Quarterly Blood Tests: When Three Months Is the Right Interval, and When It Is Not

Quarterly is the right interval for a small set of analytes and wasteful for most of the rest. The useful cadence for any given marker is set by two numbers: how fast the underlying biology moves, and how large a change the assay plus your own day-to-day variation can manufacture out of nothing. HbA1c integrates glucose exposure over roughly the preceding 90 to 120 days, so sampling it every three months gives you nearly independent observations. Sodium moves within hours and has a within-person coefficient of variation under one percent, so quarterly sampling tells you almost nothing you did not already know. Our recommendation is a tiered schedule: a compact quarterly core panel, a wider annual panel, and a handful of once-in-a-lifetime measurements, with every draw standardized hard enough that the numbers are comparable.

What sets the interval: biological variation and analyte kinetics

Every analyte has a within-subject biological variation, conventionally written CV_I, which is the percentage spread of repeated measurements on a healthy person at steady state. It sits alongside analytical variation, CV_A, from the instrument and reagent lot. Published estimates cluster in recognizable bands: roughly 0.5 to 2 percent for sodium, calcium, and HbA1c, around 4 to 7 percent for creatinine and total cholesterol, 15 to 20 percent for ALT, triglycerides, ferritin, and TSH, and above 40 percent for high-sensitivity CRP. Careful repeated-sampling studies in healthy volunteers show the same pattern for immune markers, with within-subject variation for several cell subsets and cytokines large enough that a single value is nearly uninterpretable and the derived analytical quality goals are correspondingly loose.¹

Kinetics then tell you how long to wait for a genuinely new observation. Red cell indices (MCV, hemoglobin, reticulocyte-derived quantities) reflect a circulating population with a lifespan near 120 days. The hemoglobin and reticulocyte parameters used to model plasma and red cell volume move on that timescale rather than hourly.² Lipoprotein measures respond to diet and pharmacology within two to six weeks. Liver enzymes can swing in days. Ferritin is an acute phase reactant and will follow an infection up and back down over two to three weeks. If you sample faster than the analyte’s response time, consecutive points are correlated and you are mostly measuring noise, which is what makes monthly full panels expensive without being informative.

Deciding whether a change is real

The single most useful piece of arithmetic in longitudinal self-measurement is the reference change value:

RCV = Z * sqrt(2) * sqrt(CV_A^2 + CV_I^2)

With Z = 1.96 for a two-sided 95 percent criterion, the multiplier is about 2.77. Plug in plausible numbers and the consequences are immediate. For HbA1c with CV_I near 2 percent and CV_A near 2 percent, RCV is roughly 8 percent, so a move from 5.2 to 5.5 percent is inside noise while 5.2 to 5.8 is not. For ALT with CV_I near 20 percent, you need something like a 60 percent change before a single pair of draws says anything. For hsCRP the RCV exceeds 100 percent, which is why one elevated CRP is close to uninformative and why the standard practice is to take the lower of two measurements separated by a couple of weeks.

Write the RCV table once and keep it next to your data, because it is also the antidote to multiple comparisons. A 40-analyte panel run quarterly produces 160 comparisons a year. At a nominal 5 percent per-test false positive rate you should expect roughly eight flagged analytes annually with nothing wrong. We treat a signal as worth attention only when it clears the RCV and moves in the same direction on two consecutive draws, or when it is corroborated by a mechanistically related analyte. Examples of related analytes are ferritin with transferrin saturation, ALT with AST and GGT, and creatinine with cystatin C.

Which analytes belong on a quarterly panel

We would put roughly twenty things in the quarterly core. They are chosen because they move on a three-month timescale, have modest biological variation relative to the changes you care about, and are cheap.

• CBC with differential
• a comprehensive metabolic panel
• HbA1c
• fasting glucose and fasting insulin
• apolipoprotein B with a standard lipid panel
• hsCRP
• ferritin with iron and transferrin saturation
• TSH

ApoB is the lipid measure we track rather than LDL-C, because it is a direct particle count and is not distorted by triglycerides at the time of the draw.

The annual panel adds the slower or more expensive items.

• vitamin D and B12
• homocysteine
• uric acid
• urine albumin-to-creatinine ratio
• cystatin C
• a sex-hormone panel if relevant

Lipoprotein(a) is largely genetically determined and belongs in the measure-once tier, as do most sequencing-based assays. If you are past 65, the argument for the quarterly core gets stronger rather than weaker, since renal function, hemoglobin, and glucose regulation all drift faster and the four-point-per-year series is what makes a drift distinguishable from a blip. Any value outside the reference interval, and any confirmed trend, is a conversation with a physician rather than something to act on from a spreadsheet.

Two cautions on the frontier markers. Blood-based measures of neurodegeneration include plasma p-tau217, Aβ42/40, NfL and GFAP. They have improved enough to be used in specialist clinics and trials, where they show strong discrimination against amyloid PET and CSF standards.³ Their behavior in unselected, asymptomatic people is much less settled, and the evidence gap between research performance and population screening utility is real.⁴ Preanalytic handling matters disproportionately for this class of assay, which is why the field wrote explicit standardization guidelines covering tube type, processing delay, centrifugation, and freeze-thaw cycles.⁵ We do not think these belong in a self-directed quarterly panel yet.

Controlling the preanalytic noise

Most of the variance people attribute to biology is procedural. Fix these and hold them fixed. Draw between 07:00 and 09:00, fasted 10 to 12 hours with water allowed. Sit quietly for 15 minutes before venipuncture, because posture alone shifts plasma volume and hence hemoglobin and albumin by several percent. Do no vigorous exercise for 48 hours, which otherwise elevates CK, AST, ALT, and CRP. Take no draw within two to three weeks of an infection or vaccination if you care about ferritin, WBC, or CRP. Use the same laboratory and the same instrument platform every quarter. A method change (Jaffe to enzymatic creatinine, or a ferritin immunoassay recalibration) can produce a step larger than anything physiological, and the lab will not tell you unless you ask. The preanalytic checklist developed for Alzheimer biomarker studies is a good template for the general problem, and its core message is that tube type and time-to-processing are part of the measurement.⁵

Getting the data in a form you can compute on

Ask the laboratory for an HL7 v2 ORU message or a FHIR Observation bundle rather than a PDF. Most patient portals expose a FHIR API and will return JSON with LOINC codes, which is what you want for joining across labs, since analyte names are not stable and units are not either. Normalize into one long-format table with these columns.

• collection_datetime
• loinc
• analyte
• value
• unit
• ref_low
• ref_high
• method
• lab_id
• fasting_hours
• time_of_day

Store the unit as given and convert in a view, never in place. Creatinine in mg/dL versus µmol/L, and ferritin in ng/mL versus µg/L (identical), are the two places we see silent corruption most often. Recompute eGFR yourself with the 2021 CKD-EPI equation so that historical values reported under older equations do not create a phantom step.

Analyzing a four-to-eight-point series

With four to eight points per analyte, resist modeling. Fit a Theil–Sen slope, which is the median of pairwise slopes and therefore insensitive to a single outlier, and test monotonicity with a Mann–Kendall statistic. Report the slope in analyte units per year alongside the RCV, and flag only analytes whose annual change exceeds the RCV. Plot the series with the RCV band drawn around the first observation. That single plot answers most questions faster than any statistic.

The longer-run reason to keep a clean quarterly series is that trajectories in routine blood tests carry signal well before anything is symptomatic. In a large Israeli cohort, machine learning on ordinary CBC and chemistry results identified people who would later be diagnosed with Crohn’s disease years ahead of diagnosis.⁶ Those were population-trained models on retrospective data. They establish the substrate, though they were not a personal early-warning system. The information is in the common, cheap analytes and in their movement over time. It is not in exotic assays.

Questions people also ask

Do you really need bloodwork every year? For a healthy adult, an annual wide panel plus a quarterly core panel is a defensible design. The quarterly cadence is what makes the annual numbers interpretable, because you can tell a trend from a single odd draw.

Why blood tests every three months? Because several of the most informative analytes integrate over roughly that window. HbA1c reflects about 90 to 120 days of glucose exposure, and red cell indices turn over on a similar timescale, so three months produces close to an independent observation.

Is it okay to get blood drawn two days in a row? Physiologically yes for standard panel volumes, which total on the order of 20 to 40 mL. Analytically, two consecutive days is useful for confirming a suspect result on a high-variation marker such as CRP or ferritin, and almost useless for a low-variation one such as sodium.

How often is too often for bloodwork? When the interval is shorter than the analyte’s response time, consecutive results are correlated and you are paying to measure noise, with the added cost of more false flags. Monthly full panels fall into that category for nearly everything except a marker you are deliberately tracking through a change.

Can bloodwork change a lot in a year? Yes, and the way to know is to compare the change against the reference change value rather than the population reference interval. A value can stay inside the “normal” range while moving well beyond what your own variation would produce.

Footnotes

1. Najib Aziz, Roger Detels, Joshua J. Quint, et al. Biological variation of immunological blood biomarkers in healthy individuals and quality goals for biomarker tests. BMC Immunology, 2019. https://doi.org/10.1186/s12865-019-0313-0 ↩

2. Louisa Margit Lobigs, Pierre‐Edouard Sottas, Pitre Collier Bourdon, et al. The use of biomarkers to describe plasma‐, red cell‐, and blood volume from a simple blood test. American Journal of Hematology, 2016. https://doi.org/10.1002/ajh.24577 ↩

3. Oskar Hansson, Kaj Blennow, Henrik Zetterberg, et al. Blood biomarkers for Alzheimer’s disease in clinical practice and trials. Nature Aging, 2023. https://doi.org/10.1038/s43587-023-00403-3 ↩

4. Willem A Van Gool, J A Floria Siebrand, Carol Brayne, et al. Evidence gap in blood biomarkers for Alzheimer’s disease. BMJ, 2025. https://doi.org/10.1136/bmj-2025-084781 ↩

5. Sid E. O’Bryant, Veer Gupta, Kim Henriksen, et al. Guidelines for the standardization of preanalytic variables for blood‐based biomarker studies in Alzheimer’s disease research. Alzheimer’s & Dementia, 2014. https://doi.org/10.1016/j.jalz.2014.08.099 ↩ ↩²

6. Raffi Lev-Tzion, Amir S Dolev, Shira Yuval Bar-Asher, et al. Machine learning for predicting Crohn’s disease from routine blood tests years before diagnosis: results from the epi-IIRN cohort. Journal of Crohn’s and Colitis, 2025. https://doi.org/10.1093/ecco-jcc/jjaf143 ↩