Changelog

Bug Fixes

• Fixed missing label for TRTEDTM in tern datasets.

Miscellaneous

• Implemented the lubridate package for date variables in tern datasets.

Enhancements

• Added summarize_glm_count function to analyze count data using a linear model.
• Added legend to g_step.
• Added formatting functions format_fraction_fixed_dp and format_count_fraction_fixed_dp with fixed single decimal place in percentages.
• Added na_level and labelstr arguments to summarize_vars_in_cols.
• Added analyze_num_patients to include summary at the beginning that does not repeat when paginating.
• Added h_row_first_values function as a more general helper function to retrieve first values from specific rows.
• Added option to remove "(n)" suffix from unique_count labels for s_num_patients.
• Added options to g_km to annotate with statistics (annot_stats) and add corresponding vertical lines (annot_stats_lines).

Bug Fixes

• Fixed bug causing incorrect ordering of numeric grade levels when missing grades are present in s_count_occurrences_by_grade.
• Refactored summarize_vars_in_cols to work with pagination machinery.
• Fixed bug to allow passing of conf_level argument to emmeans::contrast() in s_ancova.
• Fixed bugs in rtables_access.R caused by not checking for specific combinations (also the standard values that were never used) of column indices and names.
• Fixed single applicable record bug in count_abnormal_by_grade.
• Fixed bug in add_rowcounts that caused all row count row values to count as zero.
• Fixed bug in h_col_indices causing an error when pruning with combination columns.
• Fixed bug in test_proportion_diff missing argument for var_labels.

Documentation and Tests

• Added more tests to increase code coverage.
• Created separate documentation files for functions in different sections of pkgdown reference.
• Created separate .R files for logistic regression and cox regression helper functions.
• Fixed table tests using analyze_num_patients to generate an initial summary so there is no repetition when paginating.
• Updated tests to use testthat 3rd edition and replaced applicable tests with snapshot testing.
• Updated summarize_ancova examples to use iris dataset instead of scda data.
• Created vignette which saves cached synthetic CDISC dataset files to the data/ folder and generated cached synthetic datasets.
• Updated all examples/tests to use datasets from the data/ folder instead of scda datasets.
• Removed all template tests from tern. These tests are in internal repo scda.test.

Miscellaneous

• Renamed summarize_vars_in_cols to analyze_vars_in_cols to reflect the appropriate analyze logic.
• Removed redundant summary_in_cols helper functions.
• Exported function format_xx.
• Replaced deprecated ggplot2 functions/arguments to fix warnings.
• Replaced deprecated function forcats::fct_explicit_na with forcats::fct_na_value_to_level.
• Removed deprecated wrap_text function and related files.
• Started deprecation cycle for footnotes functions.

New Features

• Added stratified Newcombe and stratified Wilson statistics to estimate_proportion and estimate_proportion_diff with relative tests.
• Added stat_mean_pval, a new summary statistic to calculate the p-value of the mean.
• Added statistic mean_se (mean with standard error) for summarize_variables and related functions.
• Introduced again Rdpack for references.

Enhancements

• Redesign of data handling in tests by removing repetitive data loads and library calls.
• Added DescTools::BinomDiffCI function within tern.
• Added new parameter to summarize_logistic to specify which pivoted value to use during analysis.
• Updated s_coxph_pairwise to generate log-rank p-value using original log-rank test instead of Cox Proportional-Hazards Model.
• Implemented nestcolor in all examples by adapting g_km, g_ipp, g_waterfall, g_step, g_lineplot, and g_forest.
• Added parameters interaction_y and interaction_item in ANCOVA to make the interaction calculations available.
• Added new parameter footnotes to add footnotes to g_km.

Migration from assertthat to checkmate

• Implemented checkmate::assert_vector, checkmate::assert_set_equal, and checkmate::assert_int to check vector type, length, and values.
• Replaced with standard assertions from checkmate the following functions: all_elements_in_ref, is_df_with_nlevels_factor, is_df_with_no_na_level, is_proportion_vector, is_quantiles_vector, is_character_or_factor, is_nonnegative_count, is_valid_character, assert_character_or_factor, assert_equal_length and has_tabletree_colnames.
• Modified is_proportion, is_equal_length, is_df_with_no_na_level, is_df_with_nlevels_factor, is_variables, is_df_with_variables, is_df_with_factors, is_valid_factor to use assertion logic.
• Added more verbose warnings from as_factor_keep_attributes.
• Made assert_df_with_factors and assert_proportion_value internal functions.
• Renamed assertthat.R and test-assertthat.R to utils_checkmate.R and test-utils_checkmate.R.

Documentation and NAMESPACE Polishing

• Added stable badges for:
  • count_abnormal_by_marked (reference to abnormal_by_marked), count_abnormal_lab_worsen_by_baseline and h_adlb_worsen (reference to abnormal_by_worst_grade_worsen_from_baseline), count_abnormal_by_worst_grade (reference to abnormal_by_worst_grade), to_string_matrix, tidy.summary.coxph, tidy.step, surv_timepoint, (reference to survival_timepoint), surv_time (reference to survival_time), coxph_pairwise (reference to survival_coxph_pairwise), extract_survival_subgroups and tabulate_survival_subgroups (reference to survival_duration_subgroups), extract_survival_biomarkers and tabulate_survival_biomarkers (reference to survival_biomarkers_subgroups), control_summarize_vars, s_summary and a_summary (reference to summarize_variables) and kept the S3 method tree.
  • summarize_patients_exposure_in_cols, summarize_num_patients with s_num_patients, s_num_patients_content, summarize_num_patients.
  • count_cumulative, count_missed_doses, count_patients_events_in_cols, summarize_colvars, summarize_change, summarize_ancova,as.rtable, color_palette, add_footnotes.
  • (statistical function controls) control_coxreg, control_coxph, control_incidence_rate, control_lineplot_vars, control_surv_time, control_surv_timepoint, control_logisitic, control_step.
  • stat_mean_ci, stat_median_ci, split_cols_by_groups, explicit_na, sas_na, extract_rsp_subgroups, tabulate_rsp_subgroups, extract_rsp_biomarkers, tabulate_rsp_biomarkers, keep_rows, keep_content_rows, has_count_in_any_col, has_fraction_in_cols, has_fraction_in_any_col, has_fractions_difference, test_proportion_diff, pairwise, logistic_regression, estimate_incidence_rate, control_incidence_rate (reference to incidence_rate), cut_quantile_bins, estimate_multinomial_rsp, decorate_grob_set, extreme_format, fit_rsp_step, fit_survival_step, footnotes, footnotes-set, format_count_fraction, format_fraction_threshold, formatting_functions, format_fraction, combination_function (S4 method), compare_variables (S3 method), kaplan_meier.
• Internal keywords added, export removed, _pkgdown.yml updated, and tern::: added for tests/examples/vignettes where present for the following functions:
  • (chain functions, reference to abnormal_by_marked) s_count_abnormal_by_marked, a_count_abnormal_by_marked.
  • (chain functions, reference to abnormal_by_worst_grade_worsen_from_baseline) a_count_abnormal_lab_worsen_by_baseline, s_count_abnormal_lab_worsen_by_baseline.
  • (chain functions, reference to abnormal_by_worst_grade) s_count_abnormal_by_worst_grade, a_count_abnormal_by_worst_grade.
  • (chain functions, reference to survival_timepoint) s_surv_timepoint, s_surv_timepoint_diff, a_surv_timepoint, a_surv_timepoint_diff.
  • (chain functions, reference to survival_time) s_surv_time, a_surv_time.
  • (chain functions, reference to survival_coxph_pairwise) s_coxph_pairwise, a_coxph_pairwise.
  • (chain functions, reference to survival_duration_subgroups) a_survival_subgroups.
  • (chain functions, reference to count_cumulative) s_count_cumulative, a_count_cumulative.
  • (chain functions, reference to count_missed_doses) s_count_nonmissing, s_count_missed_doses, a_count_missed_doses.
  • (chain functions, reference to count_patients_events_in_cols) s_count_patients_and_multiple_events, summarize_patients_events_in_cols.
  • (chain functions, reference to incidence_rate) s_incidence_rate, a_incidence_rate.
  • (cox regression helper) cox_regression_inter, decorate_grob_factory, draw_grob, estimate_coef.
  • summary_labels, summary_formats, s_count_patients_sum_exposure, a_change_from_baseline s_change_from_baseline, a_ancova, s_ancova, arrange_grobs, as_factor_keep_attributes, combine_levels, split_text_grob, groups_list_to_df, s_cox_multivariate, is_leaf_table, a_response_subgroups, range_noinf, has_count_in_cols, has_counts_difference, prop_chisq, prop_cmh, prop_schouten, prop_fisher, s_test_proportion_diff, a_test_proportion_diff, fct_discard, fct_explicit_na_if.

Bug Fixes

• Fixed retrieval method of stats::ancova output due to version inconsistency.
• Fixed tests to respect the new standard print for NA coming from rtables.
• Fixed error in tests coming from changes in formatters::var_labels.
• Fixed prop_diff functions to respect success responses (TRUE values).
• Fixed error coming from comparing factors vector to characters vector.
• Fixed empty vector exception for cut_quantile_bins.
• Fixed exception error when empty strings are present in pivoted columns (rtables split functions)
• Fixed bug in s_ancova causing an error when the first level of the arm factor is not the control arm.
• Fixed bug in s_abnormal_by_worst_grade when there is one PARAM level.
• Fixed bug in prop_diff_wald when selecting all responders, updated tests accordingly.
• Fixed bug in h_ancova that caused an error when deselecting all covariates.

Miscellaneous

• Added deprecated badge to g_mmrm.
• Removed internal function calls in examples (tern:::) and added dontrun to internal function examples.
• Removed warnings and messages to console occurring in examples and tests.
• Deprecated functions color_palette and h_set_nest_theme in favor of nestcolor::color_palette and nestcolor::theme_nest, respectively.
• Removed deprecated functions: color_palette, color_palette_core, h_set_nest_theme, s_cox_univariate.
• Removed deprecated mmrm functions: fit_mmrm, g_mmrm_diagnostic, g_mmrm_lsmeans, as.rtable.mmrm, h_mmrm_fixed, h_mmrm_cov, h_mmrm_diagnostic, tidy.mmrm, s_mmrm_lsmeans, s_mmrm_lsmeans_single, summarize_lsmeans.
• Renamed functions arm to study_arm and extract to extract_by_name.
• Renamed rtables.R to utils_rtables.R.
• Extracted cox_regression_inter into a separate file from cox_regression.
• Renamed estimate_incidence_rate.R to incidence_rate.R to match the documentation grouping name.
• Extracted control_incidence_rate into a separate file because it produces a separate documentation file.
• Added @md and removed @order from incidence_rate.R. Modified examples accordingly.
• Removed hyperlink from prop_schouten function documentation.
• Exported draw_grob function.

Fix

• h_split_by_subgroups documentation warning fix for wrong placing of example block

Documentation and NAMESPACE polishing

• Adopting the standard of badges for @description instead of every @descriptionIn function. Corrected accordingly summarize_variables_in_cols
• Added stable badge for g_lineplot, g_step, g_waterfall, cox_regression, score_occurrences, add_rowcounts, odds_ratio, count_occurrences, count_occurrences_by_grade, explicit_na, df_explicit_na, count_patients_with_event, decorate_grob, combine_groups, append_varlabels, univariate, stack_grobs, count_abnormal (reference to abnormal), count_abnormal_by_baseline (reference to abnormal_by_baseline).
• Internal keywords added, export removed, _pkgdown.yml polished and tern::: for tests, examples, and vignettes when present for the following functions:
  • (helper functions) h_format_row, h_map_for_count_abnormal
  • (utils functions) make_names, month2day, day2month empty_vector_if_na, aesi_label, n_available, format_xx, arm.
  • count_values_funs, prop_difference, combine_counts.
  • (chain functions) s_count_abnormal, a_count_abnormal.
  • (chain functions) s_count_abnormal_by_baseline, a_count_abnormal_by_baseline, d_count_abnormal_by_baseline.
• Deprecated s_cox_univariate function has now deprecated badge.

Enhancements

• Enhanced g_lineplot with table to automatically scale the table height and return a ggplot object.
• Enhanced g_ipp with caption argument and adjust the position.
• Enhanced prop_diff, tern function and related functions to be able to apply a continuity correction in the Newcombe method.
• Enhanced summarize_numeric_in_columns and summarize_variables to allow factor/character summary and to be able to summarize the number of BLQs in AVALC from ADPC dataset.
• Updated order of summarize variables stats in manual for order consistency.
• Added a sum option to summarize_variables.
• Use consistent color palette for plotting (stream by default).
• Enhanced h_pkparam_sort function with argument key_var to allow data with different column names.

Miscellaneous

• Updated test-table_aet02.R variant 12.
• Changed the scda data version to ‘2022-02-28’.
• Added a template to the pkgdown site.
• Removed package dependencies of grDevices, stringr, and viridisLite.
• Renaming summarize_numeric_in_columns to summarize_variables_in_columns.
• Renaming summarize_vars_numeric_in_cols to summarize_vars_in_cols.
• Fixed a bug where points on the g_lineplot plot were not connected when missing values.
• Updated the package authors.

Breaking changes

• Move MMRM into a separate package tern.mmrm.

New features

• Added h_pkparam_sort to order PK PARAM value based on the order of the dataset generated by d_pkparam().
• Added d_pkparam to generate PK parameter map for sorting.

Enhancements

• Changed the nudge_y argument of h_g_ipp to be dependent on the data, fixing an issue whereby the baseline labels were offset incorrectly.
• Enhanced stat_mean_ci and s_summary.numeric to calculate the geometric mean with its confidence intervals.

Miscellaneous

• Updated dependencies and internal adjustments after the rtables package refactor.
• Removed with_label, var_labels, and var_labels<- to resolve conflict with the formatters package, a new dependency.
• Added new “Introduction to tern” and “tern tabulation” vignettes.

New features

• Added h_map_for_count_abnormal to create the map used in trim_levels_to_map split function by calling this helper function. It supports two methods: one with all observed mapping, one with at least low limit above zero and at least one non missing high limit.
• Added s_summary_numeric_in_cols and summarize_vars_numeric_in_cols functions to generate summary statistics in columns, mainly used for PK datasets.
• Added five statistics to s_summary.numeric to use in s_summary_numeric_in_cols.

Enhancements

• Enhanced functions tabulate_survival_subgroups and tabulate_rsp_subgroups (Survival Duration and Best Response analyses) to calculate N-s based on the records considered to create the model.
• Enhanced the function estimate_proportion and related functions to be able to apply a continuity correction in the Wilson method.
• Refactored count_abnormal_by_marked and related statistics and formatting functions to use a more efficient layout with .spl_context argument used for determining denominators and with trim_levels_to_map split function under split_rows_by to show the desired levels in the table. This is a breaking change.
• Refactored count_abnormal_by_worst_grade and related statistics and formatting functions to use a more efficient layout with .spl_context argument used for determining denominators and with trim_levels_to_map split function under split_rows_by to show the desired levels in the table. This is a breaking change.
• Refactored count_abnormal function and related statistics and formatting functions to use a more efficient layout with trim_levels_to_map split function under split_rows_by to show the desired levels in the table. Also updated abnormal argument to be able to consider more than one level for each direction. This is a breaking change.
• Enhanced the function estimate_incidence_rate and related functions to consider the week as time unit for data input.

Bug fixes

• Fixed bug in assertthat functions that output wrong data frame names and limited length of failure message outputs.

Miscellaneous

• Removed dependency on utils.nest by using the checkmate and purrr packages for validation and moved get_free_cores and skip_if_too_deep functions from utils.nest into tern.

New features

• Added functions to estimate continuous biomarker effects across subgroups for survival and binary response endpoints, used to produce corresponding forest plots, see survival_biomarkers_subgroups and response_biomarkers_subgroups.
• Added g_lineplot plot function, including new h_format_row helper function and control_lineplot_vars function. Removed g_summary_by.
• Added new safety helper function h_stack_by_baskets to stack events in SMQ and/or CQ basket flag in ADAE data set.

Enhancements

• Added a couple of new statistics to s_summary.numeric. Added names attribute to each element of the final list returned by the s_summary.numeric function. Added summary_formats and summary_labels helper functions.
• Added option to also convert logical variables to factor variables in df_explicit_na.
• Refactored h_append_grade_groups to improve its flexibility, robustness and clearness, and to make sure the result is ordering according to the order of grade_groups. Also, added remove_single argument which controls whether the elements of one-element grade groups are in the output or removed.
• Added var_labels and show_labels arguments to count_occurrences and count_patients_with_flags to allow for creation of a title row.
• Added na_level argument to count_abnormal_by_baseline.
• Updated h_append_grade_groups to no longer fill-in empty grade groups with zeros.

Bug Fixes

• Fixed prop_diff_cmh to handle edge case of no FALSE (or TRUE) responses.
• Enhanced g_mmrm_diagnostic to improve error handling when data is not amenable to the Locally Weighted Scatterplot Smoothing.
• Fixes in g_km:
  • Plot can now display any combination of the annotation tables for number of patients at risk, median survival time, and CoxPH summary.
  • Function will return a warning instead of an error if the arm variable includes a single level and annot_coxph = TRUE.
  • Lines in the plot now start at time 0 and probability 1.
  • Category labels can include the equals sign.

Miscellaneous

• Fixed day2month and month2day to work with NA data.
• Added parameters for stat_mean_ci and stat_median_ci so that they may return different outputs.
• Added functionality in h_row_counts to handle analysis rows with NULL cells.
• Updated LICENCE and README with new package references.
• Added error_on_lint: TRUE to .lintr.

New features

• Added new safety layout functions:
  • count_abnormal_by_marked tabulates marked laboratory abnormalities.
  • summarize_patients_exposure_in_cols tabulates patient counts and sum of exposure across all patients.

Enhancements

• Enhanced mmrm related functions for fitting models without arm variable.
• Updated cox_regression to work without covariates. Also in case of interaction model summary, p-values for main effect coefficients are no longer displayed.
• Descriptive statistics returned by summarize_vars now include quantiles. summarize_vars now accepts the control function control_summarize_vars to specify details about confidence level for mean and median and quantile details. The control argument replaces conf_level.
• Added var_labels and show_labels arguments to count_occurrences_by_grade.
• Changed indent argument in append_varlabels to accept non-negative integer to represent the indent space defined by user. Previous calls with Boolean indent will do an integer conversion and produce a warning.

Bug Fixes

• Corrected tabulate_survival_subgroups and related survival forest plot functions to use total number of events, instead of observations, as default for scaling the symbol sizes in the plot. (The user might still use total number of observations manually if they wish to do so.)
• Helper function h_adsl_adlb_merge_using_worst_flag will now impute BTOXGR for missing visits.

Miscellaneous

• Deleted count_abnormal_by_worst_grade_by_baseline and its related statistic and analysis functions as a simpler design will create lab abnormality tables.
• Switched data in examples to use scda instead of random.cdisc.data package.

New Features

• Added Subgroup Treatment Effect Pattern (STEP) model fitting functions fit_rsp_step and fit_survival_step, the corresponding tidy method tidy.step as well as the graph function g_step.
• Added new layout function compare_vars which compares variables of different types between columns and produces a p-value for the comparison to the reference column. Function built on top of the summarize_vars functionality.
• Added utility functions:
  • cut_quantile_bins cuts a numeric vector into quantile bins.
  • fct_collapse_only collapses levels of a factor and keeps those in the order provided.
  • fct_explicit_na_if inserts explicit missings in a factor based on a condition.
  • range_noinf is a kind of a wrapper function of base::range. It returns c(NA, NA) instead of c(-Inf, Inf) for zero-length data.

Enhancements

• Cox regression via fit_coxreg_univar and fit_coxreg_multivar is now also possible without treatment arm. In the univariate case this means that it fits separate univariate models for the provided covariates and tabulation of corresponding effect estimates can later occur.
• Added fraction in result returned by s_count_occurrences. It contains a list of numerators and denominators with one element per occurrence.
• Updated sum_num_patients and count_occurrences for the result unique and count_fraction to return (0, 0) when input is empty.
• Added a new argument groups_lists to extract_survival_subgroups, extract_rsp_subgroups and associated helper functions which allows to group factor levels of subgroup variables into manually defined groups, enhancing the flexibility of the resulting forest graphs.
• Forest graph function g_forest now extracts default arguments from attributes of the input table produced by tabulate_rsp_subgroups and tabulate_survival_subgroups so that the user does not have to do this manually anymore.
• In g_km:
  • Remove arm variable name from arm labels in plot legend and annotation tables.
  • Show symbol used to mark censored cases and match order of arm variable factor levels in the legend.
  • Display hazard ratio and its confidence interval to two decimal places.
  • Updated default position of hazard ratio table to stay on the left bottom corner but above x-axis.
  • Use arm value as strata name in when treatment arm variable has a single level.
• Updated s_surv_time function to use a newly created function range_noinf instead of base::range.
• New argument no_fillin_visits added to h_adsl_adlb_merge_using_worst_flag to specify excluded visits from the post-baseline worst toxicity grade output. Improved h_adsl_adlb_merge_using_worst_flag to include variables shared between adsl and adlb, along with PARAM, PARAMCD, ATOXGR, BTOXGR and optionally AVISIT, AVISITN when by_visit = TRUE. Prior output contained USUBJID, ARMCD, PARAMCD, ATOXGR, and BTOXGR.

Bug Fixes

• Fix bug in s_surv_timepoint for cases when there are zero patients at risk.
• Modified stat_median_ci function so that when passing empty var with empty name, no row names contain missing values error would show.

Miscellaneous

• Deprecated s_cox_univariate function, use fit_coxreg_univar function instead.
• Updated default formats of hr and hr_ci in a_coxph_pairwise and median in s_surv_time to align with STREAM.
• Updated the pre-processing code in the files test-table_ttet01.R and test-table_dort01.R to make sure the analysis variable EVNT1 has both levels of the factor defined.
• Improved error message when number of levels in a factor variable in a data frame is not as expected.

New TLGs

• t_summary and methods for data.frame, numeric, logical, character, factor, and Date objects.
• t_events_per_term_id, t_events_per_term_grade_id: Adverse Events & Concomitant Treatment Tables.
• t_max_grade_per_id, t_count_unique, t_events_summary elementary tables used for the Adverse Events & Concomitant Treatment Tables.
• g_waterfall: Horizontal Waterfall Plot.

New Helper Functions

• decorate_grob, decorate_grob_set, decorate_grob_factory, splitTextGrob.
• stack_grobs, arrange_grobs, draw_grob.

TLG changes

• t_tte now shows two rows with ranges for event and censored times, respectively.
• g_km works with one arm survfit objects.
• In forest plot functions, added formatting to display extreme values to “>999.9”.
• t_summarise_variables uses now n instead of N as a denominator for calculating percentages for factors by default.
• t_rsp now works when all response values are TRUE or FALSE.

Deprecated Functions

• Deprecated t_summarize_variables as t_summary is more powerful.
• Replacing t_summarize_by_visit with t_summary_by will occur in an upcoming release.