Skip to contents

Creating ADTPET and ADAPET

Source: vignettes/adpet.Rmd
adpet.Rmd

Introduction

This article describes the creation of PET Scan ADaMs.

We recommend splitting out amyloid and tau into two distinct ADaM datasets. The vignette presents a combined approach for the initial setup, parameter mapping and finals steps. It then splits into separate sections for each data type, as they require different processes: starting with ADTPET (Tau PET Scan Analysis Dataset), followed by ADAPET (Amyloid PET Scan Analysis Dataset).

We advise you first consult the admiral Creating a BDS Finding ADaM vignette. The programming workflow around creating the general set-up of a BDS ADaM dataset using admiral functions is the same. In this vignette, we focus on the neuro-specific steps to avoid repeating information and maintaining the same content in two places. As such, the code in this vignette is not completely executable; we recommend consulting the ADTPET and ADAPET template scripts to view the full workflow.

Note: All examples assume CDISC SDTM and/or ADaM format as input unless otherwise specified.

Required Packages

The examples of this vignette require the following packages.

Programming Workflow

Read in Data

To start, all data frames needed for the creation of ADTPET or ADAPET should be loaded into the global environment. Reading data will usually be a company specific process, however, for the purpose of this vignette, we will use example data from admiralneuro. We will utilize NV, SUPPNV, AG and ADSL for the basis of the PET Scan ADaM datasets.

nv <- admiralneuro::nv_neuro %>%
  convert_blanks_to_na()

suppnv <- admiralneuro::suppnv_neuro %>%
  convert_blanks_to_na()

ag <- admiralneuro::ag_neuro %>%
  convert_blanks_to_na()

adsl <- admiralneuro::adsl_neuro %>%
  convert_blanks_to_na()

While loading our input data, we can combine the NV domain and the SUPPNV supplementary domain for easier use in the next steps. Using the metatools::combine_supp() function avoids the need to manually transpose and merge the supplementary dataset with the corresponding domain.

nv_suppnv <- combine_supp(nv, suppnv)

Initial PET Scan Analysis Dataset Set-up

The following steps are used to merge ADSL variables with the source data NV (previously combined with SUPPNV), along with tracer information from the AG dataset, without distinguishing between Tau and Amyloid data at this stage. This distinction will be made later. Common analysis variables such as Analysis Date (ADT) and Relative Analysis Day (ADY) can also be derived during this process. Note that only the sections relevant to this vignette are included in the steps below. To get a detailed guidance on all the steps, refer the admiral Creating a BDS Finding ADaM vignette.

adsl_vars <- exprs(TRTSDT, TRTEDT, TRT01A, TRT01P)

adpet <- nv_suppnv %>%
  derive_vars_merged(
    dataset_add = adsl,
    new_vars = adsl_vars,
    by_vars = exprs(STUDYID, USUBJID)
  ) %>%
  derive_vars_merged(
    dataset_add = ag,
    new_vars = exprs(AGTRT, AGCAT),
    by_vars = exprs(STUDYID, USUBJID, VISIT, NVLNKID = AGLNKID)
  ) %>%
  derive_vars_dt(
    new_vars_prefix = "A",
    dtc = NVDTC
  ) %>%
  derive_vars_dy(reference_date = TRTSDT, source_vars = exprs(ADT))

Create PARAMCD, PARAM, and PARAMN variables

The next step is to create and assign parameter level variables such as PARAMCD, PARAM, and PARAMN. For this, a lookup can be created based on the SDTM --TESTCD, --CAT, --LOC, --METHOD and REFREG values to join to the source data.

param_lookup <- tibble::tribble(
  ~NVTESTCD, ~NVCAT, ~NVLOC, ~REFREG, ~NVMETHOD, ~PARAMCD, ~PARAM, ~PARAMN,
  "SUVR", "FBP", "NEOCORTICAL COMPOSITE", "Whole Cerebellum", "AVID FBP SUVR PIPELINE", "SUVRAFBP", "AVID FBP Standard Uptake Ratio Neocortical Composite Whole Cerebellum", 1,
  "SUVR", "FBB", "NEOCORTICAL COMPOSITE", "Whole Cerebellum", "AVID FBB SUVR PIPELINE", "SUVRAFBB", "AVID FBB Standard Uptake Ratio Neocortical Composite Whole Cerebellum", 2,
  "SUVR", "FBP", "NEOCORTICAL COMPOSITE", "Whole Cerebellum", "BERKELEY FBP SUVR PIPELINE", "SUVRBFBP", "BERKELEY FBP Standard Uptake Ratio Neocortical Composite Whole Cerebellum", 3,
  "SUVR", "FBB", "NEOCORTICAL COMPOSITE", "Whole Cerebellum", "BERKELEY FBB SUVR PIPELINE", "SUVRBFBB", "BERKELEY FBB Standard Uptake Ratio Neocortical Composite Whole Cerebellum", 4,
  "SUVR", "FTP", "NEOCORTICAL COMPOSITE", "Inferior Cerebellar Gray Matter", "AVID FTP SUVR PIPELINE", "SUVRAFTP", "AVID FTP Standard Uptake Ratio Neocortical Composite Inferior Cerebellar Gray Matter", 5,
  "SUVR", "FTP", "NEOCORTICAL COMPOSITE", "Inferior Cerebellar Gray Matter", "BERKELEY FTP SUVR PIPELINE", "SUVRBFTP", "BERKELEY FTP Standard Uptake Ratio Neocortical Composite Inferior Cerebellar Gray Matter", 6,
  "VR", "FBP", NA, NA, "FBP VISUAL CLASSIFICATION", "VRFBP", "FBP Qualitative Visual Classification", 7,
  "VR", "FBB", NA, NA, "FBB VISUAL CLASSIFICATION", "VRFBB", "FBB Qualitative Visual Classification", 8,
  "VR", "FTP", NA, NA, "FTP VISUAL CLASSIFICATION", "VRFTP", "FTP Qualitative Visual Classification", 9
)

This lookup may now be joined to the data and this is how the parameters will look like:

adpet <- adpet %>%
  derive_vars_merged_lookup(
    dataset_add = param_lookup,
    new_vars = exprs(PARAMCD, PARAM, PARAMN),
    by_vars = exprs(NVTESTCD, NVCAT, NVLOC, NVMETHOD, REFREG)
  )

Map AVAL and AVALC variables

Now that the parameter-level variables have been created, we can map the Analysis Value variables. For these parameters, no complex derivation is required. Note that AVALC should only be mapped if it contains non-redundant information.

adpet <- adpet %>%
  mutate(
    AVAL = NVSTRESN,
    AVALC = if_else(
      is.na(NVSTRESN) | as.character(NVSTRESN) != NVSTRESC,
      NVSTRESC,
      NA
    )
  )

Having established the foundation of a global PET Scan Analysis Dataset (adpet), we can now differentiate between tau and amyloid data.

ADTPET (Tau PET Scan Analysis Dataset)

Tau PET scan data generally do not require any further derivations, and the conversion of tau SUVR to CenTauR will be available in the next release (0.2.0) of this package. Therefore, if producing an ADTPET, we can now filter the global PET Scan Analysis Dataset adpet to retain only the tau tracer records.

adtpet <- adpet %>%
  filter(AGCAT == "TAU TRACER")

ADAPET (Amyloid PET Scan Analysis Dataset)

If producing an ADAPET, we can now filter the global PET Scan Analysis Dataset adpet to retain only the amyloid tracer records.

adapet <- adpet %>%
  filter(AGCAT == "AMYLOID TRACER")

Convert amyloid SUVR to Centiloid

SUVR (Standardized Uptake Value Ratio) is a widely used semi-quantitative measure in PET imaging that estimates the level of radiopharmaceutical tracer uptake (such as amyloid) in the brain, relative to a reference region. While useful, SUVR has certain limitations (which we will not detail here) that affect its comparability across different tracers and studies.

To address this, the Centiloid (CL) scale was developed to standardize amyloid PET measurements. The scale is anchored such that:

  1. 0 CL represents amyloid-negative individuals (usually younger, amyloid-negative individual).

  2. 100 CL corresponds to the amyloid burden observed in patients with Alzheimer’s disease.

This standardization enables consistent interpretation of amyloid burden across imaging protocols, tracers, and clinical studies.

In the following steps, we will incorporate the conversion from amyloid SUVR to Centiloid using the function admiralneuro::compute_centiloid().

For our examples, we divide the process into two main steps:

1. Define the Grouping Criteria

The first step is to specify how to group the data using admiral::slice_derivation(). In this case, we group by patient, ADSL variables, Analysis Date (ADT), Relative Analysis Day (ADY), and Visit (VISIT). We also retain records with missing Analysis Values (is.na(AVAL)) by setting keep_nas = TRUE.

2. Convert SUVR to Centiloid

The second step involves converting the original SUVR parameters to Centiloid using admiralneuro::compute_centiloid(). The formula is based on tracer, pipeline and reference region; therefore we apply filtering conditions using admiral::derivation_slice(), as following:

  1. Based on the (SUVRBFBB) tracer, the (BERKELEY FBB SUVR PIPELINE) method, and the (Whole Cerebellum) reference region.

  2. Based on the (SUVRBFBP) tracer, the (BERKELEY FBP SUVR PIPELINE) method, and the (Whole Cerebellum) reference region.

  3. Based on the (SUVRAFBB) tracer, the (AVID FBB SUVR PIPELINE) method, and the same reference region (Whole Cerebellum).

  4. Based on the (SUVRAFBP) tracer, the (AVID FBP SUVR PIPELINE) method, and the same reference region (Whole Cerebellum).

The new parameter values which include AVAL, PARAMCD, PARAM, and AVALU, are assigned within the args argument of this function.

adapet <- adapet %>%
  slice_derivation(
    derivation = derive_param_computed,
    args = params(
      by_vars = c(
        get_admiral_option("subject_keys"),
        adsl_vars,
        exprs(ADT, ADY, VISIT)
      ),
      keep_nas = TRUE
    ),
    derivation_slice(
      filter = (PARAMCD == "SUVRBFBB") & (NVMETHOD == "BERKELEY FBB SUVR PIPELINE") & (REFREG == "Whole Cerebellum"),
      args = params(
        parameters = c("SUVRBFBB"),
        set_values_to = exprs(
          AVAL = compute_centiloid(
            tracer = "18F-Florbetaben",
            pipeline = "BERKELEY FBB SUVR PIPELINE",
            ref_region = "Whole Cerebellum",
            suvr = AVAL
          ),
          PARAMCD = "CLBFBB",
          PARAM = "Centiloid (CL) based on BERKELEY FBB",
          AVALU = "CL"
        )
      )
    ),
    derivation_slice(
      filter = (PARAMCD == "SUVRBFBP") & (NVMETHOD == "BERKELEY FBP SUVR PIPELINE") & (REFREG == "Whole Cerebellum"),
      args = params(
        parameters = c("SUVRBFBP"),
        set_values_to = exprs(
          AVAL = compute_centiloid(
            tracer = "18F-Florbetapir",
            pipeline = "BERKELEY FBP SUVR PIPELINE",
            ref_region = "Whole Cerebellum",
            suvr = AVAL
          ),
          PARAMCD = "CLBFBP",
          PARAM = "Centiloid (CL) based on BERKELEY FBP",
          AVALU = "CL"
        )
      )
    ),
    derivation_slice(
      filter = (PARAMCD == "SUVRAFBB") & (NVMETHOD == "AVID FBB SUVR PIPELINE") & (REFREG == "Whole Cerebellum"),
      args = params(
        parameters = c("SUVRAFBB"),
        set_values_to = exprs(
          AVAL = compute_centiloid(
            tracer = "18F-Florbetaben",
            pipeline = "AVID FBB SUVR PIPELINE",
            ref_region = "Whole Cerebellum",
            suvr = AVAL
          ),
          PARAMCD = "CLAFBB",
          PARAM = "Centiloid (CL) based on AVID FBB",
          AVALU = "CL"
        )
      )
    ),
    derivation_slice(
      filter = (PARAMCD == "SUVRAFBP") & (NVMETHOD == "AVID FBP SUVR PIPELINE") & (REFREG == "Whole Cerebellum"),
      args = params(
        parameters = c("SUVRAFBP"),
        set_values_to = exprs(
          AVAL = compute_centiloid(
            tracer = "18F-Florbetapir",
            pipeline = "AVID FBP SUVR PIPELINE",
            ref_region = "Whole Cerebellum",
            suvr = AVAL
          ),
          PARAMCD = "CLAFBP",
          PARAM = "Centiloid (CL) based on AVID FBP",
          AVALU = "CL"
        )
      )
    )
  )

This is how the parameters and newly converted parameters will look like:

Derive Criterion Flag Variables for Amyloid Categories

Having converted the SUVR to Centiloid, we can now derive the criterion flags for the amyloid categories.

We will use Centiloid cutoff of 24.1 to define amyloid negative and positive categories. The 24.1 Centiloid cutoff comes from an autopsy-validated study in which amyloid PET SUVR values were compared with post-mortem brain amyloid plaque counts (Clark et al., 2012). The SUVR threshold corresponding to moderate-to-frequent neuritic plaques under the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) criteria was mapped to the Centiloid scale, yielding 24.1 CL as the pathology-based definition of amyloid positivity (Navitsky et al., 2018). Centiloid >= 24.1 is considered amyloid positive, while Centiloid < 24.1 is considered amyloid negative.

To do this, we will use the admiral::derive_vars_crit_flag() function. Since we want to derive these flags specifically for the Centiloid parameters, we will apply the derivation only to those records using admiral::restrict_derivation().

adapet <- adapet %>%
  restrict_derivation(
    derivation = derive_vars_crit_flag,
    args = params(
      crit_nr = 1,
      condition = if_else(PARAMCD %in% c("CLBFBB", "CLBFBP", "CLAFBB", "CLAFBP"), AVAL < 24.1, NA),
      description = "CENTILOID < 24.1",
      values_yn = TRUE # To get "Y", "N", and NA for the flag
    ),
    filter = PARAMCD %in% c("CLBFBB", "CLBFBP", "CLAFBB", "CLAFBP")
  )

This is how the criterion flags will look like:

Remaining PET Scan Analysis Dataset Set-up

The admiral Creating a BDS Finding ADaM vignette covers all the steps that are not shown here, such as merging the timing variables, analysis flags, etc.

Example Scripts

ADaM Sourcing Command
ADTPET admiral::use_ad_template("ADTPET", package = "admiralneuro")
ADAPET admiral::use_ad_template("ADAPET", package = "admiralneuro")

References

Clark, C. M. et al. (2012). Cerebral PET with florbetapir compared with neuropathology at autopsy for detection of neuritic amyloid-β plaques: a prospective cohort study. The Lancet Neurology, 11(8), 669–678.

Navitsky, M. et al. (2018). Standardization of amyloid quantitation with florbetapir standardized uptake value ratios to the Centiloid scale. Alzheimer’s & Dementia, 14(12), 1570–1577.