SNOMED CT
Content Improvement Project

Inception phase

Project ID: artf6245 IHTSDO-409 - Getting issue details... STATUS
Topic: SNOMED CT to ICD-O Mapping

Date: July 2016

Version 1.1

Amendment History

Review Timetable

© International Health Terminology Standards Development Organisation 2012. All rights reserved.
SNOMED CT® was originally created by the College of American Pathologists.
This document forms part of the International Release of SNOMED CT® distributed by the International Health Terminology Standards Development Organisation (IHTSDO), and is subject to the IHTSDO's SNOMED CT® Affiliate Licence. Details of the SNOMED CT® Affiliate Licence may be found at http://www.ihtsdo.org/our-standards/licensing/.
No part of this document may be reproduced or transmitted in any form or by any means, or stored in any kind of retrieval system, except by an Affiliate of the IHTSDO in accordance with the SNOMED CT® Affiliate Licence. Any modification of this document (including without limitation the removal or modification of this notice) is prohibited without the express written permission of the IHTSDO.
Any copy of this document that is not obtained directly from the IHTSDO [or a Member of the IHTSDO] is not controlled by the IHTSDO, and may have been modified and may be out of date. Any recipient of this document who has received it by other means is encouraged to obtain a copy directly from the IHTSDO [or a Member of the IHTSDO. Details of the Members of the IHTSDO may be found at http://www.ihtsdo.org/members/].

Table of Contents
1 Glossary
1.1 Domain Terms
2 Introduction
2.1 Purpose
2.2 Audience
3 Detailed Problem Statement
3.1 Background
3.2 Statement of Problem
3.2.1 Summary of problem as reported
3.2.2 Summary of requested solution
3.2.3 Detailed analysis of reported problem
3.2.4 Subsidiary and interrelated problems
3.2.5 Stakeholders
3.3 Risks / Benefits
3.3.1 Risks of not addressing the problem
3.3.2 Risks of addressing the problem
4 Requirements: criteria for success/completion
4.1 Criteria for success/completion
4.2 Use cases
4.2.1 Use case 1
4.2.2 Use case 2…
4.3 Test Cases
5 Envision Possible Technical Approaches
5.1 Indicative Technical Approaches
5.1.1 Approach One
5.1.2 Approach Two
6 Indicative Project Plan for Elaboration Phase
6.1 Scope
6.2 Skills required
6.2.1 Solution Specification (Elaboration)
6.2.2 Implementation
6.2.3 Preventing recurrence of problem
6.3 Project size, lifecycle, duration and resource requirements
6.4 Deployment
7 Appendices
7.1 Appendix One : Related user requests

Glossary

Domain Terms

Introduction

Purpose

The purpose of this project is to consider the scope and nature of the work required to revise and update the mapping / data interoperability relationship between SNOMED CT and ICD-O Version 3.
SNOMED CT projects transition from Inception Phase >> Elaboration Phase >> Construction Phase >> Transition Phase. This document describes the Inception Phase. The elaboration phase, in which one or more technical solutions may be developed and tested, may result in more than one document.
The purpose of the Inception Phase is to agree with stakeholders the detail of the problem to be addressed and its scope boundaries. The resulting problem description must also be of sufficient detail such that the size and impact of any resolution might have on the terminology as a whole and its users can be understood.
Subject to adequate review by stakeholders and subsequent revision, the inception phase document becomes the primary input to the Elaboration Phase of the project, in which the potential solutions are considered.

Audience

The audience for this document includes all standards terminology leaders, implementers and users but is especially targeted at those stakeholders from the histopathology and oncology domains, including national Cancer Registries.

Identification of stakeholders

Likely stakeholders include:
World Health Organisation (formand@iarc.fr ; jakobr@who.int)
AJCC American Joint Cancer Committee (ajcc@facs.org)
UICC Union Internationale Contre le Cancer (info@uicc.org)
FIGO International Federation of Gynaecology and Obstetrics
NCI Natonal Cancer Institute (Peggy Adamo seertraining@imsweb.com)
International Associated Of Cancer Registries (iacr@iarc.fr)
Pathology laboratories (cascade through NRCs)
IHTSDO iPALM Special Interest Group
Laboratory Information and Management System (LIMS) vendors
Clinical oncologists

Input from stakeholders

Consulting with the stakeholders listed above will necessarily be a slow process. Because of the number of standards bodies involved, it may be more appropriate for such a consultation to be driven at a higher level than an individual Consultant Terminologist, e.g. the IHTSDO Head of Terminology could request the IHTSDO-WHO Joint Advisory Group to consider the full set of representational issues in the oncology domain and then lead a coordinated international effort to resolve them.

Degree of consensus on the statement of problem

Not known at this stage (stakeholder consultation still pending).

Statement of the problem or need

Summary of problem or need, as reported

The issue as originally logged to the content tracker in March 2010 was expressed only in terms of an outline solution (reproduced below in s3.2). The underlying problem(s) was, therefore, only implied but appears to be:

1. the quality of the current mapping (mirroring) relationship between SNOMED CT and ICD-O must be improved, mostly to correct omissions
2. the collaboration process by which that mapping/mirror is maintained should be strengthened
3. the fundamental nature of the mapping/mirroring relationship could be reconsidered.

A supplementary note posted to the tracker item later that year (August 2010) provides further context to the supposed problem:

 With the signing of the agreement with WHO, the first meeting of the Joint Coordination Group identified coordination of ICD-O with SNOMED morphology as one of the 5 work items that the cooperative work will be addressing. 

This note refers to the joint agreement signed on 25^th July 2010 between IHTSDO and WHO. The agreement commits both parties to a sustained collaborationtoward the ultimate goal of achieving a harmonized integration between SNOMED and all WHO-FIC classifications. The continuing collaboration is today managed by a Joint Advisory Group (JAG).

When the collaboration started, back in 2010, the JAG ultimately identified two potential work streams to be their highest priority: improving the map from SNOMED CT to ICD-10, and developing a common ontological framework to underpin both SNOMED CT and ICD-11 so that, ultimately, ICD-11 might be formally expressed and maintained as a 'view' on SNOMED CT rather than as an entirely different system.

However the JAG clearly also considered (but graded of lower priority) the potential for a collaborative workstream tasked with re-examining the relationship between ICD-O and SNOMED CT: both the current relationship, and how they should interrelate in future. The solution outlined in the original tracker entry (below) suggests a working assumption that SNOMED CT should continue to closely mirror the ICD-O-3 Morphology and Topography chapters.

Summary of requested solution

The tracker item outlines a possible four-part solution:

1. Updating the mapping of SNOMED CT Body structures to ICD-O-3 topographies.

2. Updating the mapping of SNOMED CT cancer related (M-8 and M-9 SNOMEDIds) morphologic abnormality concepts to ICD-O-3 where appropriate. For example, for several years there were_ two morphology entities that existed in ICD-O (as published by WHO) but for which no corresponding concept existed in SNOMED CT with the analogous SNOMEDID (both omissions were corrected in the July 2011 release of SNOMED CT):

8155/1 Vipoma, NOS and 8935/0 Stromal tumor, benign
Additionally, out of 4,528 concepts in the descent of SNOMED's 400177003|Neoplasm and/or hamartoma (morphologic abnormality)|, at least 876 (19.3%) have a SNOMEDID that did not appear to be a valid ICD-O morphology code analog.

1. Identifying morphologic abnormalities in SNOMED CT which have non-synonymous descriptions from ICD-O-3. Flagging those descriptions as non-synonymous and creating them as new subtype or sibling concepts with appropriate ICD-O-3 mappings. e.g. apparent duplicate mappings of: vipoma (447643008 -> 8155/1, vs 31131002 -> 8155/3)

proliferating trichilemmal tumor (128638007 -> 8103/0, vs 446023005 -> 8103/1).
These may be instances of synonyms needing to be retired as inappropriate. Is "vipoma" a valid synonym of "vipoma, malignant"? Is "proliferating trichilemmal tumor" a valid synonym of "pilar tumor"?

1. Appropriate FSNs for content with M-8 and M-9 SNOMEDIds so that the FSNs mirror the intended ICD-O meaning.

Statement of problem as understood

Some Topography codes in ICDO-3 are not currently matched by a semantically equivalent SNOMED CT code or expression, and/or the equivalence is not always published.

Some Morphology codes in ICDO-3 are not currently matched by a semantically equivalent SNOMED CT code or expression, and/or the equivalence is not always published.

A significant delay often occurs occur between ICDO-3 or SNOMED publishing new codes, and the corresponding code equivalence tables being appropriately updated and published by the other party.

For some ICD-O codes (especially morphology codes), the SNOMED CT concept stated as their semantic equivalent may have current descriptions that are (a) not genuine mutual synonyms of one another and/or (b) do not appropriately mirror the meaning, if not also the actual text, of the designated ICD-O equivalent.

Because of the above failings, data originally captured using either SNOMED CT or ICD-O can not always freely interoperate by means of unambiguous, comprehensive, fully maintained and professionally tables of equivalence.

A more comprehensive and timely alignment of content is desirable, to be published as

• a many:one unidirectional SNOMED-to-ICD-O map for Topography
• a one:one bidirectional map for Morphology

Detailed analysis of reported problem, including background

Background

ICD-O

ICD-O was first published by the World Health Organisation in 1976. It is the standard tool used globally for coding neoplastic diagnoses (whether benign or malignant), both by pathology laboratories and by the tumour and cancer registries they report to. Version 2 followed in 1990, and version 3 in 2001.
Since its first version, ICD-O has been conceived as a relatively crude compositional scheme: neoplastic diseases are independently coded along two primary axes (topography and morphology) of which one (morphology) may be further characterized along two further subsidiary axes (behavior and grade). A full ICD-O encoding, therefore, comprises a concatenation of four subcodes:

• 2-3 digit topography code (anatomical location of primary tumour site; revised in ICD-O version 2. Code Range C00-C80.9; there are currently 400 distinct site codes in ICD-O-3)
• 4 digit histology / morphology code (Code range: M8000-M9989; originally only histopathological type but, since version 3, also includes other techniques for tumour typing such as cytogenetic markers. There are 765 distinct morphology codes in ICD-O-2)

• 1 digit behaviour code (modifies morphology code)

  Code	Behaviour
  /0	Benign
  /1	Uncertain whether benign or malignant; Borderline malignancy , Low malignant potential , Uncertain malignant potential
  /2	Carcinoma in situ Intraepithelial, Noninfiltrating, Noninvasive
  /3	Malignant, primary site
  /6*	Malignant, metastatic site Malignant, secondary site
  /9*	Malignant, uncertain whether primary or metastatic site

_{*Not used by cancer registries (used by some pathologists in some parts of the world)}

• 1 digit 'aggression' code (degree of differentiation or grade of the tumour; also modifies the morphology code)

In principle, therefore, ICD-O-3 allows for in excess of 9.1 million distinct combinations of a morphology, topography, behaviour and grade code. In practice, of course, many combinations are impossible or highly improbable (e.g. M-8631/01 C70.1 : Benign, well differentiated Sertoli-Leydig cell tumour of frontal lobe). Within its overarching compositional framework, therefore, ICD-O-3 is in practice more commonly delivered by WHO as a list of 2271 well known diagnostic labels, each of which is given a centrally assigned, normative, precoordinated morphology and behaviour code (but not a Topography code). Thus, if the expert consensus is that the clinical entity called 'Burkitt's Lymphoma' is always malignant, then ICD-O provides:

9687/3 Burkitt Lymphoma 

However, if such a diagnosis is confirmed histologically (as a 9687) but the subsequently observed behaviour of the specific tumour instance is other than the usual malignant, it is permissible to record (for example) 9687/2 even though this code was absent from the official distribution.
Some have, therefore, attempted to further enumerate the clinically plausible cross-product of Morphologies with Topographies and also to extend the set of recognized synonyms. These endeavours typically arrive at an enumerated list of some 9000 Topography+Morphology+Behaviour+Term combinations for the clinically plausible oncology domain. See e.g. http://www.seer.cancer.gov/icd-o-3/

Changes in ICD-O

Like all clinical terminologies, ICD-O has had to change its content as medical knowledge has evolved and the expert consensus changes. The most recent content changes, from version 2 to 3, included:
Some tumours previously graded as borderline (behavior code:1) are now reclassified as malignant (code:3). Similarly, some previously malignant (3) are now officially benign (1). For example:

203 new 4-digit Morphology codes not previously in ICD-O-2 were added to ICD-O-3
97 4-digit Morphology codes previously in ICD-O-2 were removed from ICD-O-3
1032 precoordinated 5-digit Morphology+Behaviour code combinations not previously in ICD-O-2 were added in ICD-O-3
849 precoordinated 5-digit Morphology+Behaviour code combinations previously found in ICD-O-2 are removed from ICD-O-3

More acronyms are now recognized as synonym terms; some terms are now marked obsolete.
Some terms previously present with UK English spelling (e.g. 'Polycythaemia rubra vera ') have been removed in favour of a US English spelling (e.g. 'Polycythemia rubra vera')
The classification of haematological diseases has been significantly re-worked, grouping them according to the REAL classification(ie along cell lines rather than on gross morphological features) and with some terms offering detailed differentiation by cytogenetic markers:

9866/3 = acute promyelocytic leukaemia t(15;17)(q22;q11-12)
=translocation (t) of material from long arm (q) of chromosome 15, region 22 with material from the long arm (q) of chromosome 17 between region 11 and 12

Some haematological diseases are also now described by molecular abnormalities:

9866/3 = acute promyelocytic leukaemia t(15;17)(q22;q11-12) or acute promyelocytic leukaemia PML/RAR-alpha

(Mis-)usage of ICD-O, and ICD-O internal inconsistencies

Although designed and intended to be used as a compositional system (ie where a full ICD-O encoding should always contain both a morphology and a topography code as a minimum) in practice many laboratory sites omit the Topography coding component altogether, instead opting to record only the Morphology element. This is perhaps understandable when many histopathological morphologies can either only ever arise from one anatomical locus, or so very rarely arise anywhere else that the default assumption is very rarely wrong. For example, the anatomical location of the primary is somewhat redundant to state if the morphology is given by any of:

In other cases the ICD-O morphology code itself appears to already explicitly specify the locus, such that the additional use of a Topography code can appear to be genuinely entirely redundant:

SNOMED and ICD-O

In common with several other clinical terminologies in use in the 1990s, and also probably because of SNOMED's origins as SNOP (a coding scheme specifically for histopathology), SNOMED sought to facilitate its crossmapping to ICD-O.
Being itself an early compositional system, and sharing ICD-Os ontological view that many diseases could be modeled as 'some histopathological phenomenon at some anatomical site', legacy versions of SNOMED (SNOMED 3 International, SNOMED RT) chose to closely mirror the ICD-O classification. In particular, the M-8000 to M-9989 codes in the morphology axis of SNOMED 3 were, by agreement with WHO, engineered to be a straight copy of the morphology classification in ICD-O version 2. This mirroring extended to the codestring level: the ICD-O codestring for a given Morphology-plus-Grade code combination could – originally - be trivially lexically transformed into the exactly corresponding SNOMED RT code and vice versa: ICD-O-2 M-ABDC/E becomes M-ABCDE in SNOMED-3 International, where E is the behaviour code element of the ICD-O encoding.

However, changes in SNOMED from 1993 onward - culminating in SNOMED-RT - led to identifier and content incompatibilities, particularly for non-neoplastic lesions. As a result, the electronic ICD-O-3 release files from WHO no longer includes explicit maps to SNOMED-RT codes. Since 2000, the mapping relationship between ICD-O and SNOMED CT has diverged further.

For example, the morphology code changes that occurred between ICD-O versions 2 and 3 were handled in SNOMED-CT by retiring those SNOMED codes corresponding to the (now deprecated) ICD-O-2 codes, then creating new SNOMED concepts for the new ICD-O-3 codes, and finally (where appropriate) inserting history relationships to link one to the other. Thus, for example, to reflect ICD-O-3's reclassification of PCV as a malignant condition, the SNOMED code:

31569001|Polycythemia vera (morphologic abnormality)| SNOMEDID = M-99501

…is now retired, and REPLACED_BY:

128841001|Polycythemia vera (morphologic abnormality)| SNOMEDID = M-99503

Current encoding of map from SNOMED CT to ICD-O Topography

In the same way that the ICD-O-3 release files no longer include an explicit map to SNOMED CT, so the core SNOMED CT table (sct1_concept) no longer guarantees either that all ICD-O-3 Topography or Morphology codes that currently exist will also appear somewhere in the SNOMEDID column or, where they do, that they're necessarily attached to the right underlying concept.

For this reason, it is no longer valid to crossmap to ICD-O by naïve code lookup on the SNOMEDID element of the sct1_concept table. This technique will accordingly become essentially impossible to use on RF2 distributions of SNOMED CT, within which there is no SNOMEDID column.

Instead, the International Edition of SNOMED CT (in either RF1 and RF2 release format) now contains a single crossmap table (mapSetID 102041) which encodes a mapping between 22,763 SNOMED CT anatomy concepts and 287 ICD-O-3 Topography codes, and between 1191 SNOMED CT Morphological Abnormality concepts and 1122 ICD-O-3 Morphology codes. This is a unidirectional map; it doesn't support reverse mapping from an ICD-O topography code to SNOMED CT. The map topology is, in fact, many:many rather than a strict many:one map: a minority (n=136) of SNOMED anatomy concepts have more than one possible ICD-O map:

Metastatic grading in ICD-O and SNOMED CT

The ICD-O Grade component value '6' (metastatic; secondary) isn't reliably used by cancer registries, mainly because they usually encode the nature of the condition only at the time of initial presentation and diagnosis and so often before the condition has metastasised or is clinically recognized to have done so. The central registry encoding of the disease is not always changed when metastasis is later confirmed. For different reasons, the 'metastatic behavior' element is also only variably used by histopathology laboratories across the world. As a result, sets of precoordinated ICD-O codes – including, for example, the current NCI Site/Histology validation list (at http://www.seer.cancer.gov/icd-o-3/) – do not typically contain many (or, sometimes, any) grade=6 combinations such as would be required to fully encode all metastatic disease.

By contrast, SNOMED-CT does contain some precoordinated concepts for metastatic disease, although the number is currently small. As a consequence, several SNOMED CT users have independently submitted requests to modestly expand the number of precoordinated SNOMED codes corresponding to specifically metastatic morphologies, in order to better support the coding of patients with metastatic disease. Other users have noted that, for those very few metastatic tumour morphology concepts that do already exist in SNOMED, the SNOMEDID codestring is not always a well-formed ICD-O code: they don't always end with '6'.

Leaving these content enumeration and alignment issues aside, however, a deeper ontological problem exists within the SNOMED CT content relating to metastatic disease: the adjective 'metastatic' is itself potentially ambiguous. For example, the term 'metastatic breast cancer' may indicate:

either a neoplastic disease process (morphology) characterized by a primary breast tumour with the capacity to undergo metastasis but where no distant metastasis has yet been confirmed
or a neoplastic disease process (morphology) characterized by a primary breast tumour that now, or at some point in the past, has actually undergone confirmed distant metastasis to at least one secondary site
or a discrete secondary tumour (morphology plus site) distant from the primary breast site

 
In the context of any potential mapping to ICD-O, it must be clarified which of these meanings overlaps with the meaning of the Behaviour=6 code in ICD-O-3. This unresolved issue and its other consequences are further discussed in the Inception document to artf221501.

Closed vs Open World issues

ICD-O presents as a 'closed world' taxonomy. As in most statistical classifications,
list closure is achieved principally by means of the '<morphology> not otherwise specified' and 'Other specified <anatomy>' constructs. However, the interpretation of codes such as:

C32.9 Larynx, NOS
C63.7 Other specified parts of male genital organs
8772/3 Spindle cell melanoma, NOS
8825/1 Myofibroblastic tumor, NOS

 
…is only possible if you know what other codes were available to the coder but were not selected. For example, in the case of an individual patient coded to C63.7 (above), we know that the anatomical location was specified but must inspect the range of codes available elsewhere in ICD-O (below) in order to determine that it was not the penis, prostate, testis, epididymis, spermatic cord or scrotum:

C60-C63 MALE GENITAL ORGANS

         C60 PENIS

         C61 PROSTATE GLAND

         C62 TESTIS

         C63 OTHER AND UNSPECIFIED MALE GENITAL ORGANS

               C63.0 Epididymis

               C63.1 Spermatic cord

               C63.2 Scrotum, NOS

               C63.7 Other specified parts of male genital organs

               C63.8 Overlapping lesion of male genital organs Note:

               C63.9 Male genital organs, NOS

 
This interpretation is specific to the particular version of ICD-O being used; the meaning of C63.7 would change if a code [C63.3 Bulbourethral gland] were added to ICD-O at some later date, because such an addition changes the closed world list against which the negation in 'not otherwise specified' is to be evaluated.

By contrast, the SNOMED CT description logic is 'open world'. As a direct consequence, the true specific meaning of 'NOS' codes such as C63.7 can never be represented or reliably reasoned over within SNOMED CT – partly because the logic itself does not support the negation operator, but mostly because the lists of siblings and concepts elsewhere in the taxonomy (against which any negation operator would be run) is neither the same as in the source classification nor routinely static over time.

This fundamental representational problem was recognized for ICD9/10: originally, SNOMED CT inherited thousands of codes from SNOMED RT and/or CTV3 that were clearly intended to be semantically equivalent to list closure concepts from ICD10 and other classifications, such as 194712002|Other acute rheumatic heart disease NOS (disorder)|. Although initially integrated within the active SNOMED CT taxonomy, but flagged as 'limited status', they were all formally inactivated in SNOMED CT from 2005/6.

The treatment of SNOMED CT concepts derived from ICD-O list closure codes is currently different; none are limited status. Instead, a small number (currently only about ten) of the relevant SNOMED CT concepts have been flagged with a 'no ICD-O subtype' term infix, including:

                     SNOMED CT                                                ICD-O-3

128842008|Myeloproliferative neoplasm, no ICD-O subtype|        9960/3 Chronic myeloproliferative disease, NOS

14990007|Chondrosarcoma, no ICD-O subtype|                          9220/3 Chondrosarcoma, NOS

19665009|Tubular adenoma, no ICD-O subtype|                         8211/0 Tubular adenoma, NOS

25169009|Neuroma, no ICD-O subtype|                                     9570/0 Neuroma, NOS

373381004|Myelodysplastic syndrome, no ICD-O subtype|          9989/3 Myelodysplastic syndrome, NOS

73348003|Oligodendroglioma, no ICD-O subtype|                       9450/3 Oligodendroglioma, NOS

74532006|Glioma, malignant, no ICD-O subtype|                         9380/3 Glioma, NOS

89880005|Ganglioglioma, no ICD-O subtype|                              9505/1 Ganglioglioma, NOS

This infix is, presumably, intended to alert the reader to the problem: the ICD-O hierarchy must be consulted to fully understand the meaning of the code.

However, the official WHO distribution of ICD-O-3 currently lists over 400 topography and morphology list closure codes with terms terminating in an 'NOS' postfix. The discrepancy in numbers using SNOMED CT's 'no ICD-O subtype' infix convention may be due to the fact that many ICD-O 'NOS' list closure concepts were projected into the SNOMED CT taxonomy as concepts whose descriptions have had the 'NOS' postfix stripped, for example:

          SNOMED CT (CODE + FULLY SPECIFIED NAME) SNOMEDID ICD-O-3

27090000|Transitional cell carcinoma (morphologic abnormality)|                     M-81203        8120/3 Transitional cell carcinoma, NOS

50894008|Schneiderian papilloma (morphologic abnormality)|                         M-81210        8121/0 Schneiderian papilloma, NOS

46580000|Transitional cell papilloma, inverted (morphologic abnormality)|        M-81211        8121/1 Transitional cell papilloma, inverted, NOS

24482001|Atypical adenoma (morphologic abnormality)|                               M-81401        8140/1 Bronchial adenoma, NOS

51642000|Adenocarcinoma in situ (morphologic abnormality)|                       M-81402        8140/2 Adenocarcinoma in situ, NOS

35917007|Adenocarcinoma, no subtype (morphologic abnormality)|              M-81403        8140/3 Adenocarcinoma, NOS

4590003|Adenocarcinoma, metastatic (morphologic abnormality)|                 M-81406        8140/6 Adenocarcinoma, metastatic, NOS

128878003|Islet cell tumor (morphologic abnormality)|                                 M-81501        8150/1 Islet cell tumor, NOS

25324008|Insulinoma (morphologic abnormality)|                                       M-81510        8151/0 Insulinoma, NOS

128855009|Glucagonoma (morphologic abnormality)|                                 M-81521        8152/1 Glucagonoma, NOS

16189002|Gastrinoma (morphologic abnormality)|                                     M-81531        8153/1 Gastrinoma, NOS

447643008|Vipoma (morphologic abnormality)|                                         M-81551        8155/1 Vipoma, NOS

128642005|Somatostatinoma (morphologic abnormality)|                           M-81561        8156/1 Somatostatinoma, NOS

128644006|Enteroglucagonoma (morphologic abnormality)|                       M-81571        8157/1 Enteroglucagonoma, NOS

25370001|Hepatocellular carcinoma (morphologic abnormality)|                 M-81703        8170/3 Hepatocellular carcinoma, NOS

Organisational Issues

Unless and until WHO and IHTSDO co-publish or otherwise synchronise the schedules of their new releases, SNOMED CT and ICD-O can never be 100% synchronized. There will always be some delay between e.g. a new morphology code being approved as part of ICD-O and a corresponding code being published in SNOMED CT, even if both organizations had representation on their respective editorial boards (or some other close collaboration mechanism, such as the JAG…) and so had early pre-publication sight of each other's work.

However, a bigger barrier to mirroring ICD-O-3 content within SNOMED CT is that it isn't clear which list of enumerated ICD-O codes and terms is the authoritative one that exists to be mirrored, and who publishes it: the machine readable distribution currently downloadable from WHO is relatively sparse and also has some significant technical issues; several other sources distribute different and significantly more comprehensive lists. Periodic official WHO-FIC modifications to ICD-O, including new codes, are often documented first (and often only ever) in exclusively human-readable PDF files.

Any ongoing programme to achieve a more comprehensive and timely mirroring would be significantly facilitated by improved content governance and technical stewardship of the ICD-O half of the equation.

Technical considerations for future encodings of a SNOMED-ICDO map

Both the original 'ICD-O/SNOMEDID codestring equivalence' approach and the current mapping table solution have significant inherent limitations for any future scenario in which SNOMED's own representation of oncology morphology becomes explicitly compositional (as ICD-Os own 3-part morphology code representation already is). For example, future representation of metastatic morphologies and tumour instances by SNOMED CT may be best achieved not by full enumeration of the relevant content as precoordinated codes, but instead be available only as postcoordinated expressions to be constructed by end users (e.g. as an existing base 'behaviour agnostic' morphology code qualified by a suitable 'metastatic' qualifier). In this scenario, the existing RF1 crossmapping approach ceases to be a viable way to encode a map between SNOMED and ICD-O - partly because it can not technically hold SNOMED expressions and partly because, even if it could, there would be too many of them to exhaustively list, map and assure.

By contrast, SNOMED's RF2 RefSet mapping mechanism could at least technically represent a future mapping relationship between two explicitly postcoordinating schemes. However, significant scaling issues may still remain due to the sheer number of potential expressions and the inherent challenge of manually verifying the ICD-O code for each.

[To represent such a map from postcoordinated SNOMED CT expressions, an appropriately modified RF2 RefSet design pattern would also be required. A further trivial extension of the existing iisssccRefSet or iissscRefSet design patterns for map artefacts would not be sufficient, because referencedComponentID element of the current RefSet design presupposes that the source concept of a SNOMED CT map can only ever be a precoordinated concept and not also a postcoordinated expression; by contrast, the ICD-O map target is already permitted to be an expression.

A possible technical workaround to this RF2 design limitation has been discussed in other circles, notably those supporting the LOINC integration. In essence it involves e.g. an siisssccRefSet extension of the existing iissscc 'ExtendedMap' pattern, within which the obligatory referencedComponentID column would be either always NULL or populated with a meaningless identifier, and the postcoordinated SNOMED CT source expression would be encoded instead within the new leading STRING datatype column, as SNOMED CT Compositional Grammar expressions.]

Review of terms

Part 3 of the proposed solution identifies a problem with the additional synonyms that have been added over the years to SNOMED CT concepts that have ICD-O-3 correspondents, where the term added is not truly synonymous with the reference ICD-O-3 meaning. In some cases, this addition of further synonyms in SNOMED CT has resulted in the same term now being associated with two different SNOMED codes and so, by implication, with two different ICD-O morphologies. The WHO's own reference data for ICD-O-3 does not have this problem: no term is associated with more than one morphology code.

No list of known infractions is known to exist; it would very likely be incomplete even if it did exist.

Subsidiary and interrelated problems

There is a dependency on the prior successful resolution of the issues in:
artf6220 Malignant neoplasm, primary malignant
artf6221 Recurrent malignant neoplasm
artf221501 Metastatic malignant neoplasm
Other relevant issues include:
artf221574 Update grade and stage concepts based on AJCC 7th Ed

Risks / Benefits

Risks of not addressing the problem

At the time of writing, the oncology surveillance community is not logging high volumes of complaints through NRCs or directly to IHTSDO in relation to issues with the existing mapping relationship between SNOMED CT and ICD-O. A few highly specialist oncology subdomains (e.g. haematological oncology) are reporting some issues in data migration or other interoperability settings, where the content and/or the mapping exhibits significant divergence: for example, some registries are attempting to support receiving raw SNOMED CT data from secondary care sites and then internally mapping these onward to ICD-O. The known deficiencies in the current SNOMED-to-ICDO map therefore continue to be rated a low priority item by IHTSDO and the JAG.

Risks of addressing the problem

If this work item is progressed before the interrelated issues (3.5 above) have been satisfactorily addressed, then a risk arises that later changes in the modelling of e.g. metastatic disease would require significant re-working of the map to ICD-O.

Requirements: criteria for success and completion

Criteria for success/completion

1. A table of equivalence assigns each ICD-O-3 topography code to a unique SNOMED CT code, (except list closure concepts from ICD-O-3)
2. A table of equivalence assigns each ICD-O-3 morphology code to a unique SNOMED CT expression (except list closure concepts from ICD-O-3).
3. For each SNOMED CT code or expression in these tables of equivalences, its descriptions have been checked to ensure true mutual synonymy with each other and with the corresponding ICD-O code.
4. A simple map refset assigns each SNOMED CT concept below 91723000|Anatomical structure (body structure)| to one (or, occasionally, more than one) ICD-O code.
5. An IHTSDO-WHO collaboration mechanism exists to ensure that this realignment is maintained in a timely fashion.

Strategic and/or specific operational use cases

Use case 1

Fit with IHTSDO strategy

This item corresponds to content development area 3A identified in the 2012 Content Development Plan:

Outline Possible Technical Approaches and Concept Model

Indicative Solutions

Approach One

Once an authoritative source has been identified from which to obtain the current ICD-O-3 codelists, and future updates, the solution can be divided into a smaller (topography) and a larger (morphology) workpackage. Each workpackage could then be further divided into two activities: a manual inspection of the existing mapping table in order to identify and list the infractions (missing codes; non-synonymous synonyms), and an authoring task to correct them.
An exhaustive inspection – especially for the morphology chapter – would be laborious and costly. Further, we might expect a generally low level of infractions and, as a consequence, would predict that any attempt at an exhaustive inspection may have relatively low sensitivity due to 'reviewer fatigue'. Consideration should therefore be given as to whether the inspection component could be targeted: maps could be automatically validated (ie not require manual inspection) if the ICD-O and SNOMED terms met agreed lexical similarity criteria, and the extensional definitions (taxonomic milieu) of each pair of concepts were, structurally, also sufficiently similar. Only maps that failed one or other criteria would be offered for manual review.

Indicative Project Plan

Scope of elaboration phase

Further elaboration of a solution specification would benefit from an opinion from the JAG as well as technical input on the potential for targeted review.
To reduce reviewer fatigue and so improve the sensitivity of the review phase, it would be better to have more reviewers doing small volumes of reviewing.
Some items within the domain to be reviewed are very specialist; some of the reviewers will need to be clinical oncology domain specialists.

Projection of overall project size and resource requirements

Expected project resource requirement

If the project decides to map a typical extended catalogue of ICD-O morphologies, then the worst case (brute force) approach would involve the manual inspection of almost 10,000 mapping relationships. Based on similar reviewing exercises undertaken in the UK, this represents approximately 250 hours of often very rare and highly skilled reviewer time.
Assuming a 5% infraction rate and 15 minutes to correct each, the project would require about 125 hours of authoring time.
The project resource requirement is therefore classed as SMALL – less than 6 person months
Whilst a targeted approach may allow a significant reduction in the reviewer effort required, any resource saving may be largely offset by the new requirement for a technical resource to implement the targeting algorithm.

Expected project impact and benefit

The project impact is MEDIUM – significant improvement to a minority but high profile use case

Indicative resource estimates for elaboration, construction, transition and maintenance:

Elaboration phase: 3 person months' effort, 6-12 months elapsed time
Construction and transition phase: 500 concepts to be authored or corrected
Maintenance phase: 500-1000 new 'frequent usage' concept requests in 1^st 3 years