Skip to content

LOINC Tranche K Closing Ledger: Prenatal and Fetal Screening

Date: 2026-10-17 Scope: projects/LOINC_DIAGNOSTICS.md tranche K — chromosomal disorders and screening-programme panels, bounded by kb/disorders/ entries carrying a cfDNA/NIPT, maternal-serum, nuchal-translucency, or karyotype diagnostic row. PRs: #14630 (batch 1, 6 entries), #14633 (batch 2, 8 entries) — both open, neither merged as of this writing.

Entries examined and their outcome

Entry Outcome Code / reason
Down_syndrome Bound LOINC:77011-5 Fetal Chromosome 21 trisomy [Presence] based on Plasma cell-free DNA by Sequencing
Trisomy_18 Bound (split into two rows) LOINC:29770-5 Karyotype; LOINC:50580-0 generic multi-condition maternal screen panel (no triple/quad/penta specificity stated by the source)
Anencephaly Bound (split into two rows) Maternal Serum AFP Screening row: LOINC:48802-3 serum AFP panel. Amniotic Fluid AFP and Acetylcholinesterase row: LOINC:58735-2 amniotic AFP panel, LOINC:76478-7 amniotic acetylcholinesterase panel
Spina_Bifida_Cystica Bound LOINC:48802-3, LOINC:58735-2 (same two codes as Anencephaly, added to the entry's existing single combined row)
Hemolytic_Disease_of_the_Fetus_and_Newborn Bound LOINC:100670-9 RHD gene allele [Identifier] in Plasma cell-free DNA Nominal--fetus
47_XYY_Syndrome Bound (karyotype only) + citable absence (screening) LOINC:29770-5 (blood), LOINC:33773-3 (amniotic fluid) for the two confirmatory karyotype rows; no cfDNA screening code exists for the aneuploidy itself
22q11.2_Deletion_Syndrome Bound LOINC:75576-9 Fetal 22q11.2 deletion risk [Likelihood] — not the chromosome-21-specific code, and not the joint aneuploidy-and-22q11.2 panel, which asserts a test this row does not claim
15q11q13_Microduplication_Syndrome Citable absence no code for incidental maternal CNV detection during genome-wide NIPS (the row describes a maternal, not fetal, finding)
Thanatophoric_Dysplasia Citable absence no FGFR3 code specifies maternal plasma cell-free DNA as the specimen; only Blood/Tissue (postnatal) and Amniotic fluid/CVS (the invasive route the test exists to avoid) exist
45_X_46_XY_Mixed_Gonadal_Dysgenesis Citable absence no code for the 45,X/46,XY mosaic pattern itself, distinct from LOINC's simple Monosomy X risk family
Achondrogenesis_Type_IA Bound LOINC:33069-6 Fetal nuchal translucency measured by US — the entry reports an actual measured value (4.3 mm at 14 weeks)
Tonne-Kalscheuer_Syndrome Declined nuchal translucency named only inside one quoted case-report sentence, no dedicated row, no reported value
CHRNA1-associated_Fetal_Hypo-akinesia_Disorder_of_Prenatal_Onset Declined has a dedicated ultrasound diagnosis row, but nuchal translucency is named only categorically, with no number for the quantitative measurement code to attach to
Congenital_Isolated_ACTH_Deficiency Bound LOINC:LP15559-5 Estriol.unconjugated, the Part code — the entry's claim is qualitative ("isolated low" estriol), with no number or Multiple-of-the-median value cited, so the MoM family does not apply
Aflatoxin_Related_HCC Declined — collision "cell-free DNA" matched on a liquid-biopsy ctDNA assay (droplet digital PCR for somatic TP53 R249S in hepatocellular carcinoma), not prenatal cfDNA
Brain_Arteriovenous_Malformation Declined — collision "cell-free DNA" matched on somatic lesion genotyping from peripheral blood (a liquid-biopsy route to the lesion's own somatic driver mutation), not prenatal cfDNA

Across 16 entries examined, 9 carry at least one new LOINC binding, 4 are recorded as citable absences (3 full entries plus the screening half of 47_XYY_Syndrome, whose confirmatory karyotype rows are bound), and 4 decline outright with a stated reason — 2 for want of a performed measurement and 2 for the same acronym-collision shape found earlier in this campaign (99mTc-DPD scintigraphy matching "DPD"; here, liquid-biopsy ctDNA matching "cell-free DNA").

Finding 1: the AFP/estriol [Multiple of the median] family, unused this tranche

LOINC carries a family of prenatal-screening-specific codes distinct from the plain analyte concentration: [Multiple of the median], with adjusted, unadjusted, weight-adjusted, and (for some analytes) diabetes-adjusted variants, for both AFP and unconjugated estriol, in both serum and amniotic fluid. MoM is the reporting convention prenatal screening programmes actually use in practice — a bare "AFP panel - Serum or Plasma" panel code is identical whether the context is prenatal screening or tumour-marker monitoring.

None of the four entries in this tranche that cite AFP or estriol (Anencephaly, Spina_Bifida_Cystica, Trisomy_18, Congenital_Isolated_ACTH_Deficiency) states a MoM value or a numeric threshold (a "2.0" or "2.5 MoM" style cutoff). Each makes a qualitative claim ("elevated," "isolated low") without a number. Per the rule established on this tranche — bind the MoM code only when the entry's own row states a MoM value or threshold — the panel/Part codes stand in all four entries, and no interpretation_bands were added.

This is a lead, not a defect. If a future pass sources a specific screening-programme cutoff for any of these four entries (the quad-screen MoM threshold for Down syndrome, trisomy 18, or an open neural tube defect is widely published), the MoM-family code is the correct rebind and the cutoff would support a genuinely sourced interpretation_bands entry — something this tranche never had the citation to add.

Finding 2: four citable LOINC absences, dated

Each of the following was confirmed by a fully-enumerated search (every result read, not a truncated page) against the Monarch KG search API (https://api-v3.monarchinitiative.org/v3/api/search, which carries the LOINC table) on 2026-10-17. LOINC adds cfDNA screening codes release to release, so a later search could find a code where this one did not; these are dated for exactly that reason, and are recorded as a note on the affected row rather than left as a silent gap.

  • 15q11q13_Microduplication_Syndrome — "15q11 duplication cell-free DNA", "genome-wide cell-free DNA", and "microdeletion microduplication cell-free DNA" each return zero results of any kind (total=0 for all three, nothing to enumerate).
  • Thanatophoric_Dysplasia — "FGFR3 cell-free DNA" returns zero results (total=0). The broader "FGFR3" search returns 277 results, all read to the end; the only FGFR3 molecular-testing LOINC codes in that set specify Blood or Tissue (postnatal) or Amniotic fluid/Chorionic villus sample (the invasive route the cfDNA test exists to avoid) as the specimen — none names cell-free DNA.
  • 45_X_46_XY_Mixed_Gonadal_Dysgenesis — "45,X/46,XY mosaicism cell-free DNA" and "sex chromosome mosaicism cell-free DNA" each return zero results (total=0 for both).
  • 47_XYY_Syndrome (the screening half of the entry; the confirmatory karyotype rows are bound) — "XYY cell-free DNA", "XXY cell-free DNA", "XXX cell-free DNA", and "fetal sex chromosome aneuploidy" each return zero results (total=0 for all four).

A pattern across three of these four: LOINC's cfDNA aneuploidy family covers the autosomal trisomies (21, 18, 13) and simple Monosomy X. It does not cover microdeletion/microduplication panels beyond 22q11.2 (which does have its own code family, bound on 22q11.2_Deletion_Syndrome), single-gene skeletal dysplasias, or the extra-sex-chromosome aneuploidies (47,XXY; 47,XXX; 47,XYY). A candidate pool built on "this entry mentions cfDNA/NIPT" will keep surfacing entries where the technology is real but LOINC has not (yet) coded that specific target.

Open design decision, not performed: bulk karyotype binding

Roughly 140 kb/disorders/ entries name karyotype somewhere in their text. LOINC:29770-5 (Karyotype [Identifier] in Blood or Tissue Nominal) and LOINC:33773-3 (Karyotype [Identifier] in Amniotic fluid Nominal) — both bound individually, on their own merits, to specific rows in this tranche — would attach mechanically to nearly all of them: find a diagnosis row naming karyotype, read its specimen (blood/tissue vs. amniotic fluid/CVS), bind the matching code. The mechanical part of that is almost entirely pattern matching, not judgment.

That is exactly why it should not simply be dispatched as a tranche batch. Every binding this campaign has made has rested on reading one entry's own claim and choosing among a family of candidate codes for it — the AFP panel-vs-MoM question, the VKORC1-vs-CYP2C9-panel question, the karyotype blood-vs-amniotic-fluid question this tranche's own entries raised. A sweep that binds the same two codes to ~140 entries by pattern match is a different kind of act: it would be the first campaign binding made by rule rather than by reading each entry, and it would do it at a scale no other single PR in this campaign has approached.

The choice: either (a) treat karyotype as in-scope for a future dedicated pass that still reads each entry's row individually — slower, but consistent with every binding this campaign has made so far — or (b) treat it as a deliberately different, rule-based operation and decide separately whether to run it as a bulk mechanical pass with its own review standard.

Recommendation: (a). The two karyotype bindings made in this tranche each needed a real decision — which specimen, whether the row described a screen or a confirmation, whether a more specific code existed for the disease's own named test. A bulk pass by pattern match would get the specimen choice right by coincidence as often as by reading, and Karyotype Analysis is exactly the generic row name a pattern match would latch onto without checking whether the entry's own karyotype claim says blood, amniotic fluid, or something the pattern can't tell apart (a postmortem sample, a buccal swab, a tissue biopsy). If the 140-entry number makes a manual pass impractical, that argues for narrowing the pattern (a karyotype row with its own presence: value naming a specific abnormal result, the shape every binding in this tranche actually had) rather than lowering the bar on what "binding a code" means everywhere else in this campaign.

What landed, in one line

9 of 16 examined entries carry at least one new LOINC binding across PRs

14630 and #14633 (both open); 4 are recorded as citable absences dated

2026-10-17; 4 decline outright, 2 of them the same acronym-collision shape (cell-free DNA / ctDNA) already seen once in this campaign as 99mTc-DPD / dihydropyrimidine dehydrogenase. The MoM family and the karyotype bulk-binding question are recorded above for whoever next touches a prenatal-screening or chromosomal entry.