Skip to main content

Table 1 REMARK profile—improving the Nottingham Prognostic Index (NPI), adapted from Winzer et al. [16]

From: Structured reporting to improve transparency of analyses in prognostic marker studies

Part a: Patients, treatments, and variables
Study and marker Remarks
  Marker handled  = NPI Continuous and categorical. Cutpoints as predefined in the literature. For details see Blamey et al. [27] in [16]
  Further variables v1 = Tumor size, v2 = no. of pos. Lymph nodes, v3 = tumor grade, v4 = age, v5 = histology, v6 = hormone receptor status, v7 = menopausal status, v8 = vessel invasion, v9 = lymphatic vessel invasion
Patients n Remarks
  Assessed for eligibility 2062 Disease: primary breast cancer
Patient source: Database Surgical clinic Charité, Berlin. All patients with surgery from 1st Jan. 1984 to 31st Dec. 1998.
  Excluded 502 63 metastasis, 73 previous carcinomas other than breast cancer, 86 primary breast cancer prior to the study, 134 breast cancer in situ, 8 pt0, 123 older than 80 years, 20 neo-adjuvant chemotherapy, 71 death within the first months of surgery, three or more standard prognostic factors missing. For some patients, more than one exclusion criterion applied
  Included 1560 Previously untreated. Treatment: local therapy: BCT or mastectomy with or without radiotherapy, adjuvant therapy: chemo (y/n), hormone (y/n). For details see, Add file 1 and table 2 in Winzer et al. [28] in [16]
  Outcome events 221 Overall survival: death from any cause
Part b: Statistical analysesh
Analysis Patients Events Variables considered Results/remarks
  IDA 1a: imputation for missing values 1560 NRb v1(94), v2 (68), v3(217), v6(490), v7(54) Variables (number of patients) with imputed values
  A1c: NPI (3) 1560 221 NPI Prognostic value of NPI in 3 categories (Tables 2 and 3, Fig. 1)
  A2: NPI (6) 1560 221 NPI 6 categories (Fig. 1, Table 3)
  C1d: check of PHe in NPI (3) and in NPI (6) 1560 221 NPI Fig. 2, S4 and non-significant result of FPTf (see last paragraph 4.2)
  A3: NPIcont 1560 221 NPI More information from continuous data? (Table 3)
  C2: NPIcont. has a linear effect 1560 221 NPI FP2 function not significantly better, see 4.3.1
  C3: check of PHe in NPIcont 1560 221 NPI Non-significant result of FPTf (see last paragraph 4.3.1)
  A4: MFP7 of the three NPI variables (univ. and multivariable) 1560 221 v1, v2, v3 Table 4
  A5: functional form for nodes 1560 221 v2 Fig. 3
  A6: prognostic value and additional value of further variables (univ. and multiv.) 1560 221 NPI, v4, v5, v6, v7, v8, v9 Table 5, Fig. 4
  A7: MFP using all available information 1560 221 v1, v2, v3, v4, v5, v6, v7, v8, v9 Final MFP model in Table 6, see 4.5
  A8: measures of separation 1560 221 NPI, v1, v2, v3, v4, v5, v6, v7, v8, v9 Table 7, see 4.6
  C4: check of PHe in MFP model 1560 221 v1, v2, v3, v6 Non-significant result of FPTf (see end of 4.5)
  1. aIDA = initial data analysis
  2. bNR = not relevant
  3. cA1, A2, … = number of analysis
  4. dC1, C2, … = number of check
  5. ePH = proportional hazards,
  6. fFPT = fractional polynomial time procedure
  7. gMFP = multivariable fractional polynomial procedure
  8. hAll analyses using a Cox model are stratified for strata according to therapy. There are 8 strata defined by the combination of surgery, radiotherapy (y/n), and systemic therapy (y/n (no chemotherapy and no hormone therapy))