バイオマーカートラッキングアプリ：患者が必要とする9つの機能

What a good biomarker tracking app should do first

I’m Thomas Klein, MD, Chief Medical Officer at Kantesti, and the mistake I see most often is simple: patients compare a flagged value from one lab with an unflagged value from another lab and assume their body changed. Often the measurement context changed instead; units, fasting status, assay method, time of day, and hydration can all move results by clinically visible amounts.

A useful app should answer 4 patient questions in under 60 seconds: what changed, how much it changed, whether that change exceeds expected biological variation, and what should be discussed with a clinician. For a deeper patient primer on separating signal from noise, our guide to 実際の検査トレンド shows why a single abnormal flag rarely tells the whole story.

The practical buyer test is blunt. If an app cannot show the original unit, the lab-specific reference interval, the date, and the clinical context beside the graph, it is not really tracking health; it is decorating numbers.

Feature 1: upload that preserves the original report

Manual entry is where quiet errors creep in. I have reviewed patient spreadsheets where sodium 140 mmol/L became 140 mg/dL, vitamin D 25 nmol/L was treated like 25 ng/mL, and a platelet count of 145 x10⁹/L was entered as 145,000 without context; each error changed the patient’s anxiety level more than their care plan.

A serious app should let you upload a full report and later reopen the image or PDF beside the interpretation. Kantesti’s 血液検査PDFのアップロード workflow is built around that audit trail because physicians do not trust orphan numbers either.

Look for optical reading that captures footnotes and specimen notes, not just the table. Phrases such as hemolyzed specimen, non-fasting, estimated GFR calculated, or result repeated can completely change what a number means.

Feature 2: unit conversion and reference-range mapping

Vitamin D is the classic trap: 50 nmol/L equals 20 ng/mL, not 50 ng/mL. Glucose 5.6 mmol/L equals about 101 mg/dL, and cholesterol 5.2 mmol/L equals about 201 mg/dL; a graph that mixes those units without conversion is clinically unsafe.

Reference ranges also differ by method and population. Some European laboratories use lower upper limits for ALT than older US panels, and creatinine ranges vary by sex, muscle mass, and assay; our article on 異なる検査室の単位 walks through the common conversions patients misread.

A good app should show both absolute value and position within that lab’s range. In my experience, this prevents a lot of needless worry when a value looks higher only because the new lab narrowed its normal interval.

Feature 3: trend graphs that respect biological variation

Many analytes have a reference change value, which estimates how large a change must be before it likely reflects a real physiological shift. The rough formula uses analytical variation plus within-person biological variation; patients do not need the math, but the app should behave as if it knows the math.

ALT may fluctuate 20–30% after exercise, alcohol exposure, fatty meals, or a viral illness. TSH can move by time of day and sleep disruption, while ferritin can rise during tissue response even when iron stores have not improved; our explainer on 血液検査のばらつき gives patient-level examples.

Kantesti AI treats trends differently from isolated flags by comparing date sequence, magnitude, biomarker family, and clinical context. That matters because a 2-point CRP fall from 6 to 4 mg/L after an infection is reassuring, while a potassium shift from 4.7 to 6.2 mmol/L needs immediate verification.

Feature 4: context tags for fasting, exercise, illness and time

Triglycerides can rise substantially after meals, fasting glucose can rise after poor sleep, and cortisol interpretation is almost meaningless without collection time. Morning testosterone is preferred because levels can be 20–40% lower later in the day in many men, especially when sleep is short.

I once reviewed a 52-year-old marathon runner with AST 89 IU/L and ALT 42 IU/L. The liver panic stopped when CK came back over 2,000 IU/L after a race; our guide to は、適応（順応）と病理を切り分けるのに役立ちます。 covers why muscle can masquerade as liver injury.

The app should ask small questions at the right time, not bury patients in forms. Fasting status, workout within 48 hours, illness within 2 weeks, and new supplements within 30 days explain a surprising share of borderline changes.

Feature 5: cross-lab comparison without false alarms

Creatinine is a good example. Jaffe and enzymatic methods can produce slightly different values, and eGFR calculations depend on the equation used; a shift from eGFR 78 to 69 mL/min/1.73 m² may be method noise, hydration, or true kidney change depending on repeat results.

Thyroid antibodies, vitamin D, and some hormone assays vary enough between platforms that a clean comparison needs method awareness. If you often use large commercial laboratories, our patient guide to lab result flags and trends explains why the same result can be presented differently.

The best display is a layered view: original result, converted result, lab-specific range, and a confidence note about comparability. That is less flashy than a smooth graph, but it is much closer to how clinicians think.

Feature 6: pattern reading across related biomarkers

Ferritin below 30 ng/mL with rising RDW and low transferrin saturation points toward early iron deficiency, even when hemoglobin is still normal. Ferritin 250 ng/mL with CRP 18 mg/L tells a different story because ferritin behaves as an acute-phase reactant.

A1c and fasting glucose can disagree. A1c 5.4% with fasting glucose 118 mg/dL may reflect early insulin resistance, dawn phenomenon, anemia effects, or recent diet change; our guide to 血液検査の数値パターン helps patients read those disagreements calmly.

Kantestiの AIを活用した血液検査の解釈 platform weighs biomarker families rather than treating every red flag as equal. The reason we worry about high AST plus high CK is muscle injury, whereas high AST plus high bilirubin and high INR raises a different level of concern.

Feature 7: risk alerts tied to clinical thresholds

The 2018 AHA/ACC cholesterol guideline treats LDL-C ≥190 mg/dL as severe hypercholesterolemia that usually warrants prompt risk assessment and therapy discussion, not casual annual watching (Grundy et al., 2019). ApoB can add value when triglycerides exceed 200 mg/dL because LDL-C may understate particle burden.

The same logic applies to diabetes and kidney risk. HbA1c ≥6.5% diagnoses diabetes when confirmed appropriately, while eGFR <60 mlmin1.73 m² for 3 months or urine albumin-creatinine ratio ≥30 mgg suggests chronic kidney disease; our 医学的な検証基準 describe how we separate guideline-based alerts from wellness commentary.

A good alert says what to do next: repeat soon, call your clinician, seek urgent care, or interpret with context. Red without action is just noise with better lighting.

Feature 8: family profiles and caregiver permissions

A hemoglobin of 11.2 g/dL may be interpreted differently in a toddler, a pregnant adult, an older man, and a person receiving chemotherapy. Pediatric alkaline phosphatase can be much higher during growth, and teenage iron deficiency can appear before hemoglobin falls.

Caregivers also need audit trails. If you track a parent’s eGFR, potassium, INR, or hemoglobin, the app should show who uploaded the result, when it was interpreted, and whether any recommendation was shared; our 家族の医療記録アプリ guide goes deeper on consent and safety.

Most families do not need 200 biomarkers. They need the right 12–20 markers, trended reliably, with separate profiles and clear permissions.

Feature 9: plain-language explanations with clinician-ready export

A good summary does not say your liver is bad. It says ALT rose from 32 to 58 IU/L over 4 months, AST is 41 IU/L, bilirubin and ALP are normal, recent strenuous exercise was reported, and repeat testing in 2–6 weeks may be reasonable if symptoms are absent.

Our doctors and reviewers, including the clinicians listed on Kantesti’s 医療諮問委員会, push for explanations that show uncertainty instead of hiding it. Sometimes the honest answer is: this could be noise, medication effect, early disease, or a sample issue, and the next best step is a targeted repeat.

Clinician-ready export should be short enough to read in a 10-minute visit. If your app produces 12 pages of generic advice for a borderline bicarbonate of 21 mmol/L, it is serving the software more than the patient.

Which biomarkers are worth tracking every year

The ADA Standards of Care define HbA1c <5.7% as usual normal, 5.7–6.4% as prediabetes, and ≥6.5% as diabetes when confirmed in the right clinical setting (American Diabetes Association Professional Practice Committee, 2024). That makes HbA1c one of the few markers where a small threshold crossing can change the diagnosis conversation.

KDIGO 2024 emphasizes both eGFR and albuminuria because creatinine alone can miss early kidney damage (KDIGO CKD Work Group, 2024). An eGFR of 72 mL/min/1.73 m² may be acceptable in many older adults, while urine albumin-creatinine ratio ≥30 mg/g deserves attention even when creatinine looks normal.

For a practical starting list, our guide to the most useful blood tests prioritizes markers that change management. I would rather see 15 well-tracked biomarkers over 120 random wellness markers with no plan.

When a change is probably normal noise

White blood cell count commonly ranges from about 4.0 to 11.0 x10⁹/L in adults, and a shift from 5.8 to 7.2 x10⁹/L after poor sleep or a mild cold is rarely meaningful by itself. Platelets can move within the 150–450 x10⁹/L range without implying a clotting disorder.

TSH is another troublemaker. A TSH of 3.8 mIU/L followed by 4.4 mIU/L may be less informative than the paired free T4, thyroid antibodies, medication timing, and whether biotin was taken; our guide on 異常結果の再検 gives realistic retest windows.

My rule in clinic is to avoid making a life decision from a single borderline result unless the marker is dangerous, diagnostic, or tied to symptoms. Two or 3 data points over 8–12 weeks often tell a cleaner story than one dramatic screenshot.

How AI should flag lab errors and impossible combinations

Pseudohyperkalemia is common enough that every tracking app should know it exists. Potassium above 6.0 mmol/L can be dangerous, but if the sample was hemolyzed and kidney function is stable, the safest interpretation is urgent verification rather than instant diagnosis.

Biotin can falsely lower TSH and falsely raise free T4 in some immunoassays, which may imitate hyperthyroidism. High-dose hair and nail supplements often contain 5,000–10,000 micrograms, far above ordinary dietary intake; our article on AIの検査エラー確認 ではよくある落とし穴を扱っています。.

Kantesti’s validation work also tests hyperdiagnosis traps, where the tempting answer is wrong because one result conflicts with the rest of the panel. The AI血液検査ベンチマーク explains how clinical reasoning is tested across specialties rather than only by marker lookup.

Privacy, security and regulation checks before you upload

A lab report can reveal pregnancy status, HIV testing, kidney disease, cancer markers, medication exposure, genetic clues, and family risk. That is why Kantesti Ltd, UK Company No. 17090423, operates with GDPR, HIPAA, ISO 27001, and CE Mark requirements rather than vague privacy promises.

Ask 5 questions before upload: where is my data stored, who can view it, can I delete it, is it used for model training, and how are family profiles separated? Our guide to 安全に保管するためのガイド gives a patient checklist for 2026.

Also read the legal documents, dull as that sounds. Kantesti’s ソフトウェアライセンスの条件 explain permitted use, limitations, and user responsibilities; in medical AI, the boring pages often contain the safety details.

Buyer checklist: choose the app that reduces confusion

A practical test is to upload 2 old reports from different labs and ask whether the app explains why some values cannot be compared cleanly. If it overreacts to every borderline flag, underreacts to potassium >6.0 mmol/L, or hides the original units, keep looking.

Kantesti now supports PDF and photo upload, trend analysis, family health risk, nutrition plans, and interpretation in 75+ languages across 127+ countries. You can try a real upload through our 無料の血液検査の見方 and see whether the output helps you ask better questions, not just collect more numbers.

Bottom line: the right app should make you calmer and better prepared for your clinician. If you want to understand how our team built this, read more about カンテスティ and our medical mission.

よくある質問