Radicality of ovarian, fallopian tube and primary peritoneal cancer surgery in England

A population-based analysis¹of survival outcomes from 11 gynaecological cancer centres with known surgical ethos demonstrated extensive variation in surgical practice and association with differences in patient survival.

The surgical analysis phase of the OCAFP project therefore proposed to utilise routinely collected HES APC data to derive estimates of surgical radicality within gynaecological cancer centres at a hospital episode level, elucidating the extent to which extensive, ultra-radical surgery was offered in England. Additionally, the project sought to investigate whether surgical radicality was associated with survival.

In advance of this project, a list of recommended OPCS-4 surgical procedure codes was regularly circulated to the BGCS membership, who were in turn encouraged to disseminate the list to coding staff within hospital trusts to harmonise surgical data capture and submission. It was envisaged that the promotion of an agreed set of codes would permit a robust nationwide analysis that could be performed with rigour.

Selecting surgical procedures

To validate the coding within HES, a list of codes for all distinct procedures recorded during the first nine months following diagnosis was extracted for a sample cohort of over 13,000 ovarian cancer patients (excluding borderline cases) diagnosed 2017 to 2018.

The resulting list of more than 1,000 procedure codes was then manually audited by three surgical gynaecological oncologists to identify entries of relevance to the surgical treatment of ovarian cancer. A total of 51 such codes were identified, representing 95% of all distinct surgical procedures captured.

Each of the 51 procedure codes was assigned a score between one and three, as determined by consensus by the three clinicians (Appendix 1). Higher values indicated greater radicality. Each procedure was also allocated to a distinct category of surgery, which denoted the type of operation undertaken.

Although relevant to the surgical treatment of ovarian cancer, descriptions for three of the 51 selected procedures were deemed by the clinicians to be of insufficient detail to determine radicality. In these circumstances, two scores were assigned to each of the three procedures with the aim of reflecting the minimum or maximum probable level of radicality given the description available.

Scores were then summed at a hospital episode level, with each episode categorised as being of low (<4), intermediate (4-7) or high (≥8) complexity. When summing scores within each episode, to avoid a false impression of increased radicality from separate coding of routine elements of hysterectomy and salpingo-oophorectomy surgery, a maximum of one procedure was considered for codes denoting the following procedure categories: “hysterectomy”, “salpingectomy”, “TH-BSO, USO or BSO”.

This categorisation was undertaken twice, selecting either the minimum or maximum scores assigned by clinicians to each of the 51 procedures of interest, yielding two sets of results within which the ‘true’ level of surgical radicality should be situated.

It was envisaged that an analysis of these codes and their accompanying scores would provide an accurate summary of the radicality of surgical episodes.

To confirm that the correct codes were selected and that the assigned scores were appropriate, a validation exercise was undertaken. This involved extracting episode-level procedure data from HES APC for one sample cancer centre in England. These events and their corresponding scores were then linked to detailed local hospital data that comprehensively described the actual procedures undertaken.

Two gynaecological oncology consultants audited these data and found poor concordance between HES APC and local hospital data, with discrepancies observed for more than 10% of patients. This prevented rigorous validation of the proposed method.

As the code list included those recommended for use by the BGCS and captured more than 95% of all procedure events in HES APC for a sample ovarian cohort, this discrepancy was hypothesised to be a consequence of incomplete local clinical coding to populate the institution’s HES records.

Further validation and refinement to check this hypothesis required more resource than was available to the OCAFP, needing validation within other cancer centres to determine whether coding discrepancies were localised or widespread, and in which ways coding practices could be improved.

The OCAFP was not funded to validate surgical radicality scores within centres and directly advise on improvements to coding practice. It is clear from the work undertaken that any analysis of surgical radicality using nationally collected surgical data in HES APC should include a validation exercise such that radicality can be robustly defined. In the absence of such interrogation, there is a high likelihood that derived radicalities will be unreliable.

Nevertheless, the question of variation in surgical practice across England and its association with cancer survival remains an extremely important and unaddressed question. Our recommendation for the forthcoming nationally funded national ovarian cancer audit is to identify a set of cancer centres which will be willing to work with the national audit to share granular local data, such that a method of radicality scoring of routinely captured HES and cancer registration data can be refined. Moreover, to improve and standardise the submission of surgical data to routine data sources.