cold, with a squeeze nobody measured. In most services the next gain in point-of-care quality will not come from a better device. It will come from someone watching how the blood gets onto the strip.

Sources and notes

Most of the quantitative findings here come from single studies with modest sample sizes, in specific populations and on specific devices, and should be read as demonstrations of mechanism and magnitude rather than estimates for any particular service or product. The capillary versus venous haemoglobin literature is genuinely inconsistent in direction, and this article presents both sides. The Bond and Richards-Kortum ratios are the maximum values reported in the abstract. The fruit studies used deliberate contamination; the Laming study shows that on clean hands, drop selection and hand preparation make little difference. CLSI GP42 is a paid standard. Procedural details are cited through the open-access Croatian recommendations, which adapt the sixth edition (GP42-A6, 2008) with permission; wording in the current seventh edition (2020) has not been separately verified. The potassium calculation combines figures from two different studies and is an illustrative extrapolation. The Lee glucose gradients are meter-derived plasma-equivalent values, and the Raa volunteer figure includes post-exercise samples. The worked examples are arithmetic illustrations using published figures and, for haemoglobin, a hypothetical patient and an assumed dilution fraction. Figure 1 is a schematic; Figures 2, 3 and 4 plot published data.

1. Krleza JL, Dorotic A, Grzunov A, Maradin M, Croatian Society of Medical Biochemistry and Laboratory Medicine. Capillary blood sampling: national recommendations on behalf of the Croatian Society of Medical Biochemistry and Laboratory Medicine. Biochemia Medica, 2015 (based on CLSI GP42-A6, 2008; composition of capillary blood; first drop; milking; heel depth 2.0 mm, perichondrium 2.4 mm and 1.2 mm, vascular bed 0.35 to 1.6 mm; lancet table; warming up to seven-fold flow; 56% of neonatal procedures; potassium and calcium confirmation).
2. World Health Organization. Practical guidance on capillary sampling (finger and heel-prick), WHO guidelines on drawing blood: best practices in phlebotomy. WHO, 2010 (wipe away first drop; avoid squeezing; heel depth under 2.4 mm; 0.85 mm for premature neonates; finger 2.2 mm in adults; middle or ring finger).
3. Clinical and Laboratory Standards Institute. GP42, Collection of Capillary Blood Specimens, seventh edition. CLSI, 2020.
4. Kupke IR, Kather B, Zeugner S. On the composition of capillary and venous blood serum. Clinica Chimica Acta, 1981 (total protein, bilirubin, calcium, sodium, chloride lower by up to 5% in capillary serum; potassium nearly identical when not haemolysed).
5. Lee I, Lunt H, Chan H, Heenan H, Berkeley J, Frampton CM. Postprandial capillary-venous glucose gradient in Type 1 diabetes: magnitude and clinical associations in a real world setting. Diabetic Medicine, 2016 (n = 43; meter-derived gradient 0.21, 0.87 and 0.52 mmol/L at 0, 60 and 120 min from baseline, breakfast about 15 min after baseline; laboratory 0.40 mmol/L at 120 min).
6. Swaminathan A, Lunt H, Chang WS, Logan FJ, Frampton CM, Florkowski CM. Impact of prandial status on the comparison of capillary glucose meter and venous plasma glucose measurements in healthy volunteers. Annals of Clinical Biochemistry, 2013 (n = 103; minus 0.51 to plus 0.81 and minus 0.13 to plus 1.19 mmol/L).
7. Colagiuri S, Sandbaek A, Carstensen B and colleagues. Comparability of venous and capillary glucose measurements in blood. Diabetic Medicine, 2003 (2 h venous 11.1 mmol/L corresponding to measured capillary 11.7 mmol/L).
8. Raa A, Sunde GA, Bolann B and colleagues. Validation of a point-of-care capillary lactate measuring device (Lactate Pro 2). Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine, 2020 (fingertip 47%, earlobe 27% above arterial in ICU; fingertip 14% in volunteers sampled at rest and after a maximal treadmill test; 95% CIs).
9. Neufeld LM, Larson LM, Kurpad A, Mburu S, Martorell R, Brown KH. Hemoglobin concentration and anemia diagnosis in venous and capillary blood: biological basis and policy implications. Annals of the New York Academy of Sciences, 2019.
10. Azizi K, Lukindo T, Hendry S and colleagues. Comparison of capillary and venous hemoglobin determination among school-age children: insights from Zanzibar, Tanzania. Advances in Laboratory Medicine, 2026 (n = 810; mean difference minus 0.909 g/dL; anaemia 44.7% capillary vs 20.5% venous).
11. Bond MM, Richards-Kortum RR. Drop-to-drop variation in the cellular components of fingerprick blood: implications for point-of-care diagnostic development. American Journal of Clinical Pathology, 2015 (CV up to 3.4x Hb, 5.7x WBC, 3x lymphocytes, 7.7x granulocytes, 4x platelets; up to 5x Hb on haemoglobinometer; stable at 60 to 100 microlitres; more than 2 g/dL between successive drops in some donors, per the Rice University release).
12. Hortensius J, Slingerland RJ, Kleefstra N and colleagues. Self-monitoring of blood glucose: the use of the first or the second drop of blood. Diabetes Care, 2011 (n = 123; 11% and 4% unwashed; 88% and 11% fruit; median 15.0 vs 8.9 mmol/L; pressure 5 to 13%; 1% between hands).
13. Hirose T, Mita T, Fujitani Y, Kawamori R, Watada H. Glucose monitoring after fruit peeling: pseudohyperglycemia when neglecting hand washing before fingertip blood sampling. Diabetes Care, 2011 (grape median 360 vs 87 mg/dL; 274 mg/dL after one alcohol swab).
14. Laming B, Smee S, Riddell H, Spence AL, Brade CJ, Davey RJ. The effect of preanalytical factors on capillary blood glucose readings from point-of-care devices. Journal of Diabetes Science and Technology, 2025 (n = 30; mean differences minus 0.06 to plus 0.06 mmol/L).
15. Dietzen DJ, Jackups R, Zaydman MA. Clinical implications of inaccurate potassium determination in hemolyzed pediatric blood specimens. Clinica Chimica Acta, 2024 (20% significant haemolysis; nearing 50% in neonatal nursery; 8% discordant pairs; bias minus 1.0 to plus 4.0 mmol/L).
16. Mansour MM, Azzazy HM, Kazmierczak SC. Correction factors for estimating potassium concentrations in samples with in vitro hemolysis: a detriment to patient safety. Archives of Pathology and Laboratory Medicine, 2009 (0.40 to 0.51 mmol/L per 0.1 g/dL plasma haemoglobin).
17. Algeciras-Schimnich A, Cook WJ, Milz TC, Saenger AK, Karon BS. Evaluation of hemoglobin interference in capillary heel-stick samples collected for determination of neonatal bilirubin. Clinical Biochemistry, 2007 (mean free haemoglobin 1.62 g/L in 151 heel-stick samples).
18. Hattemer A, Wardat S. Evaluation of hematocrit influence on self-monitoring of blood glucose based on ISO 15197:2013. Journal of Diabetes Science and Technology, 2018 (criteria 10 mg/dL or 10% against 42% haematocrit control; two of six systems exceeded 10%).
19. Jendrike N, Baumstark A, Pleus S, Mende J, Haug C, Freckmann G. Assessment of system accuracy, intermediate measurement precision, and measurement repeatability of a blood glucose monitoring system based on ISO 15197. Journal of Diabetes Science and Technology, 2019 (ISO 15197:2013: 95% within 15 mg/dL or 15%; 99% in consensus error grid zones A and B).
20. US Food and Drug Administration. Blood glucose monitoring test systems for prescription point-of-care use: guidance for industry and FDA staff. FDA, 2020 (95% within 12%, 98% within 15%; labelling for hypotension, dehydration and shock).
21. Medicines and Healthcare products Regulatory Agency. Point of care testing: blood glucose meters, advice for healthcare professionals. MHRA, 2013 (plasma calibration about 12% higher; avoid excessive massaging; contra-indications including peripheral circulatory failure and extremes of haematocrit).
22. Inoue S, Egi M, Kotani J, Morita K. Accuracy of blood-glucose measurements using glucose meters and arterial blood gas analyzers in critically ill adult patients: systematic review. Critical Care, 2013 (21 studies; odds ratios for error 0.04 and 0.36 vs capillary meter).
23. World Health Organization. Guideline on haemoglobin cutoffs to define anaemia in individuals and populations. WHO, 2024 (non-pregnant women below 120 g/L; venous blood on automated analysers preferred in laboratory settings, as summarised by the DHS Program).
24. Mendez-Gomez-Humaran I, De la Cruz-Gongora V, Dary O, Shamah-Levy T. Capillary drops, capillary pooled, and venous blood samples for determining hemoglobin concentration using HemoCue: a measurement system analysis. PLOS One, 2024 (single capillary drops 0.585 g/dL above venous after adjustment; pooled not different).