NEWS

---

cgmguru 1.3.0 (2026-08-21)

• Added interval_down() for downsampling Dexcom-style 5-minute CGM data into clock-aligned 15-minute intervals. It averages available readings in each interval, labels results at the interval end, and can optionally return the number of contributing readings with n_observed = TRUE.

cgmguru 1.2.1

• Corrected standard Level 1/2 hypo- and hyperglycemia episode labelling so short, unsuccessful recovery runs remain within an episode until sustained recovery occurs. Level 1-exclusive events now use the full labelled Level 1/2 intervals for overlap assessment, aligning episode counts and reported boundaries with iglu::episode_calculation().
• Aligned episode-rate rounding with R's rounding behaviour.
• Relicensed cgmguru from MIT to GPL-2 to reflect that parts of the Rcpp implementation port, translate, or structurally adapt code from the GPL-2 licensed iglu package.
• Updated package metadata, license files, README, and package documentation to describe iglu as an upstream GPL-2 implementation source and cgmguru as a GPL-2 licensed package.
• Added LICENSE.note and source-file notices for cgmguru implementation components derived from, ported from, or structurally adapted from iglu, including preserved upstream copyright notices.
• Corrected the documentation for uGMI to identify it as the Updated Glucose Management Indicator. The existing uGMI calculation is unchanged.
• Added the Bergenstal et al. (2026) reference to the detect_all_events() documentation and package publication list.

cgmguru 1.2.0 (2026-07-08)

• Added rebound_events() to detect rebound hypoglycemia and rebound hyperglycemia using cgmguru Level 1 initial events followed by an opposite threshold crossing within 120 minutes. The function returns the preprocessed event grid by default. detect_all_events() includes rebound rows and wide summary columns.
• Added summary_digits to detect_all_events() to control rounding for numeric summary outputs. The default remains 2 decimal places; use NULL or "none" to return unrounded values.
• Fixed detect_all_events() documentation so formulas involving mean_glucose render the column name literally instead of treating the underscore as a subscript in Rd output.
• Added Rcpp-backed conga_rcpp() for iglu-compatible Continuous Overall Net Glycemic Action (CONGA) calculation. The function interpolates CGM data to a regular day-aligned grid and returns the standard deviation of glucose differences separated by n hours.
• Added Rcpp-backed mage_rcpp() for iglu-compatible Mean Amplitude of Glycemic Excursions (MAGE) calculation. The function supports the moving-average (version = "ma") and legacy standard-deviation (version = "naive") algorithms, segment-level output with return_type = "df", and avg/service/max/plus/minus direction summaries. Plotting options from iglu::mage() are intentionally not implemented.
• Added Rcpp-backed modd_rcpp() for iglu-compatible Mean of Daily Differences (MODD) calculation. The function uses cgmguru's iglu-compatible day-grid preprocessing and returns same-time-of-day absolute glucose differences averaged across the requested day lag.
• Updated excursion() episode-start output to include the peak glucose value within 2 hours of the excursion start (maxima_glucose), its time (maxima_time), time to peak in minutes (time_to_peak_min), and maxima_index. Excursion starts now require both the previous glucose value and the starting glucose value to be at least 70 mg/dL before checking for a rise greater than 70 mg/dL within 2 hours.
• Added iglu parity tests for conga_rcpp(), mage_rcpp(), and modd_rcpp() using the iglu example CGM datasets.

cgmguru 1.1.1

• Updated sensor_wear() tests to avoid timezone-dependent one-to-one start_date comparisons against iglu::active_percent() manual windows. Fixed-window sensor wear tests now compare the calculated observed/expected reading counts directly, making the checks stable across DST-sensitive timezones.
• Expanded the package-level cgmguru vignette into a practical CGM analysis guide covering data requirements, sensor wear, event summaries, event-grid inspection, GRID analysis, postprandial maxima workflows, excursions, visualization, and scaling to larger datasets.

cgmguru 1.1.0 (2026-06-09)

• Updated maxima_grid() and detect_between_maxima() to include all subject IDs in episode_counts, returning 0 for subjects with no detected episodes or between-maxima results.
• Fixed detect_all_events() to summarize event counts segment-by-segment after interpolation gaps, preventing events that end at a gap boundary from being merged into the next segment.
• Updated extended hypoglycemia event detection to match iglu by requiring duration strictly greater than 120 minutes below 70 mg/dL, rather than greater than or equal to 120 minutes.
• Avoided materializing the standalone hypo-/hyperglycemic event grid when return_interpolated = FALSE, improving speed and memory use for calls that do not request the interpolated data.
• Optimized returned interpolated event grids by preallocating storage, avoiding repeated ID strings in C++ storage, and skipping unused grid metadata.
• Changed detect_all_events() summary glucose metrics to use original raw CGM values by default, with summary_metrics_source = "preprocessed" for the previous internal event-grid behavior.
• Rounded detect_all_events() CGM summary metrics and sensor wear outputs to two decimal places.
• Added sensor_wear_ndays to detect_all_events() to calculate sensor_wear_percent over a fixed retrospective window, such as the last 90 days; when omitted, sensor_wear_percent continues to use the original timestamp span.
• Updated sensor_wear() so the default calculation uses each subject's original timestamp span. Supplying ndays now switches to the fixed-window calculation.
• Renamed detect_all_events() return tables to subject_summary and glycemic_event_summary.
• Renamed detect_all_events() summary columns for clarity: sensor_wear_percent, *_total_episodes, and avg_minutes_below_54_per_episode; CV is now reported as a percent.

cgmguru 1.0.1 (2026-05-14)

• Renamed event count output columns to total_episodes for standalone hypo-/hyperglycemic event summaries and detect_all_events() long-format event output.
• Updated documentation, examples, vignettes, and tests to use total_episodes consistently.

cgmguru 1.0.0 (2026-05-14)

• Added iglu-compatible event-grid interpolation to the event detection pipeline, including automatic reading interval detection, linear interpolation up to inter_gap, gap masking, and segment-wise event classification.
• Added interpolate_cgm() as a standalone helper for inspecting the interpolated event grid used by glycemic event functions.
• Added sensor_wear() and included observed-data sensor wear in detect_all_events() summary output.
• Updated detect_all_events() to calculate CGM summary metrics on the interpolated event grid while returning event and summary tables only.
• Added iglu parity and interpolation-focused tests for glycemic event detection.

cgmguru 0.2.0 (2026-05-07)

• Added preset event definitions with type = "lv1", "lv2", and "extended" to detect_hyperglycemic_events() and detect_hypoglycemic_events().
• Updated event boundary reporting so end_glucose and end_index identify the final dysglycemic reading immediately before the confirmed recovery period begins.
• Renamed public output index columns to singular forms: indices to index, start_indices to start_index, end_indices to end_index, max_indices to max_index, and min_indices to min_index.
• Added tests for pre-recovery event boundaries in extended hyperglycemic events with 5-minute and 15-minute sampling intervals.

cgmguru 0.1.0 (2025-11-05)

• Initial CRAN submission.