Note that backward compatibility of function arguments will not be (near-)guaranteed until version 1.0. Argument and function changes will continue until then. These will be especially noticeable around the inclusion of IV NCA parameters and additional specifications of the dosing including dose amount and route.

PKNCA 0.12.0

Breaking changes

• PKNCA will now give an error when there are unexpected interval columns. The keep_interval_cols option can be used to mitigate this error.
• NA results from calculating c0 will now add an exclusion reason.
• AUC for intravenous dosing (all the auciv* parameters) now more robustly calculate c0 and does not raise an error when is.na(c0) (#353).
• Manual calculation of half.life no longer allows negative half-live values (#373).

New Features

• PKNCAconc() and PKNCAdose() can now accept unit specifications as either column names or units to use (#336).
• PKNCA options can now use tmax as a reference for BLQ handling by using new names in the conc.blq argument (before.tmax,after.tmax)
• A new parameter count_conc_measured was added to enable quality checks, typically on AUC measurements. An associated exclusion function, exclude_nca_count_conc_measured() was also added.
• The PKNCAconc() arguments of include_half.life and exclude_half.life now allow NA values. If all values are NA, then no inclusion or exclusion is applied (the interval is treated as-is, like the argument had not been given). If some values are NA for the interval, those are treated as FALSE.
• group_vars() methods were added for PKNCAdata and PKNCAresults objects.
• If intervals have attributes on the columns, there will no longer be an error during parameter calculation, and the attributes are preserved (#381)
• When adding units, if some but not all units are provided, then an error will be raised. This error can be converted to a warning using the option allow_partial_missing_units = TRUE. (#398)
• A new function get_halflife_points() lets users know which points were used for half-life calculation. (#387)

Minor changes (unlikely to affect PKNCA use)

• PKNCA will now verify the intervals data.frame when creating PKNCAdata. The checking includes confirming intended column naming and ensuring the correct data types.
• PKNCA now contains a getGroups.PKNCAdata function to capture grouping columns.
• Duplicate data checks now account for excluded rows. So, if a row is duplicated and all but one of the duplicated rows is excluded, it is not an error. (#298)
• Removed native pipes (|>) so that PKNCA will work with older versions of R (#304).
• Missing dosing times to pk.calc.c0() will not cause an error (#344)

PKNCA 0.11.0

• PKNCA will now indicate the number of observations included in a summary (“n”) when it is not the same as the number of subjects included in the summary (“N”) and the caption will also indicate the definition of “N” and “n”. Note that counting of “n” includes all non-missing values that were not excluded from summarization; this will included all zeros that are e.g. excluded from geometric statistics.
  • If n == 1, spread statistics are no longer calculated in the summary.
• A new AUC integration method, “lin-log”, has been added using the linear method through tmax and log after tmax, with required exceptions for zeros (fix #23)
• The parameter vd was removed (it was not specific like vz or vss, and it was effectively a duplicate of vz). Use vz, instead.
• The count_conc NCA parameter was added to assist in data quality checking.
• Subject count (“N”) previously counted the number of rows of data, but in unusual circumstances, the number of subjects in an NCA result could be fewer than the number of rows. Now the number of subjects is counted (fix #223).
• Extra column in the intervals argument to PKNCAdata() will no longer cause an error (fix #238)
• Many new assert_* functions were added to standardize input checking in the style of the checkmate library.
• Interpolation of zero concentrations in the middle of a set of concentrations is now more extensively supported.
• PKNCA has begun the process of deprecating dots in favor of underscores in function and parameter names. Functions with dots instead of underscores should continue to work for the foreseeable future (until version 1.0) with warnings.
• AUCint will now extrapolate the AUC beyond Tlast using logarithmic extrapolation, regardless of the method used (fix #203).
• Imputation will now automatically search for a column named “impute” in the interval definition (fix #257).
• Imputation now can look outside the concentration-time of the interval to the full concentration-time profile for the group with the conc.group and time.group arguments to the imputation functions. And, PKNCA_impute_method_start_predose() imputation performs more reasonably when the end of the interval is infinite.
• A progress bar is now available via the PKNCA.options(progress = ) option (fix #193).
• Additional versions of average concentration based on AUCint are now available (fix #45).
• A new option “keep_interval_cols” was added to allow keeping a column from the intervals in the NCA results. Note that these are not included in the summary groups by default.
• A new argument “filter_requested” for as.data.frame.PKNCAresults() allows you to filter only to requested results from a PKNCAresults object.
• pknca_units_table() now has four new arguments to allow for simplified automatic conversion from source units to desired reporting units, concu_pref, doseu_pref, amountu_pref, and timeu_pref (#197)
• Extraction of PKNCA objects from within other PKNCA objects is now supported by various as_PKNCA* functions like as_PKNCAconc() which can be used to extract the concentration data from within a PKNCAdata or PKNCAresults object (#278)
• A new “totdose” parameter gives the total dose administered during an interval
• You may exclude parameters from a summary with the new drop_param argument to summary() for PKNCAresults objects.
• The as.data.frame() method for PKNCAresults objects has a new filter_excluded argument to remove excluded results from the extracted data.frame. The default behavior is to keep the excluded results with the exclude column indicating the reason they were excluded.

Bugs fixed

• superpostion() and the interp.extrap.conc() family of functions now respect the interpolation and extrapolation types requested rather than using default.
• Concentration extrapolation with extrapolate.conc() using the “AUCall” method now has decreasing instead of increasing concentrations (#249).
• The aucint.inf.obs parameter when calculated with all zero concentrations returns zero and aucint.inf.pred returns NA_real_ (#253)

Breaking changes

• The arguments interp.method and extrap.method have been replaced with method and auc.type in the interp.extrap.conc() family of functions for consistency with the rest of PKNCA (fix #244)
• The AIC.list() function is no longer exported (it was never intended to be an external function).
• The depends argument to add.interval.col() must either be NULL or a character vector.
• The names of the fun.linear, fun.log, and fun.inf arguments to pk.calc.auxc were changed to use underscores. (If you were using those directly, please reach out as they were intended to be internal arguments, and I would like to know your use case for changing them.)
• check.conc.time() is defunct (it was never intended to be an external function). It has been replaced by assert_conc(), assert_time() and assert_conc_time().
• The clast.obs parameter is now zero when all concentrations are zero (see #253 for part of the reason).
• (This is not likely to be important for most users.) The business... functions (e.g. business.geomean()) now include an attribute in non-NA results with n, the number of values included in the statistic.

Changes under the hood

• Multiple changes were made to speed up calculations. These will mainly be noticed when performing NCA on many subjects (for instance, following simulations). None of these should have external effects that users will notice:
  • Adding in dependent parameters required for requested parameters is now more efficient (approx 40% time savings)
  • Sorting interval dependencies happens less often (approx 5% time savings)
  • Determining if a parameter is needed for calculation when looking across all parameters is more efficient (negligible time savings)
• An internal change was made to make AUC integration and concentration interpolation simpler and simplify the ability to create new AUC integration or concentration interpolation methods

PKNCA 0.10.2

• A minor change to pk.calc.aucpext() was made so that it now returns NA_real_ instead of NaN.
• A minor change was made so that AUC and amount excreted (ae) calculations will provide an exclusion reason the result is NA.
• A minor new feature makes the specification of imputation easier. You can give the imputation with or without the “PKNCA_impute_method_” prefix. So, “PKNCA_impute_method_start_predose” and “start_predose” are equivalent.

PKNCA 0.10.1

• A new parameter aucabove.trough.all was added to calculate the NCA above the trough concentration.
• Testing updates were made to work with dplyr version 1.1.0 (fix #198)
• Internal changes to how columns are identified were made, and the parseFormula function was subsequently removed (parseFormula was never intended for external use).

PKNCA 0.10.0

Bugs Fixed

• When calculating AUC with only a single concentration measurement, NA is now returned instead of 0. (fix #176)

New Features

• Initial support for unit assignment and conversion has been added. See the units argument to the PKNCAdata() function and the function pknca_units_table().
• Initial support for imputation has been added. See the impute argument to the PKNCAdata() function and the Data Imputation vignette.
• With the addition of units, several outputs now will differ, if units are used:
  • summary() on a PKNCAresults object shows the units in the column heading.
  • When running as.data.frame() on a PKNCAresults object with the argument out.format="wide", if standardized units values are available, they will be used. And if any unit are available, they will be in the column names.
• Summary tables with units use the “pretty_name” for a parameter which is intended for clearer representation in reports. “pretty_name” use can be controlled with the “pretty_names” argument to summary().
  • Note that the pretty names themselves may be modified to help clarify and/or shorten the names to make the table heading more useful. If you intend to modify column headers programmatically, set pretty_names=FALSE when calling the summary() function.
• New, IV AUC calculation methods have been added.
• pk.calc.time_above() now uses the default AUC calculation method for interpolation of time above. And, it can use ‘lin up/log down’ interpolation.
• PKNCA can now calculate parameters that require extra information by adding the extra information to the intervals data.frame. For example, add conc_above as a column to the intervals to allow calculation of time_above. With this change, the “conc_above” PKNCA.options() value has been removed.
• Added dplyr joins, filter, mutate, group_by, and ungroup to allow modification of PKNCA objects after creation. (Note that these functions will make the provenance no longer match for PKNCAresults objects.)

Breaking Changes

• Some functions that were intended to be internal were removed:
  • All getData() functions were removed.
  • The getDataName() function for PKNCAdata objects was removed.
• interpolate.conc() and interp.extrap.conc() now give more errors with missing (NA) input. This should not affect typical NCA (where NA values are dropped), but it may affect direct calls to the functions themselves.

Internal Breaking Changes (these should not affect PKNCA users)

• print.parseFormula() was removed from the package.

PKNCA 0.9.5

• The internals of how PKNCA performs calculations had a significant update. The only user-visible change should be that PKNCA does not perform parallel computations as of this version. Parallel computation is planned to return in the near future.
  • Breaking change: As part of this change, the split methods for PKNCAconc and PKNCAdose objects were removed along with the merge.splitList function.
• Single-subject (ungrouped) analysis works without creating a dummy group (#74)
• PKNCAconc objects are checked earlier for valid data (#154)
• Add time_above parameter to calculate time above a given concentration.
• Fix numeric BLQ replacement when the value is a number and different values are given for first, middle, and last (related to #145). This only affects datasets where BLQ is being replaced with a nonzero value (not a common scenario).
• Fix issue where intervals could not be tibbles (#141)
• Fix minor issue where only the first exclusion reason would show in the exclusion column and other reasons would be ignored (#113). Note that the impact of this bug is minimal as the result would have been excluded from summaries for the first reason found, but if there were multiple reasons for exclusion the subsequent reasons would not be recorded.

PKNCA 0.9.4

• Additional changes required for compatibility dplyr version 1.0 and CRAN checks. No functionality changed.
• Minor typographical and documentation consistency cleanups throughout.

PKNCA 0.9.3

• Changes required for compatibility dplyr version 1.0. No functionality changed.

PKNCA 0.9.2

• New feature: the time_calc() function will help convert time values to be relative to events (such as calculating time after and before doses)
• Fix issue summarizing results when “start” and “end” are dropped and there are multiple interval rows matched for a single group.
• Enable exclusions to be prevented when the input arguments suggest exclusion, but the parameter calculating function may be aware of better information about exclusion.
• Ensure that exclusions are maintained if an earlier parameter is excluded during the initial parameter calculations (Fix #112).
• Two-point half-life calculation works and adjusted r-squared gives a warning instead of an error with 2 points (Fix #114).
• Half-life calculation time was decreased by using .lm.fit() instead of lm() decreasing time for a full NCA run by ~30% (and half-life by ~50%).
• For R version 4.0, much more care was taken not to create factors from strings unless required (see https://developer.r-project.org/Blog/public/2020/02/16/stringsasfactors/index.html)

PKNCA 0.9.1

• Correct vignette building.

PKNCA 0.9.0

• Breaking Change: plot.PKNCAconc() was moved to the pknca.reporting package (https://github.com/billdenney/pknca.reporting)
• Breaking Change: summary.PKNCAresults() now provides a caption including the summary method for each parameter. If you change summary functions using PKNCA.set.summary(), you must now use the description option to set the description of the summary.
• Breaking Change: ptr now accurately uses ctrough instead of cmin (fix #106)
• Issue fixed where aucint* calculations now respect BLQ and NA rules like other calculations. (#104)
• When half.life is not calculated due to insufficient number of points (default < 3), an exclusion reason is added. (#102)
• tibbles now work as the interval argument for PKNCAdata() (fix #72)
• Issue fixed with summarization of data that has exclusions. Exclusions are now correctly handled as missing instead of never calculated.
• parseFormula now internally uses NULL for no-group formula definitions.
• signifString and roundString now have sci_range (deprecating si_range) and sci_sep arguments.
• Documentation is improved, especially around the selection of parameters for intervals.
• Multiple dose data with a single concentration measurement no longer generates an error (fixes #84).
• The “start” and “end” columns may now be dropped from the summary of PKNCAresults objects.
• PKNCAdata() is more restrictive on unknown arguments issuing an error when unknown arguments are present.
• intervals argument to PKNCAdata() may now be a tibble (fixes #72).
• Documentation has been extensively updated (fixes #81).
• CRAN changes: Vignettes now better respect not loading suggested packages. Tests are now more permissive in timing.

PKNCA 0.8.5

• Cleaned AUCint names
• Added dose-count within interval (to warn of multiple doses within an interval)
• Various documentation updates
• signifString and roundString now by default use scientific notation for values >=1e6 and <=1e-6
• Fix bug in option handling within pk.nca (Fix #68)

PKNCA 0.8.4

• Added AUCint flavors
• Parameter names for NCA parameters will likely be changing in the next version; code will still work, but some calculation methods and therefore results may be subtly different. These changes will be fully documented.)

PKNCA 0.8.2

• BACKWARD INCOMPATIBILITY: The function supplied to the exclude argument ‘FUN’ now requires two arguments and operates on the level of a single group rather than the full object. The function can also return the reason as a character string instead of a logical mask of when to exclude data.

• BACKWARD INCOMPATIBILITY: Added back-end functionality to only require one function to handle many NCA parameters that are related (e.g. combine pk.calc.aucpext, pk.calc.aucpext.obs, pk.calc.aucpext.pred, etc.). If your current code calls a specific function (like pk.calc.aucpext.pred), you must change to using the generic function (like pk.calc.aucpext)

• BACKWARD INCOMPATIBILITY: Functions that previously may have returned Infinity due to dividing by zero (e.g. when AUC=0 calculating clearance) now return NA.

• Added Validation vignette.

• Corrected issue where time to steady-state with a single estimate may have given more than one estimated time to steady-state.

• Corrected issue with exclude handling where now a blank string is also accepted as included (not excluded).

• PKNCAconc now accepts a “volume” argument and pk.nca can now calculate urine/feces-related parameters (fe, ae, clr)

• exclude_nca* functions added (Fixes issue #20)

• Add manual half-life point selection (Fixes issue #18)

• Improved summary settings (Fixes issue #54)

• Add parameters for Ceoi and intravenous MRT

• Updated vignettes to improve clarity

• Added dose-normalized PK parameters (Fixes issue #41)

• Added checks to confirm that concentration and time are numeric (Fixes feature request #40)

• Improved test coverage

PKNCA 0.8.1

• A PKNCAdose is no longer required for calculations.
• Data may now be excluded from calculations.

PKNCA 0.8

This release is not backward compatible. The switch to observed and predicted-related NCA parameters (like aucinf.obs and aucinf.pred) changed the format of the intervals specification.

• Remove dependency on doBy library
• Dose-aware interpolation and extrapolation was added with the interp.extrap.conc.dose function.
• Added Clast.pred related NCA calculations
• Added N to summary of PKNCAresults
• Added parameter selection between Clast,observed and Clast,predicted across all parameters
• Enabled PKNCAdose to be specified with one-sided formula
• Improved error reporting so that the group and time (interval specification) is reported in addition to the error.
• PKNCAdose now allows route of administration and IV infusion parameters of rate/duration to be specified

PKNCA 0.7.1

• Updated vignettes
• Standardize rounding and significance with missing values in signifString and roundString
• Enable wide data output with as.data.frame(PKNCAresults, out.format=“wide”)
• Correct calculation of Vz
• Various CRAN-related cleanups

PKNCA 0.7

• Features added
  • Additional PK parameters to support IV dosing added
  • Fix #11, Intervals can be specified manually, and will apply across appropriate parts of the grouping variables
  • Enable dose and dose.time as parameters to NCA calculations
  • More NCA parameters are calculated, especially related to IV dosing
  • Fix #8, Reporting times for time-based parameters are now within the current interval rather than since first dose (e.g. Tmax on day 14 should be between 0 and 24 not 2724+c(0, 24))
  • Added several vignettes
• Bugs fixed
  • Dosing without concentration is probably placebo; warn and continue
  • Fix #6, make merge.splitByData work with more than one dosing level
  • Export some generic classes that were not previously exported to simplify their use
  • Superposition extensions when lambda.z cannot be calculated
  • Significance rounding into character strings works when the rounding moves up one order of magnitude.
  • Fix #9, summarization of parameters that are not calculated show not calculated instead of missing.

PKNCA 0.6

First release targeting CRAN