The Nottingham Prognostic Index: five- and ten-year data for all-cause Survival within a Screened Population

BTW holds a carefully maintained central database for all cancers diagnosed in women who have attended breast screening. All women aged 50–65 years who attended breast screening between April 1998 and April 2001 were followed up for ≥10 years. BTW received notifications of all BCs diagnosed as well as any deaths that occurred during follow-up. The study period of 1998 to 2001 was chosen because that particular screening round was the tenth year since the inception of BTW. At this juncture, BTW would be considered to be running without major problems.

BC patients underwent standardised treatment according to local protocols as agreed by discussion within the regional multidisciplinary team. Surgical treatment for BCs was conducted at respective treatment centres. Subsequent adjuvant therapy was carried out in a regional oncology centre.

All cases in the present study were identified via the BTW and National Breast Screening System (NBSS) database. NBSS holds all the death records of women who attended screening regardless of their age, thereby providing all-cause survival data.

If possible, histological assessment was carried out to provide parameters for calculation of the NPI using the following formula:

NPI = maximum invasive cancer size (S) × 0.2 + lymph node stage (LN = 1, 2, or 3) + histological grade (H = 1, 2, or 3).

Patients were grouped into four categories according to the NPI score: I (excellent) ≤2.4; II (good) >2.4 but ≤3.4; III (moderate) >3.4 but ≤5.4; and IV (poor) >5.4. Kaplan–Meier survival curves were constructed using SPSS v19 (IBM, Armonk, NY, USA) using all-cause survival data.

The total number of women who attended screening in the years 1998 to 2001 was 199,082. The overall uptake was 77.3%.

A total of 1,712 patients were diagnosed with invasive BCs having undergone screening in the three-year period. There were 1,020 screening-detected BCs, and the overall prevalence of cancer detection on screening for these three years was 0.51% (5.1 per 1000 women screened). However, 692 women with ‘normal’ screening mammograms subsequently developed interval BCs, which equates to a prevalence of interval BC of 0.38% (3.48 per 1,000 women screened).

Data were incomplete for 166 patients so the NPI could not be calculated. Therefore, 1,546 cases were included in this study. Table 2 shows the number of cancers in each NPI category as well as overall five- and ten-year survival.

Figure 1 shows the Kaplan–Meier all-cause survival curve of BCs diagnosed in women who underwent screening within one screening cycle, and is broken down into each NPI category. It demonstrates clear segregation of the survival curve for each NPI category with no significant crossover.

Tabar et al. ¹³ described one of the longest follow-ups of BCs patients within a screening programme. After 29 years of follow-up, they demonstrated that the relative risk of BC mortality in screened women was 0.69–0.73 compared with unscreened women. This figure was comparable with their publication of a 30% reduction in BC mortality in 1985. ¹⁴

Overall all-cause ten-year survival in the present study was 82%, which is better than that observed in a study in Sweden and that of recent UK data (77%). ¹¹ Five-year survival of 94% was a marked improvement from the five-year UK figure of 85%. ¹⁰ These UK figures relate to cancer-specific survival, and our data are for all-cause survival. Purists would argue that such comparisons are not compatible but we stress: (i) that any improvement in overall survival is always an achievement worth noting; (ii) if available, BC-specific survival data for our cohort would be even better than that presented here. This trend was also observed in the study from a single centre in Nottingham. Our data may be used as a reference for updating all-cause survival data specifically within a screened population in which treatment was carried out in multiple centres.

1998 to 2001 was the third screening round for BTW. Each screening round is three years. With the first two screening rounds broadly achieving National Health Service Breast Screening Programme standards, we are confident in the quality of the data provided here. Importantly, two-view mammography was introduced for all rounds in BTW in 1998. Prevalent round screens used two view mammography since 1989; data from previous screening rounds were not included in the present study to prevent the introduction of bias. Data from more recent screening rounds were not included because the women who attended screening would not have completed ten-year follow-up to provide the required survival figure.

The validity of the NPI as a prognostic tool for BC has been reviewed. ^1,6–9 The most recent major review was carried out in Nottingham, ¹² with a detailed analysis and internal validation demonstrating the robustness of their data. Our data were comparable with their data with the exception of NPI VI. This difference could be because our patients were diagnosed more recently than those reviewed by the Nottingham team. Also, our patients were thought to have improvements in treatment (especially with introduction of analyses of the status of hormone receptors and hormonal manipulation), so NPI VI patients may have benefited greatly.

Most ‘developed’ countries are running BC screening, so our data highlight the validity of the NPI. The NPI is applicable in most screened populations in which one key centre oversees the screening programme, and where surgical/oncological treatments are carried out in multiple local centres. We analysed all cancers within one screening cycle in a population-based screening programme, and our patients were treated in multiple local centres. Each treatment centre would review each case in their respective multidisciplinary meeting and recommend appropriate treatment according to agreed protocols. This strategy represents how most screening programmes are run.

Other prognostication indices are available and are being used in conjunction with the NPI in the planning of treatment of BCs. For example, AO is highly valuable in helping to formulate adjuvant treatment and is used widely in multidisciplinary meetings. We propose that the NPI should remain one of many tools available to clinicians when formulating treatment plans. The NPI is easy to use and does not require cumbersome and significant clinical, epidemiological or histological assessment.

Overall five-year and ten-year all-cause survival has improved even when compared with BC-specific survival. We found that the NPI remains valid and valuable for BC treatment. However, our data suggest that the NPI should be updated for all-cause survival of women diagnosed with BCs within a screened population.