Is the Manufacturer's Range Right for Our Lab? Verifying the Serum-Creatinine Reference Interval with CLSI EP28

Every serum-creatinine report in a hospital is judged against a reference interval — usually the range printed in the analyser manufacturer's insert. That range decides who is called normal and whose result triggers an eGFR calculation and a chronic-kidney-disease workup. Yet few Indian laboratories ever check whether the inherited range actually fits their instrument, their method (kinetic Jaffe or enzymatic) and their population, even though CLSI EP28, the international standard, defines exactly how such a check is done.

This two-member B.Sc MLT project performs that verification for serum creatinine in the clinical biochemistry laboratory of a 300-bed hospital (fictional: Nandini Multispeciality Hospital). First the analytical system is proved fit: within-run imprecision (20 replicates each of two pooled-serum levels) and daily Levey–Jennings QC with Westgard multirules throughout the study. Then follows the EP28 transference study: 20 healthy adult men and 20 healthy adult women (volunteers, consented, screened by a short exclusion questionnaire) each provide one sample, tested against the manufacturer's interval — no more than 2 of 20 results outside verifies transference for each partition.

As a supporting, fully anonymised check, six months of outpatient creatinine results (ages 18–60) are exported from the LIS without any identifiers, extreme values are removed by Tukey fences, and the 2.5th–97.5th percentiles per sex are compared with the transference outcome. Finally, the clinical consequence is made concrete by computing eGFR (CKD-EPI 2021) at the interval's boundary values. Ethics: IEC approval, written consent for volunteers, aggregate anonymised data only.

The study laboratory reports serum creatinine against the reference interval printed in the analyser manufacturer's insert. That interval was derived on a different instrument generation, a different method variant and a different population, yet no verification has ever been documented on the installed system. The interval directly determines which outpatients are labelled normal and which proceed to eGFR-based chronic-kidney-disease classification, so a misfit is not a cosmetic issue: a range even slightly too wide can mask early renal decline, and one too narrow generates false concern and repeat testing.

CLSI EP28 defines the minimal professional check — transference verification with 20 healthy subjects per partition — and modern practice adds an indirect percentile analysis of the laboratory's own anonymised historical data. The laboratory needs this documented verification for serum creatinine, together with the QC record that shows the analytical system was in control while it was performed, so that the interval it reports can be defended in accreditation terms and in clinical terms.

Abstract

A laboratory-based verification study will test whether the manufacturer's serum-creatinine reference interval is valid on the laboratory's own chemistry analyser, following the transference protocol of CLSI EP28. After confirming analytical performance (within-run imprecision of 20 replicates at two pooled-serum levels, and daily Levey–Jennings QC interpreted with Westgard multirules), 40 healthy adult volunteers (20 men, 20 women, ages 18–60) will provide single samples under written consent. For each sex partition, no more than 2 of 20 results outside the manufacturer's interval will verify transference. A supporting indirect analysis of six months of anonymised outpatient creatinine results from the LIS (Tukey outlier removal, 2.5th–97.5th percentiles per sex) will be compared with the direct outcome, and the clinical impact of any boundary shift will be demonstrated by computing eGFR with the CKD-EPI 2021 equation at the interval limits.

Introduction

Creatinine is the workhorse renal marker: it drives the eGFR reported alongside every result and therefore the CKD classification that follows. The eGFR equations assume a correctly calibrated creatinine assay judged against a valid reference interval. A reference interval that is too wide hides early renal impairment; one that is too narrow manufactures worry. Reference intervals are population- and method-dependent — muscle mass, diet, hydration and assay chemistry (Jaffe's picrate reaction is vulnerable to non-creatinine chromogens; enzymatic methods differ again) all shift the distribution.

Literature gap

Large Indian hospital studies have established local creatinine intervals, and laboratory textbooks describe EP28 in principle. What is rare is a small-lab, protocol-first verification study: most routine laboratories inherit insert ranges without the documented 20-subject transference check that CLSI EP28 treats as the minimum professional standard, and student projects rarely combine the direct check with an indirect (data-mining) percentile analysis of the laboratory's own historical results. This project does both, at a scale a B.Sc MLT team can actually complete.

Proposed work

• Prove the analytical system: within-run imprecision at two levels; documented Westgard-controlled QC for the study period.
• Run the EP28 transference study: 20 healthy volunteers per sex partition; accept if ≤ 2 of 20 fall outside the manufacturer's interval; if either partition fails, escalate to a 120-subject local establishment (outlined as contingency).
• Cross-check with an indirect analysis of six months of anonymised outpatient results (Tukey screening, percentile estimation per sex).
• Quantify the clinical stakes: eGFR by CKD-EPI 2021 at the interval boundaries.

Feasibility

• Technical: the analyser, QC material and LIS are already in routine use; phlebotomy and sample handling are core MLT skills.
• Economic: no reagent beyond routine consumption; volunteers are staff and students donating ~3 mL blood each; consumables cost a few hundred rupees.
• Time: volunteer recruitment and runs fit inside eight weeks of the 16-week project; the indirect analysis is purely computational.
• Ethical: IEC approval with consent for the volunteer arm; the LIS arm uses anonymised, aggregate results with no patient identifiers — no additional blood is drawn from any patient.

In scope

• Serum creatinine on the laboratory's routine chemistry analyser, one method (as installed).
• Direct transference: 40 healthy adult volunteers (20 per sex, 18–60 years), single sample each.
• Indirect analysis: six months of anonymised outpatient creatinine results for ages 18–60, partitioned by sex.
• Daily internal QC and within-run imprecision during the study window.
• eGFR (CKD-EPI 2021) computations at the interval boundaries as a clinical-impact illustration.

Out of scope

• Establishing a new reference interval from 120 subjects per partition (outlined as contingency only, not executed unless a partition fails).
• Other renal analytes (urea, uric acid, cystatin C) and urine albumin–creatinine ratio.
• Paediatric, geriatric (60+) and pregnancy partitions.
• Method comparison between Jaffe and enzymatic creatinine (proposed as future work).
• Any patient-level diagnosis review or follow-up of individuals flagged by the indirect data.

Study design: analytical verification study with a direct EP28 transference arm and a retrospective, fully anonymised indirect arm.

Setting: clinical biochemistry laboratory of a 300-bed hospital (fictional: Nandini Multispeciality Hospital); the routine creatinine method (kinetic Jaffe or enzymatic, as installed) is used throughout without reagent-lot change mid-study where avoidable.

Phase 1 — Analytical fitness (weeks 3–4)

• Within-run imprecision: 20 replicate determinations on each of two pooled-serum levels (approximate medical decision levels); report mean, SD, CV%; accept CV within the method's stated imprecision goal.
• Establish QC baselines from at least 20 prior runs of the current lot; plot daily two-level QC on Levey–Jennings charts and interpret with Westgard multirules (1-3s, 2-2s, R-4s, 4-1s, 10x); runs violating rejection rules are corrected before any study sample is run, and every violation plus corrective action is logged.

Phase 2 — Direct transference, CLSI EP28 (weeks 5–9)

• Volunteers: 20 healthy adult men and 20 healthy adult women, 18–60 years, recruited from hospital and college staff/students. Exclusion (short questionnaire): known kidney disease or hypertension, diabetes, pregnancy, regular medication other than vitamins, acute illness or fever in the past two weeks, extreme athletic training, prolonged fasting or high-protein supplementation before sampling.
• Consent and ethics: IEC approval before recruitment; written informed consent; each volunteer donates one ~3 mL clotted-venous sample; results outside the interval are communicated with advice to consult a physician.
• Procedure: samples centrifuged and run within the routine turnaround window; each sample analysed once (EP28 allows single determination); two samples per partition run in a single analytical session alongside QC to bracket session validity.
• Criterion: per partition, count results outside the manufacturer's 95% interval; ≤ 2 of 20 outside verifies transference; 3 or more outside fails, triggering the contingency plan (120-subject local establishment by nonparametric percentiles, outsourced to a follow-on study and outlined in the report).

Phase 3 — Indirect check on LIS data (weeks 10–12)

• Export six months of outpatient creatinine results, ages 18–60, with all identifiers removed at export (no names, no UHIDs, no dates of birth — age band and sex only).
• Partition by sex; remove outliers by Tukey fences (below Q1 − 1.5×IQR or above Q3 + 1.5×IQR); report the removed proportion.
• Estimate the 2.5th and 97.5th percentiles (with bootstrap 90% confidence intervals in jamovi) per sex; compare with the manufacturer's interval and the direct result.

Phase 4 — Interpretation and clinical impact (weeks 13–14)

• Table the three sources of truth: manufacturer insert, direct transference, indirect percentiles.
• Compute eGFR by CKD-EPI 2021 (NKF reference) at the interval's lower and upper creatinine boundaries for representative ages (40 and 60, both sexes) to show how a boundary shift moves CKD classification.
• Draft the recommendation: retain, flag, or initiate re-establishment.

Statistics

Descriptives (mean, SD, median, IQR, percentiles) in Excel and jamovi/SPSS; CV% for imprecision; binomial assessment of out-of-range counts; percentile estimation with bootstrap confidence intervals; no inferential test beyond these is claimed, in keeping with EP28's design philosophy — every statistical choice is justified in the report, as examiners require.

Ethics summary

IEC approval and waiver policy documented per ICMR national ethical guidelines; written consent for all volunteers; the LIS arm is retrospective, anonymised aggregate data with no patient identifiers; no extra blood is drawn from any patient; the study's flagged volunteers are counselled to consult a physician.

The study has two evidence streams — a direct volunteer arm and an indirect anonymised-data arm — that must converge on the same verdict under continuous QC oversight. The figure belongs in the Methodology chapter.

flowchart TD
  A["Daily two-level QC: Levey-Jennings charts"] --> B{"Westgard rejection rule violated?"}
  B -->|Yes| C["Corrective action, re-run controls"]
  C --> A
  B -->|No| D["Analytical system in control"]
  E["Within-run imprecision: 20 replicates x 2 pooled levels"] --> F{"CV% within goal?"}
  F -->|No| G["Investigate method before proceeding"]
  F -->|Yes| D
  D --> H["EP28 transference: 20 male + 20 female volunteers"]
  I["IEC approval and written consent"] --> H
  H --> J{"Per partition: >2 of 20 outside interval?"}
  J -->|No: <=2 outside| K["Transference verified for that partition"]
  J -->|Yes: 3+ outside| L["Transference fails - plan 120-subject local establishment"]
  M["LIS export: 6 months outpatient creatinine, 18-60, no identifiers"] --> N["Partition by sex, Tukey outlier removal"]
  N --> O["2.5th-97.5th percentiles with bootstrap CI"]
  K --> P["Compare: insert vs transference vs indirect percentiles"]
  O --> P
  L --> P
  P --> Q{"Intervals agree?"}
  Q -->|Yes| R["Retain interval, document verification"]
  Q -->|No| S["Flag interval, recommend re-establishment"]
  R --> T["eGFR impact check: CKD-EPI 2021 at boundary values"]
  S --> T

Decision rule at a glance

The indirect percentiles never overrule the direct study — they corroborate or question it, and disagreements are reported honestly with possible causes (outpatient mix, outlier handling, seasonal hydration).

• Extend to a full 120-subject local establishment study if any partition fails transference, using the nonparametric method of CLSI EP28.
• Add a Jaffe-versus-enzymatic creatinine method comparison on the same anonymised samples — the natural companion study to interval verification.
• Add cystatin C and the eGFR-cystatin C equations for patients where creatinine is unreliable (low muscle mass, extremes of age).
• Partition by age bands (18–40, 41–60) and study seasonal hydration effects in a hot-climate population.
• Build a verification template (protocol, consent form, Excel workbook) that the laboratory can reuse annually for every analyte — turning one student project into a standing accreditation asset.