predict.models.predict — predict-model 2.1.3

A cljs version of the predict model, enhanced with radiotherapy and bisphosphonates, and extended to 15 years.

abs cljs

age cljs

(age y)

base-m-cum-br cljs

(base-m-cum-br erstat tm)

baseline survival. Actually baseline-mortality! R: base.m.cum.br

cljs-predict cljs

(cljs-predict {:keys [age size nodes grade erstat detection her2 ki67 rtime
                      radio? bis? chemoGen horm radio bis tra]
               :as inputs})

clojure/script implementation of predict-v2 model.

Predicts survival based on patient input parameters.

Arguments age, size and nodes are entered as values; the others as lookups

This is how the model assigns some input parameters (or ranges) into variables

i.e. parameter (or ranges) -> web form setting -> Predict model variable setting

Tumour Grade (1,2,3,unknown) -> (1,2,3,9) -> (1.0,2.0,3.0,2.13)

ER Status (-ve,+ve) -> (0,1) -> (0,1) n.b. unknown not allowed

Detection (Clinical,Screening,Other) -> (0,1,2) -> (0.0,1.0,0.204)

HER2 Status (-ve,+ve,unknown) -> (0,1,9)

KI67 Status (-ve,+ve,unknown) -> (0,1,9)

We are now passing in the selected treatments so we don't have to calculate all possible treatment combinations on each call. Instead, we calculate the treatment combinations that could make up the current set in hrctb order

This means that if we see non-null horm, chemoGen, bis values, we will calculate h, hc, hb, hcb only.

Note: For uncertainties in the coefficients h,c,t etc, see docs/Predictv2-uncertainties.docx

deltas cljs

(deltas start v)

detection cljs

(detection d)

exp cljs

grade-a cljs

(grade-a grade)

her2-rh cljs

(her2-rh her2)

ki67-rh cljs

(ki67-rh erstat ki67)

ln cljs

log-age-10 cljs

(log-age-10 age)

log(age/10)

m-oth-prognostic-index cljs

(m-oth-prognostic-index age radio?)

pow cljs

prognostic-index cljs

(prognostic-index {:keys [age size nodes grade erstat detection her2-rh ki67-rh
                          grade-a radio? bis?]
                   :or {detection 0
                        ki67-rh -0.11333
                        age 65
                        bis? true
                        her2-rh -0.0762
                        grade-a 0
                        radio? true
                        size 19
                        nodes 1
                        grade 1
                        erstat 1}})

Calculate the breast cancer mortality prognostic index (pi). Comments relate this code to the corresponding R variables.

r-base-br cljs

(r-base-br radio?)

R: r.base.br Base breast cancer mortality coefficient for radiotherapy enabled or disabled.

r-base-oth cljs

(r-base-oth radio?)

R: r.base.oth Base other cause mortality coefficient

r-br cljs

(r-br radio?)

R: r.br breast mortality coefficient for radiotherapy treatment if enabled

r-oth cljs

(r-oth radio?)

rec-age-10-sq cljs

(rec-age-10-sq age)

1/(age/10)^2

types-rx cljs

(types-rx {:keys [erstat her2 horm chemoGen radio? radio bis? bis tra]})

Calculate treatment coefficients radio indicates radiotherapy is available in the interface and selected bis indicates bisphosphonates is available in the interface and selected c = chemo, h = hormone therapy, t = trastuzumab, r = radiotherapy, b = bisphosphonates

Note that we always calculate the same columns, but if a treatment is not selected, then its associated treatment coefficients will be zero.

e.g. The treatment combination hcb will be calculated as hrctb, but with r and c coefficients zeroed.