Predicting menstrual bleeding patterns with levonorgestrel-releasing intrauterine systems

Figures & data

Figure 1. Bleeding intensity during days 0–360 post-placement of levonorgestrel-releasing intrauterine system, compiled from bleeding-diary entries from all participants using LNG-IUS 12 in two in clinical trials [13,28]. Each row represents data from one individual participant.

Figure 2. Bleeding intensity from daily bleeding diaries during days 0–270 post-placement of LNG-IUS 12 in clinical trials [13,28], clustered by bleeding pattern for the fixed-cluster model. Each row represents data from one individual participant.

Figure 3. Bleeding patterns for participants using LNG-IUS 12 who were assigned to the ‘predominantly bleeding’ cluster, normalised with respect to cycle length and start of first bleeding/spotting episode within days 91–270. Each row represents data from one individual participant. LNG-IUS, levonorgestrel-releasing intrauterine system.

Table 1. Fixed-cluster model performance analysis with LNG-IUS data showing proportion of women allocated to each cluster and correct classification rates.

		Predicted cluster
		Predominantly amenorrhoea	Predominantly spotting	Predominantly bleeding	Total
True cluster	LNG-IUS 12
	Predominantly amenorrhoea, n	55	71	13	139
	Predominantly spotting, n	22	324	132	478
	Predominantly bleeding, n	5	157	572	734
	Total, n	82	552	717	1351
	% correct prediction	67.1	58.7	79.8	70.4
	LNG-IUS 20
	Predominantly amenorrhoea, n	16	12	0	28
	Predominantly spotting, n	6	76	11	93
	Predominantly bleeding, n	0	31	64	95
	Total, n	22	119	75	216
% correct prediction		72.7	63.9	85.3	72.2
	LNG-IUS 8
	Predominantly amenorrhoea, n	41	45	4	90
	Predominantly spotting, n	18	252	91	361
	Predominantly bleeding, n	12	233	604	849
	Total, n	71	530	699	1300
% correct prediction		57.7	47.5	86.4	69.0

LNG-IUS: levonorgestrel-releasing intrauterine system.

Figure 4. Sizes of the true and predicted clusters by device and cluster type. LNG-IUS, levonorgestrel-releasing intrauterine system.

Figure 5. Proportion of women with regular and irregular cycles by device in the regularity assessment. LNG-IUS, levonorgestrel-releasing intrauterine system.

Figure 6. Logistic regression of regularity by device in the fixed-cluster model. The model’s ability to predict regularity is indicated by the amount of overlap between the coloured bars, with greater separation between the coloured bars indicating better prediction. LNG-IUS, levonorgestrel-releasing intrauterine system.

Figure 7. Receiver operating characteristics plot for LASSO regularity regression model in the fixed-cluster model demonstrating highest sensitivity and specificity is achieved for LNG-IUS 12. LASSO, least absolute shrinkage and selection operator; LNG-IUS, levonorgestrel-releasing intrauterine system.

Figure 8. Histogram and median values of autocorrelation values by group during days 91–270 in the regression-tree model. The model’s ability to predict regularity is indicated by the amount of overlap between the coloured bars, with greater separation between the coloured bars indicating better prediction. Autocorr., autocorrection.

Figure 9. Distribution and median values of the number of bleeding days by subgroup during days 91–270 in the regression-tree model.

Figure 10. Distribution and median values of number of spotting days by subgroup during days 91–270 in the regression-tree model.

Gemzell-Danielsson K, Schellschmidt I, Apter D. A randomized, phase II study describing the efficacy, bleeding profile, and safety of two low-dose levonorgestrel-releasing intrauterine contraceptive systems and Mirena. Fertil Steril. 2012;97(3):616.e3–622.e3.

 Web of Science ®Google Scholar

Li S, Lee H. Learning daily activity patterns with probabilistic grammars. Transportation. 2017;44(1):49–68.

 Web of Science ®Google Scholar

Data availability

The datasets generated and/or analysed during the current study are available in the Clinical Study Data Request (CSDR) repository, https://clinicalstudydatarequest.com