CTCAE v5.0 Grading

1library(yamaa)23lb <- yamaa_domain("spec.yaml")$output4lb

1import yamaa23lb = yamaa.yamaa_domain("spec.yaml").output4lb

1# Reference solution for the yamaa benchmark sdtm-lb-grading (R track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67library(dplyr, warn.conflicts = FALSE)8library(readr)910raw <- read_csv(11  "/app/input/lb_raw.csv",12  col_types = cols(13    LBSEQ = col_integer(),14    LBSTRESN = col_double(),15    .default = col_character()16  )17)1819toxgrade <- function(test, sex, value) {20  if (is.na(value)) return(NA_character_)21  if (test == "ANC") {22    if (value < 0.5) return("4")23    if (value < 1.0) return("3")24    if (value < 1.5) return("2")25    if (value < 1.8) return("1")26    return("0")27  }28  if (test == "HGB") {29    if (!sex %in% c("M", "F")) return(NA_character_)30    if (value < 8.0) return("3")31    if (value < 10.0) return("2")32    limit <- if (sex == "M") 13.5 else 12.033    if (value < limit) return("1")34    return("0")35  }36  NA_character_37}3839lb <- raw |>40  filter(LBTESTCD %in% c("ANC", "HGB"), !(LBTESTCD == "HGB" & !SEX %in% c("M", "F"))) |>41  rowwise() |>42  mutate(LBTOXGR = toxgrade(LBTESTCD, SEX, LBSTRESN)) |>43  ungroup() |>44  mutate(DOMAIN = "LB") |>45  select(DOMAIN, STUDYID, USUBJID, LBSEQ, LBTESTCD, LBSTRESN, LBTOXGR) |>46  arrange(STUDYID, USUBJID, LBSEQ)4748dir.create("/app/output", showWarnings = FALSE)49write_csv(lb, "/app/output/lb.csv", na = "")

1# Reference solution for the yamaa benchmark sdtm-lb-grading (Python track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67from pathlib import Path89import polars as pl1011raw = pl.read_csv("/app/input/lb_raw.csv", infer_schema=False).with_columns(12    pl.col("LBSEQ").cast(pl.Int64),13    pl.col("LBSTRESN").cast(pl.Float64, strict=False),14)151617def toxgrade(test, sex, value):18    if value is None:19        return None20    if test == "ANC":21        if value < 0.5:22            return "4"23        if value < 1.0:24            return "3"25        if value < 1.5:26            return "2"27        if value < 1.8:28            return "1"29        return "0"30    if test == "HGB":31        if sex not in ("M", "F"):32            return None33        if value < 8.0:34            return "3"35        if value < 10.0:36            return "2"37        limit = 13.5 if sex == "M" else 12.038        if value < limit:39            return "1"40        return "0"41    return None424344rows = []45for r in raw.to_dicts():46    grade = toxgrade(r["LBTESTCD"], r["SEX"], r["LBSTRESN"])47    if grade is None and r["LBTESTCD"] not in ("ANC", "HGB"):48        continue49    if r["LBTESTCD"] == "HGB" and r["SEX"] not in ("M", "F"):50        continue51    rows.append(52        {53            "DOMAIN": "LB",54            "STUDYID": r["STUDYID"],55            "USUBJID": r["USUBJID"],56            "LBSEQ": r["LBSEQ"],57            "LBTESTCD": r["LBTESTCD"],58            "LBSTRESN": r["LBSTRESN"],59            "LBTOXGR": grade,60        }61    )6263lb = pl.DataFrame(64    rows,65    schema={66        "DOMAIN": pl.String,67        "STUDYID": pl.String,68        "USUBJID": pl.String,69        "LBSEQ": pl.Int64,70        "LBTESTCD": pl.String,71        "LBSTRESN": pl.Float64,72        "LBTOXGR": pl.String,73    },74).sort(["STUDYID", "USUBJID", "LBSEQ"])7576Path("/app/output").mkdir(parents=True, exist_ok=True)77lb.write_csv("/app/output/lb.csv")

1/* opensas reference solution for the yamaa benchmark sdtm-lb-grading.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data lb_raw;6  length STUDYID USUBJID LBTESTCD SEX REPROGRADE $1024;7  infile "/app/input/lb_raw.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDYID : $char1024. USUBJID : $char1024. LBSEQ LBTESTCD : $char1024. SEX : $char1024. LBSTRESN REPROGRADE : $char1024.;9run;1011data result; set lb_raw; length DOMAIN $2; DOMAIN='LB'; LBTOXGR=.;12if LBTESTCD='ANC' then do; if missing(LBSTRESN) then LBTOXGR=.; else if LBSTRESN<0.5 then LBTOXGR=4; else if LBSTRESN<1 then LBTOXGR=3; else if LBSTRESN<1.5 then LBTOXGR=2; else if LBSTRESN<1.8 then LBTOXGR=1; else LBTOXGR=0; end;13else if LBTESTCD='HGB' and SEX in ('M','F') then do; if missing(LBSTRESN) then LBTOXGR=.; else if LBSTRESN<8 then LBTOXGR=3; else if LBSTRESN<10 then LBTOXGR=2; else if (SEX='M' and LBSTRESN<13.5) or (SEX='F' and LBSTRESN<12) then LBTOXGR=1; else LBTOXGR=0; end; else delete; run;14proc sql;15  create table final as select DOMAIN, STUDYID, USUBJID, LBSEQ, LBTESTCD, LBSTRESN, LBTOXGR from result;16quit;17proc export data=final outfile="/app/output/lb.csv" dbms=csv replace; run;