Prediction of ovulation can be important for women who are trying to conceive, in order for them to time intercourse to their fertile period to maximise their chances of conception. It can also have clinical importance for appropriate timing of procedures, such as intrauterine insemination, frozen embryo transfer or donor sperm insemination.
Over the years, several methods have been developed to predict ovulation. One of the most common methods used in fertility awareness is the sympto-thermal method (STM) because abundant evidence of its reliability has been presented [22–25]. However, the STM requires training in use, detailed recording of symptoms of self-observation and the application of several calculation rules, all lowering general acceptance rates. With the advancement of technology, other reliable, objective and less user-dependent methods have been introduced, which rely on the analysis of the levels of reproductive hormones measured in urine [26]. The study of hormonal profiles in urine during the menstrual cycle has proved to be very valuable in understanding menstrual cycle dynamics, as well as variability in ovulation [19, 27, 28]. Likewise, urinary hormone testing has become an established application for personal monitoring of fertility awareness [9]. Both temperature and hormonal analysis using repeated measurements in menstrual cycles are currently under intensive research as part of the evolving field of digital signal tracking in gynaecological endocrinology [21].
In premenopausal women, the menstrual cycle has a recognisable general pattern, and an average duration; however, there are important variations in cycle length and dynamics, both between and within women [29–33]. Likewise, there are differences in hormonal patterns, even in ovulatory cycles, which introduce a further source of difficulty into the analysis [34, 35]. This compromises the utility of all fertility awareness methods that solely rely on statistical parameters or calculation rules of the menstrual cycle (i.e. monitoring cycle days), and strengthens the importance of looking at specific events of physiological relevance leading to ovulation, such as the LH surge. It is the LH surge which always precedes ovulation, rather than the LH peak, and which is a signal of impending ovulation and peak fertility. Therefore, it is important to have a reliable algorithm to determine the onset of the LH surge in population studies. A standardised algorithm for determining the timing of the urinary LH surge in studies that have examined the hormone profiles in women is still not available.
This study examined the scientific literature, searching for studies focusing on the detection of the onset of the LH surge during the cycle, in urine samples, and found 12 relevant studies. In these papers, there were three major methodological categories designed and used to determine the onset of the surge. The key difference between these methods was the reference used for estimating the baseline values of urinary LH during the follicular phase. Important differences in the performance of these methods were identified. Methods that used the LH surge as a reference to estimate baseline values of LH clearly had the highest level of reliability in this respect, detecting on average up to 76.1 % of the surges, compared with only 64.8 % when the LH peak was used as a reference and 48.4 % when using fixed days of the cycle. The findings of this study might help to improve the analysis of the human menstrual cycle for monitoring to assist with conception and contraception, as well as for biomedical research. In addition, a common methodology would facilitate comparisons between studies.
Many studies report LH peak [29, 36]; however, as a prospective measure, peak LH has less value as it often occurs post ovulation [37]. In addition, assignment of the day of peak urinary LH is dependent on the assay used [4]. Some urinary LH assays use antibodies that recognise the beta-core fragment of LH (LH-βcf), which is a by-product of LH metabolism, and the predominant molecular form of LH in the urine [38]. Levels of LH-βcf continue to rise after the level of the physiologically relevant intact LH reaches peak concentration, sometimes achieving maximal levels up to 5 days later [4].
The comparatively poor performance of this prospectively applicable approach that uses fixed days during the menstrual cycle to determine baseline levels of LH is somewhat expected, since this approach does not take into consideration important factors, such as the variability in the length of the cycle, which mostly affects the follicular phase and exerts an influence over the hormonal dynamics [30]. Ovulation has been seen as early as day 8 of the menstrual cycle and, as the fixed days method generally includes day 8, it is clear that this method will fail under such circumstances.
The higher reliability of a mid-cycle reference (method #2) to determine baseline LH was anticipated, because data from the complete menstrual cycle have to be considered for its application, and due to its proximity with the fertile window, which is a very stable stage of the cycle, and unlike the early follicular phase, is not affected by age [39, 40]. It is also clear why the approach that uses the estimated surge (method #3) as a reference performs better than the method using the peak (method #2) because in almost 25 % of normal menstrual cycles the peak may occur postovulatory (Roos et al. submitted); differences in the LH dynamics around the time of the surge and peak in LH account for this. Some complex LH peaks have been described when researchers have used an assay that recognises LH-βcf, with a longer duration than normal, sometimes described as ‘biphasic’ or in ‘plateau’ [6, 11]; this often presents a significant delay between the surge and the peak of LH, compared with the ‘spike’ type of peaks. It is possible that when the reference is taken starting 3 or 4 days before the peak, in this type of cycle, levels of LH are already high, decreasing the accuracy of the estimation of the baseline levels. Likewise, when using a large number of days away from the peak (such as 6 or 5), in short cycles or those of normal duration, the levels estimated might be closer to those found in the fixed approach.
For fertility monitors, however, a retrospective approach with analysis of data from a complete menstrual cycle is not applicable. Therefore, learning monitors should consider the location reference points from previous cycles for higher reliability of surge detection. Current research focuses on a formal mathematical approach (signal tracking), finding distinct patterns in hormonal and temperature curves to predict change points and distinguish them from the random ‘noise’ (change-point analysis) [21].
A further way to improve the results includes other hormonal markers in urine, in addition to LH. These are estradiol and its metabolites, since a rise in the levels of estradiol triggers the rise in the levels of LH [38], and peak estrone-3-glucuronide (E3G) levels occur on the day of ovulation, as do those of follicle-stimulating hormone (FSH). However, this level of detail would add considerable complexity and additional cost to the analysis, especially by repeated measurement.
Ultrasound-observed ovulation is the gold-standard method used to determine the timing of ovulation and it would be of interest to observe how different approaches to LH surge definition compare to this reference. However, there is variability between women in the timing of the LH surge in relation to ultrasound-observed ovulation [6], thus further work could consider this relationship.
In conclusion, this study found that there are many methods for calculating the LH surge in urine reported in the literature, and their effectiveness at correctly identifying the surge varies between methods. Some of these methods are in use in current fertility monitors. Care should be taken when comparing between studies reporting urinary LH surge data, as results will be dependent on the method employed for surge identification. This study supports the recommendation for using methods that involve an initial estimation of the LH surge to establish which days should be used to determine baseline.