Sir,

We read with interest the paper by Bendifallah et al, 2013 on the validation of our previous nomogram (Obermair et al, 2013) to estimate the risk of recurrence after surgery for Borderline Ovarian Tumours (BOT) and we congratulate the authors on their work.

A diagnosis of BOT is typically established only postoperatively and many women diagnosed with BOT are in their childbearing years. Although most BOT patients will expect excellent outcomes, a small proportion of women will recur. Prediction of relapse is critical. Patients with remotely low risk of relapse can be discharged from regular follow-up. By contrast, patients at high risk of relapse may benefit from extended surgery or regular, lifelong follow-up because recurrences may develop late after surgery (Silva et al, 2006).

Our nomogram was the first attempt to quantify a patient’s individual risk of relapse and included covariates from readily available clinical, biological and pathological characteristics. We made every attempt to create a representative sample and therefore included all consecutive patients from six gynaecological cancer centres. Hence, almost 80% of patients in our group were classified stage 1.

The French group of clinicians abstracted information from 314 patients from two French institutions between 1980 and 2008. To validate our nomogram, they repeated our study using identical covariates. However, their patient sample was distinctly different to ours. Stage 2, 3 and 4 was almost five times as common in the French study than in ours. The pre-operative median serum CA125 was more than double as high in the French paper than in ours (77.6 U ml−1 vs 36 U ml−1). Expectedly, relapses developed in 5.5% vs 29.9% (French study).

After discussions with the French authors, it became clear that pathologists at those two French institutions regularly review high-risk BOT cases referred from other institutions. It seems that those cases were included in the reporting of this series, thus resulting in a very significant over-representation of high-risk cases.

While the Australian series reported the outcomes of a representative sample of all BOT, the French cohort was not representative of all BOT cases but provided an over-representation of high-risk patients.

Although both samples overlap to a degree, the French cohort was not a comparable patient cohort and therefore was not suited to validate the Australian cohort. Both samples were profoundly different in regards to patients’ characteristics and outcomes. A comparison of those two samples should have excluded samples from external review to level the field.