Duration of Response

1library(yamaa)23adtte <- yamaa_domain("spec.yaml")$output4adtte

1import yamaa23adtte = yamaa.yamaa_domain("spec.yaml").output4adtte

1# Reference solution for the yamaa benchmark adam-adtte-dor (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)910adsl <- read_csv(11  "/app/input/adsl.csv",12  col_types = cols(RESPDT = col_date(), NACTDT = col_date(), .default = col_character())13)14adrs <- read_csv(15  "/app/input/adrs_raw.csv",16  col_types = cols(ASEQ = col_integer(), ADT = col_date(), .default = col_character())17)18ds <- read_csv(19  "/app/input/ds.csv",20  col_types = cols(DSSEQ = col_integer(), DSSTDTC = col_date(), .default = col_character())21)2223# Earliest progression and death, lowest sequence number on a tie.24progression <- adrs |>25  filter(AVALC %in% "PD", !is.na(ADT)) |>26  arrange(USUBJID, ADT, ASEQ) |>27  distinct(USUBJID, .keep_all = TRUE) |>28  select(USUBJID, PDDT = ADT, PDSEQ = ASEQ)2930death <- ds |>31  filter(DSDECOD %in% "DEATH", !is.na(DSSTDTC)) |>32  arrange(USUBJID, DSSTDTC, DSSEQ) |>33  distinct(USUBJID, .keep_all = TRUE) |>34  select(USUBJID, DTHDT = DSSTDTC, DTHSEQ = DSSEQ)3536# Last evaluable assessment (a date and a response other than NE), highest37# sequence number on a tie.38last_assessment <- adrs |>39  filter(!is.na(ADT), !is.na(AVALC), AVALC != "NE") |>40  arrange(USUBJID, desc(ADT), desc(ASEQ)) |>41  distinct(USUBJID, .keep_all = TRUE) |>42  select(USUBJID, LASTDT = ADT, LASTSEQ = ASEQ)4344# Where ADT comes from, for each outcome.45sources <- tribble(46  ~OUTCOME,                           ~SRCDOM, ~SRCVAR,47  "DISEASE PROGRESSION",              "ADRS",  "ADT",48  "DEATH",                            "DS",    "DSSTDTC",49  "LAST TUMOUR ASSESSMENT",           "ADRS",  "ADT",50  "START OF NEW ANTI-CANCER THERAPY", "ADSL",  "NACTDT"51)5253adtte <- adsl |>54  filter(!is.na(RESPDT)) |>55  left_join(progression, by = "USUBJID") |>56  left_join(death, by = "USUBJID") |>57  left_join(last_assessment, by = "USUBJID") |>58  mutate(59    # The first event; progression wins a same-day tie with death.60    EVENT = case_when(61      !is.na(PDDT) & (is.na(DTHDT) | PDDT <= DTHDT) ~ "DISEASE PROGRESSION",62      !is.na(DTHDT) ~ "DEATH"63    ),64    EVENTDT = if_else(EVENT == "DEATH", DTHDT, PDDT),65    # An event after the start of new therapy does not count.66    EVENT = if_else(!is.na(NACTDT) & EVENTDT > NACTDT, NA_character_, EVENT),67    # Without a counted event, the record is censored at the earlier of the68    # last evaluable assessment and the start of new therapy.69    OUTCOME = case_when(70      !is.na(EVENT) ~ EVENT,71      !is.na(LASTDT) & (is.na(NACTDT) | LASTDT <= NACTDT) ~ "LAST TUMOUR ASSESSMENT",72      !is.na(NACTDT) ~ "START OF NEW ANTI-CANCER THERAPY"73    ),74    ADT = case_when(75      OUTCOME == "DISEASE PROGRESSION" ~ PDDT,76      OUTCOME == "DEATH" ~ DTHDT,77      OUTCOME == "LAST TUMOUR ASSESSMENT" ~ LASTDT,78      OUTCOME == "START OF NEW ANTI-CANCER THERAPY" ~ NACTDT79    ),80    SRCSEQ = case_when(81      OUTCOME == "DISEASE PROGRESSION" ~ PDSEQ,82      OUTCOME == "DEATH" ~ DTHSEQ,83      OUTCOME == "LAST TUMOUR ASSESSMENT" ~ LASTSEQ84    ),85    PARAMCD = "DOR",86    PARAM = "Duration of Response",87    STARTDT = RESPDT,88    AVAL = as.integer(ADT - STARTDT) + 1L,89    CNSR = if_else(is.na(EVENT), 1L, 0L),90    EVNTDESC = coalesce(EVENT, "CENSORED"),91    CNSDTDSC = if_else(is.na(EVENT), OUTCOME, NA_character_)92  ) |>93  left_join(sources, by = "OUTCOME") |>94  select(95    STUDYID, USUBJID, PARAMCD, PARAM, STARTDT, ADT, AVAL, CNSR, EVNTDESC,96    CNSDTDSC, SRCDOM, SRCVAR, SRCSEQ97  )9899dir.create("/app/output", showWarnings = FALSE)100write_csv(adtte, "/app/output/adtte.csv", na = "")

1# Reference solution for the yamaa benchmark adam-adtte-dor (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 pl101112def read(name: str) -> pl.DataFrame:13    return pl.read_csv(f"/app/input/{name}", infer_schema=False)141516def first_per_subject(17    records: pl.DataFrame, by: list[str], descending: bool = False18) -> pl.DataFrame:19    return records.sort(by, descending=descending).unique(20        "USUBJID", keep="first", maintain_order=True21    )222324adsl = read("adsl.csv").with_columns(pl.col("RESPDT", "NACTDT").str.to_date())25adrs = read("adrs_raw.csv").with_columns(26    pl.col("ADT").str.to_date(), pl.col("ASEQ").cast(pl.Int64)27)28ds = read("ds.csv").with_columns(29    pl.col("DSSTDTC").str.to_date(), pl.col("DSSEQ").cast(pl.Int64)30)3132# Earliest progression and death, lowest sequence number on a tie.33progression = first_per_subject(34    adrs.filter(pl.col("AVALC") == "PD", pl.col("ADT").is_not_null()),35    ["ADT", "ASEQ"],36).select("USUBJID", PDDT="ADT", PDSEQ="ASEQ")3738death = first_per_subject(39    ds.filter(pl.col("DSDECOD") == "DEATH", pl.col("DSSTDTC").is_not_null()),40    ["DSSTDTC", "DSSEQ"],41).select("USUBJID", DTHDT="DSSTDTC", DTHSEQ="DSSEQ")4243# Last evaluable assessment (a date and a response other than NE), highest44# sequence number on a tie.45last_assessment = first_per_subject(46    adrs.filter(pl.col("ADT").is_not_null(), pl.col("AVALC") != "NE"),47    ["ADT", "ASEQ"],48    descending=True,49).select("USUBJID", LASTDT="ADT", LASTSEQ="ASEQ")5051# Where ADT comes from, for each outcome.52sources = pl.DataFrame(53    {54        "OUTCOME": [55            "DISEASE PROGRESSION",56            "DEATH",57            "LAST TUMOUR ASSESSMENT",58            "START OF NEW ANTI-CANCER THERAPY",59        ],60        "SRCDOM": ["ADRS", "DS", "ADRS", "ADSL"],61        "SRCVAR": ["ADT", "DSSTDTC", "ADT", "NACTDT"],62    }63)6465PDDT, DTHDT, LASTDT, NACTDT = (66    pl.col(name) for name in ("PDDT", "DTHDT", "LASTDT", "NACTDT")67)68EVENT, OUTCOME = pl.col("EVENT"), pl.col("OUTCOME")6970adtte = (71    adsl.filter(pl.col("RESPDT").is_not_null())72    .join(progression, on="USUBJID", how="left")73    .join(death, on="USUBJID", how="left")74    .join(last_assessment, on="USUBJID", how="left")75    # The first event; progression wins a same-day tie with death.76    .with_columns(77        EVENT=pl.when(PDDT.is_not_null() & (DTHDT.is_null() | (PDDT <= DTHDT)))78        .then(pl.lit("DISEASE PROGRESSION"))79        .when(DTHDT.is_not_null())80        .then(pl.lit("DEATH"))81    )82    .with_columns(EVENTDT=pl.when(EVENT == "DEATH").then(DTHDT).otherwise(PDDT))83    # An event after the start of new therapy does not count.84    .with_columns(85        EVENT=pl.when(NACTDT.is_not_null() & (pl.col("EVENTDT") > NACTDT))86        .then(None)87        .otherwise(EVENT)88    )89    # Without a counted event, the record is censored at the earlier of the90    # last evaluable assessment and the start of new therapy.91    .with_columns(92        OUTCOME=pl.when(EVENT.is_not_null())93        .then(EVENT)94        .when(LASTDT.is_not_null() & (NACTDT.is_null() | (LASTDT <= NACTDT)))95        .then(pl.lit("LAST TUMOUR ASSESSMENT"))96        .when(NACTDT.is_not_null())97        .then(pl.lit("START OF NEW ANTI-CANCER THERAPY"))98    )99    .with_columns(100        ADT=pl.when(OUTCOME == "DISEASE PROGRESSION")101        .then(PDDT)102        .when(OUTCOME == "DEATH")103        .then(DTHDT)104        .when(OUTCOME == "LAST TUMOUR ASSESSMENT")105        .then(LASTDT)106        .when(OUTCOME == "START OF NEW ANTI-CANCER THERAPY")107        .then(NACTDT),108        SRCSEQ=pl.when(OUTCOME == "DISEASE PROGRESSION")109        .then(pl.col("PDSEQ"))110        .when(OUTCOME == "DEATH")111        .then(pl.col("DTHSEQ"))112        .when(OUTCOME == "LAST TUMOUR ASSESSMENT")113        .then(pl.col("LASTSEQ")),114        PARAMCD=pl.lit("DOR"),115        PARAM=pl.lit("Duration of Response"),116        STARTDT=pl.col("RESPDT"),117        CNSR=pl.when(EVENT.is_null()).then(1).otherwise(0),118        EVNTDESC=pl.coalesce(EVENT, pl.lit("CENSORED")),119        CNSDTDSC=pl.when(EVENT.is_null()).then(OUTCOME),120    )121    .with_columns(AVAL=(pl.col("ADT") - pl.col("STARTDT")).dt.total_days() + 1)122    .join(sources, on="OUTCOME", how="left")123    .select(124        "STUDYID",125        "USUBJID",126        "PARAMCD",127        "PARAM",128        "STARTDT",129        "ADT",130        "AVAL",131        "CNSR",132        "EVNTDESC",133        "CNSDTDSC",134        "SRCDOM",135        "SRCVAR",136        "SRCSEQ",137    )138)139140Path("/app/output").mkdir(exist_ok=True)141adtte.write_csv("/app/output/adtte.csv")

1/* opensas reference solution for the yamaa benchmark adam-adtte-dor.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data adrs_raw;6  length STUDYID USUBJID ADT AVALC $1024;7  infile "/app/input/adrs_raw.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDYID : $char1024. USUBJID : $char1024. ASEQ ADT : $char1024. AVALC : $char1024.;9run;1011data adsl;12  length STUDYID USUBJID RESPDT NACTDT $1024;13  infile "/app/input/adsl.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;14  input STUDYID : $char1024. USUBJID : $char1024. RESPDT : $char1024. NACTDT : $char1024.;15run;1617data ds;18  length STUDYID USUBJID DSDECOD DSSTDTC $1024;19  infile "/app/input/ds.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;20  input STUDYID : $char1024. USUBJID : $char1024. DSSEQ DSDECOD : $char1024. DSSTDTC : $char1024.;21run;2223proc sql;24  create table usable as select a.* from adrs_raw as a inner join adsl as b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID where a.ADT ne '' and a.AVALC ne 'NE';25  create table progression as select a.* from adrs_raw as a inner join adsl as b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID where a.ADT ne '' and a.AVALC='PD' and a.AVALC ne 'NE' and (missing(b.NACTDT) or a.ADT<=b.NACTDT);26  create table deaths as select a.* from ds as a inner join adsl as b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID where a.DSDECOD='DEATH' and a.DSSTDTC ne '' and (missing(b.NACTDT) or a.DSSTDTC<=b.NACTDT);27quit;28proc sort data=progression; by STUDYID USUBJID ADT ASEQ; run;29data pd; set progression; by STUDYID USUBJID; if first.USUBJID; keep STUDYID USUBJID ADT ASEQ; rename ADT=PDDT ASEQ=PDSEQ; run;30proc sort data=deaths; by STUDYID USUBJID DSSTDTC DSSEQ; run;31data death; set deaths; by STUDYID USUBJID; if first.USUBJID; keep STUDYID USUBJID DSSTDTC DSSEQ; rename DSSTDTC=DEATHDT DSSEQ=DEATHSEQ; run;32proc sort data=usable; by STUDYID USUBJID descending ADT descending ASEQ; run;33data lastassessment; set usable; by STUDYID USUBJID; if first.USUBJID; keep STUDYID USUBJID ADT ASEQ; rename ADT=LASTDT ASEQ=LASTSEQ; run;34proc sql; create table joined as select a.*,b.PDDT,b.PDSEQ,c.DEATHDT,c.DEATHSEQ,d.LASTDT,d.LASTSEQ from adsl as a35  left join pd as b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID left join death as c on a.STUDYID=c.STUDYID and a.USUBJID=c.USUBJID left join lastassessment as d on a.STUDYID=d.STUDYID and a.USUBJID=d.USUBJID; quit;36data result;37  length PARAMCD $8 PARAM $40 STARTDT ADT $10 EVNTDESC CNSDTDSC $40 SRCDOM $4 SRCVAR $16;38  set joined; if not missing(RESPDT);39  PARAMCD='DOR'; PARAM='Duration of Response'; STARTDT=RESPDT;40  if not missing(PDDT) and (missing(DEATHDT) or PDDT<=DEATHDT) then do; ADT=PDDT; CNSR=0; EVNTDESC='DISEASE PROGRESSION'; SRCDOM='ADRS'; SRCVAR='ADT'; SRCSEQ=PDSEQ; end;41  else if not missing(DEATHDT) then do; ADT=DEATHDT; CNSR=0; EVNTDESC='DEATH'; SRCDOM='DS'; SRCVAR='DSSTDTC'; SRCSEQ=DEATHSEQ; end;42  else do;43    CNSR=1; EVNTDESC='CENSORED';44    if not missing(NACTDT) and (missing(LASTDT) or NACTDT<LASTDT) then do; ADT=NACTDT; CNSDTDSC='START OF NEW ANTI-CANCER THERAPY'; SRCDOM='ADSL'; SRCVAR='NACTDT'; end;45    else do; ADT=LASTDT; CNSDTDSC='LAST TUMOUR ASSESSMENT'; SRCDOM='ADRS'; SRCVAR='ADT'; SRCSEQ=LASTSEQ; end;46  end;47  if not missing(ADT) and not missing(STARTDT) then AVAL=input(ADT,yymmdd10.)-input(STARTDT,yymmdd10.)+1;48run;4950proc sql;51  create table final as select STUDYID, USUBJID, PARAMCD, PARAM, STARTDT, ADT, AVAL, CNSR, EVNTDESC, CNSDTDSC, SRCDOM, SRCVAR, SRCSEQ from result;52quit;53proc export data=final outfile="/app/output/adtte.csv" dbms=csv replace; run;