A handheld milk conductivity sensing device (Mylee) for measuring secretory activation progress in lactating women: a device validation study

Abstract

Background

Human milk electrolytes are known biomarkers of stages of lactation in the first weeks after birth. However, methods for measuring milk electrolytes are available only in laboratory or expert settings. A small handheld milk sensing device (Mylee) capable of determining on-site individual secretory activation progress from sensing the conductivity of a tiny milk specimen was developed. Here we evaluate the validity of a novel milk-sensing device (Mylee) for measuring the progress of milk maturation and secretory activation status.

Methods

Retrospective data analysis of laboratory records generated using the Mylee device. Device conductivity measurements were assessed for accuracy, reliability and stability in rigorous laboratory tests with standard materials. A set of human milk specimens (n = 167) was used to analyze the agreement between the milk maturation score and laboratory measurements of the secretory activation biomarker milk sodium [Na+].

Results

The Mylee device was demonstrated to have excellent reproducibility (CV_95%<5%) and accuracy (error < 5%) for conductivity measurements of a small specimen (350 µl), with good device stability and almost perfect inter-device unit reliability (ICC > 0.90). With regression analysis, we revealed excellent agreement between Mylee milk maturation (MM%) output or its raw conductivity signal and laboratory measurements of conductivity and sodium [Na+] in a dataset of milk specimens (n = 167; R² > 0.9). The Mylee MM% score showed good predictive ability for secretary activation status, as determined by sodium threshold (18 mmol/L) in human milk specimens.

Conclusions

In this study, we demonstrated the reliability and validity of the Mylee device and its ability to detect on-site milk secretory activation in a manner comparable to that of electrolyte-based methods. The novel MyLee device offers the potential to generate real-time information about the lactation stage, measured by mothers at the commodity of their home.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12884-025-07141-x.

Background

Breastfeeding and human milk are the gold standards of infant care [1, 2]; however, breastfeeding rates are low [3], and insufficient milk supply and consequences thereof are the main reasons for early discontinuation of breastfeeding [4, 5], with a huge drop in rates evident in the first month after birth [6, 7]. Post glandular (or secretory) lactation insufficiency, caused by ineffective milk removal in the first period after birth, is a main contributor [8–10]. Breastfeeding is managed by the mother at her home, with periodical oversight by a care provider [11]. Currently, there is no available tool deployed for monitoring individual secretory activation progress for driving personalized lactation care.

Secretory activation is a complex biological process in which mammalian epithelial cells gradually differentiate toward the production of copious amounts of mature milk during the first days to weeks after birth [12]. Various studies have used human milk electrolytes, mainly sodium ions (Na+), and electrolyte balance as biochemical markers for successful or failed secretory activation [13–17], predicting prolonged breastfeeding success [17–22]. To date, various testing methods have been implemented in research, many of which require a laboratory setting, and others designated for field use [23]. These require technical expertise for handling, precluding them from practical non-expert use.

MyMilk laboratories Ltd. developed Mylee, a novel smartphone-operated, handheld, milk conductivity sensing device that measures the individual secretory activation score of lactating women in real time (Fig. 1). The device defines the milk maturation progress for each breast side separately by measuring the milk conductivity within seconds of a tiny (approximately 6 drops, 350 µl) human milk specimen captured in the designated milk specimen holding chamber (milk well). The raw milk conductivity data are transmitted via Bluetooth to a smartphone application designed for instant computation of the milk maturation score (MM%), reflecting the specimen maturation status as a percentage (%) within the full dynamic range from initial colostrum to fully mature milk, as was previously described [24]; the lower the MM% score is, the less advanced the mother is in her lactogenesis (secretory activation) process and supply progress. A previous study described the development of the algorithm and the core conductivity technology integrated in the MyLee and presented the feasibility of the system for identifying milk secretory activation delay and breastfeeding problem status in real-world operation [24].

Fig. 1.

Schematic representation of the Mylee device and app. A milk well is designed to hold a predefined small milk specimen volume when the Mylee device is placed on top. The device is connected via Bluetooth to a smartphone and operated via the Mylee app. The milk data are transferred from the device to the app, and the Milk maturation percentage is computed based on server-based algorithm logic

The MyLee system was designed for simple operation and is associated with a user-friendly app interface. It was designed not to require ongoing calibration and intended for home use and is designed to generate technical alerts at misuse, with the aim of creating a reliable system to be operated directly by untrained users in uncontrolled environments.

The aim of this study was to evaluate the reliability, accuracy, stability and validity of the MyLee system using a retrospective data analysis of laboratory accumulated data.

Methods

Study design

This retrospective study was based on a dataset extracted from MyMilk’s laboratory database to assess the performance of the Mylee novel system and between methods agreement using various results from routine laboratory analysis. Records were generated empirically by in-house human milk testing procedures and QA tests performed at MyMilk Laboratories between February 2022 and October 2023 using the Mylee Milk Sensing Device (Product ID: MLK-01-00). An additional dataset was used for extracting previous records of laboratory grade conductivity and sodium measurements performed for different aliquots of the same milk specimens in MyMilk Laboratory between 2015 and 2023. When available, baby age, reflecting specimen collection time after birth in days, was included in the records.

Setting and context

This retrospective analysis included data records from milk specimens obtained from breastfeeding mothers in the state of Israel. According to formal figures published by the Israeli Ministry of Health, the breastfeeding initiation rate in Israel is 93%, and exclusive breastfeeding rates drop to 55% by 1 month and to 20% at 6 months (data collected by the software “Healthy thinking” within “Tipat Halav” National Well Baby Visit Centers [25]).

Sample

For the retrospective analysis, a dataset of 227 milk specimens with MyLee measurement records was extracted. A total of 118 of the records had day after birth (age) records (average age 50 days; full range 3-300 days; 78% within 2 months of age). For a comprehensive analysis, all measurements were treated as independent, and multiple measurements from the same mother were not excluded, introducing a possible bias. As certain additional conditions, such as mastitis, breast inflammation and milk stasis, may also be reflected in abnormal levels of milk electrolytes [26, 27], for comprehensive analysis covering the extended physiological range, a separate subset of records associated with breast pain was included in the dataset (but conditions were not independently assessed).

Unless otherwise specifically noted, data was extracted for Milk specimens that were either aliquots from full breast expression, or milk sample expressed before a feed or at least 1 h after the last feed. This is done to avoid possible variability stemming from possible extreme fat content towards the end of milk expression that may have significant effect on result accuracy.

For device performance retrospective analysis, an additional dataset of standard material laboratory records was included.

Sample sizes were determined to comply with commonly described sample sizes required for in vitro diagnostic performance verification.

Measurement

The variables assessed in the current study focused on a previously described metric, the milk maturation percentage (MM%) score and raw conductivity, as measured by the MyLee system.

The Na + concentration and conductivity were measured via laboratory-grade instruments and the performance of the MyLee device was assessed within-device and between methods, as detailed herein.

Milk maturation percent (MM%)

MM% is a novel index calculated from device-measured conductivity using a proprietary software embedded equation that we previously defined [24] based on a unique empirical dataset of hundreds of mother’s milk specimens along the first period after birth. MM% was designed to intuitively follow the directional progression of mothers’ milk on a continuous 0–100% scale (from early colostrum to what we call fully mature milk) and is inversely related to conductivity; The higher the conductivity level is, the lower the MM%. Data previously gathered with the system revealed a typical time-dependent increase in the milk maturation parameter (MM%), characterized by an initial steep increase, followed by a moderate increase, and reaching a plateau during the first weeks postpartum [24].

MyLee device and software

The MyLee device was designed to be a robust calibration-free conductivity sensor for up to hundreds of tests over months; this device was intended for use with a small human milk specimen for on-site evaluation of the milk maturation % score, as previously described [24]. The handheld device includes 2 fixed electrodes and a temperature probe. When the device is placed on top of the milk well, it continuously records the conductance of a small prefixed volume of specimen encapsulated by it. The Mylee Kit includes two identical separate milk wells, enabling sampling from each breast separately.

The device is intended for use in an uncontrolled environment directly by untrained users. The system was designed to be simple to operate, with a single ON/OFF button, a Bluetooth connection, and seamless measurement that is operated by a smartphone application. The Mylee device internally evaluates the validity of a measurement; the milk well is designed to hold a prefixed liquid volume at the time of sensing, and discard the specimen remains. The device is engineered to monitor sufficient specimen volume for reliable measurement with an inability to perform a valid measurement when using insufficient specimen volume (< 250 µl) and showing an alert to the user. As fluid conductance is influenced by fluid temperature, the device includes a sensitive temperature probe and corrects the conductivity to 25c. In addition, the system alerts and does not allow for measurement in temperatures below 10c or above 40c. The system also shows an alert at extreme measurements that may stem from external contamination of the sample, and advise the user to use a new sample and perform a new measurement.

The device is operated through the MyLee App (Mothers App: MyLee, iOS, ID1533231342, MyMilk laboratories Ltd.) via Bluetooth. The Mylee app guides the user through the measurement session, and the specimen’s computed MM% result is recorded on the app within seconds. MM% was developed as an intuitive marker of how effective a mother is in her efforts to establish her milk supply during breastfeeding establishment and is presented on a continuous 0–100% scale (inversely related to conductivity). The MM% is presented in the App separately for each breast, and progress is tracked in a chart. The software assigns a color indication of specimen status relative to day-match reference benchmarks [24]. The data are synced and stored in a secure cloud-based server (AWS). The system is not intended to diagnose or treat any medical condition or to serve as a substitute for medical advice or to guide medical care.

The device was used according to the provided user guide. The specimen size was 350 µl in all laboratory tests.

Mother’s milk specimens were obtained from the MyMilk stored specimen set for laboratory testing. Milk specimens were frozen (-20c) in 1 ml aliquots and were brought to stable room temperature before performing measurement. Measurements were performed on whole-milk specimens.

Milk wells were washed and dried between each specimen, and the device sensors were cleaned by the cleaning accessory with 3–4 ml of tap water between specimens and at the end of use, as detailed in the user guide instructions.

Device performance analysis

The device conductivity reliability and accuracy were tested with KCl solutions and conductivity reference standards. Linearity was defined by a series of 7 concentrations of KCl in the measuring range of 10–100 mM (1400–12880 µS/cm) obtained by serial dilutions of KCl solution. The conductivity standard materials used for routine QA and device testing were N.I.S.T. Certified (Level 1 XS Basic EC 1413 µS/cm (25 °C), GiorgioBormac; Level 2 5.00 mS/cm (25 °C)).

For accuracy and precision, data from 9 devices tested over 5 separate dates were included. For stability analysis, records from 5 devices with two levels of standard solutions were evaluated over 2 months (32 independent dates, with an average of N = 120 measurements per device).

Device performance was further evaluated with records of mother’s milk specimens. The inter-day analysis included 5 milk specimens and 3 standard materials over 10 independent days. Test-retest analysis was performed for 4 devices with a set of 28 milk specimens, and the inter-device correlation was tested.

Laboratory grade na + and conductivity measurements

The performance of the MyLee system was compared to that of laboratory grade conductivity analyzer (LAQUA Twin conductivity meter EC-33, HORIBA) and sodium analyzer (Na + ion selective electrode analyzer, LAQUA Twin, NA-11 HORIBA) apparatuses, which previous reports for accurate analysis of human milk sodium levels have validated [23]. Instruments were calibrated and used according to the manufacturer’s instructions, with standard materials provided by the manufacturer.

Data collection

This study utilized datasets collected with a Mylee device (MLK-01-00) in the laboratory between Feb 2022 and Oct 2023. This dataset included records of milk specimens measured with the Mylee device (n = 227) and records of standard material measurements over a period of time and from various protocols designed to evaluate device setting and performance, reliability, accuracy, inter-device variability, and prolonged device stability.

An additional dataset, including laboratory records from between 2015 and 2023, was used for extracting records of laboratory-grade conductivity and sodium (Na+) measurements performed independently on aliquots of the same milk specimens. When available, records of day-after-birth of milk expression were obtained.

The MyMilk Laboratory stored milk specimen set was derived from specimens voluntarily sent by mothers for various informational laboratory tests. All specimens in the dataset were anonymized and identified by specimen identification numbers.

The data were anonymized and manually organized for analysis based on the available records, specific protocol, and analysis of interest.

Data analysis

For the MyLee system, accuracy and precision analysis data from 9 devices tested over 5 separate days with standard materials were included in the study. The coefficient of variation (%CV_95%) was evaluated for imprecision. For accuracy, the error was determined. The acceptance criterion was set at < 5%. For stability analysis, records from 5 devices with two levels of standard materials over 2 months (32 independent dates, with an average of N = 120 measurements per device) were evaluated for repeatability; imprecision, measured as %CV_95%, was evaluated, and the Levey–Jennings chart was constructed.

Device performance was further evaluated with records of mother’s milk specimens. Inter-day analysis included 5 milk specimens and 3 standard materials over 10 independent dates, and repeatability was quantified by using the intraclass correlation coefficient (ICC), a widely used test-retest indicator used for measuring the reproducibility of measurements [28]. Test-retest analysis was performed for 4 devices with a set of 28 milk specimens, and the inter-device correlation was tested.

For between-method agreement analysis and system validation analysis, the dataset of the first milk specimen records was manually matched to corresponding records from the second datasets. Due to the unstructured dataset and incomplete records, some uncontrolled variability and bias may have been introduced in an uncontrolled manner, including the following shortcomings: the analysis date and number of specimens per analysis were not uniform between methods; specimens were not controlled for demographic information; and, to offer comprehensive analysis, all measurements were treated as valid independent measurements, including multiple measurements from the same mother, different breast sides, and different conditions (such as pain), presenting a possible bias toward mothers who sampled more frequently or to certain conditions not differentially regarded in the analysis.

For between-method agreement analysis, the results obtained by the evaluated MyLee systems (MyLee) of a large number of mother’s milk specimens were compared to the records of laboratory grade measurements (LAQUA Twin conductivity meter, LAQUA Twin Na + ion selective electrode apparatus previously validated for human milk sodium analysis [23]). Regression and Bland‒Altman plot analyses were performed to determine the agreement and differences between the methods. For Bland–Altman analysis, the Na equivalent (Na + Eqv) was calculated by an equation derived from plotting conductivity and Na + records measured in a separate set of breastmilk specimens (n = 72 specimens, [Na+] = 5–50 mmol/L, dataset previously described in [24]. The Na + eqv was calculated for a subset of the MyLee measured dataset according to the equation limits (5–50 mmol/L, N = 132). Differences between the Na eqv calculated from conductivity and directly measured by [Na+] ion-selective electrode records were plotted against their averages.

For system validation analysis, a subset of this dataset was retrospectively categorized into two groups according to baby age records: days 5–10 and days 11–60 after birth. The dataset was further retrospectively categorized into two subgroups based on milk Na + levels, a previously suggested objective biomarker to evaluate the possibility of delayed lactation and a measurement of milk production in the early postpartum period [13, 17, 21]. A concentration of 18 mmol/L was set as the threshold above which previous reports categorized delayed lactogenesis according to milk Na levels beyond day 5 [14, 15, 17, 19, 21, 22, 29, 30]. Accordingly, in the present analysis, [Na+] > 18 mmol/L was set as positive for the condition (‘Yes’), and [Na+] < 18 mmol/L was set as no condition (‘No’). Group classifications do not imply a diagnosis or a confirmed clinical condition, and no direct measurements of milk volume or per-feed milk transfer were performed. The ability of the MM% parameter, as measured by the device, to distinguish between positive and negative conditions according to Na was assessed, and the accuracy of different cutoff values was evaluated through the calculation of sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for each period after birth.

See supplementary figure S4 for a detailed visualization of sample types, data records, and performance analysis described in the manuscript.

Statistical analysis

Statistical analysis was conducted using JASP Graphical Statistical Software Version 0.17.1.0 (JASP Team, University of Amsterdam, Amsterdam, the Netherlands). Descriptive statistics included the mean, standard deviation, and coefficient of variation. Precision was evaluated by calculating a strict coefficient of variation covering 95% of the results (± 1.96 SD) (%CV_95%), with the criterion of %CV_95% < 5%. The uncertainty was 1.96 SD, and the confidence limits (95%, ± 1.96 SD), LUL (lower uncertainty limit) and UUL (upper uncertainty limit) were reported. %Error was used for accuracy assessment (with a pass criterion of %Error < 5%). The intraclass correlation coefficient (ICC) and test-retest analysis were used to measure the reproducibility of the measurements. Correlation and regression analyses were used for between-measurement agreement analysis. ANCOVA was used to estimate the mean change in MM% based on group class to the condition (‘Yes’/‘No) and days after birth class. The specificity, sensitivity, and positive predictive value (PPV) and negative predictive value (NPV) for assessing the MM% cutoff were reported.

Results

The MyLee system

The Mylee device (Fig. 1) was designed as a single-button device entirely operated by a designated mobile application intended for use directly by untrained users, with no need for ongoing calibration, for real-time milk maturation score computation based on reliable milk conductivity measurements. As colostrum (first milk) and early milk are produced in small volumes during the first days after birth, the Mylee device was engineered to sense conductivity in several drops of human milk reliably. Since deviations in specimen volume strongly influence liquid conductivity, the milk well was designed to hold a specific liquid volume, set at 350 µl when encapsulated by the Mylee device and to alert when the specimen is not sufficient for reliable measurement (< ∼250 µl).

To demonstrate measurement reliability in a range of milk volumes tolerated by the milk wells (Fig. 2A), conductivity was assessed for a range of specimen volumes and found consistent between 260 µl and 450 µl in two levels of reference material within the measurement range. Between 260 and 450 µl: Average %Error ± SD; 0.7%± 0.8 and 1.6%± 0.8, 1413uS/cm and 5.00mS/cm, respectively; a 10.3%± 3.8 Error introduced with Specimen volume 230–250 µl; under 230 µl system generated insufficient specimen alert in line with system design (Fig. 2B). A specimen size of 300–400 µl (6–8 drops) was chosen to be included in the instructions of use, and a 350 µl specimen size was used for consistent laboratory testing.

Fig. 2.

Device measurements are reliable for a range of specimen temperature and volume. (A) Milk well 3D illustration and middle slice cross view illustrating the central cavity volume (300–350 µl). (B) The conductivity of specimens gradually elevated from 200 µl to 450 µl as recorded by the Mylee device in two levels of standard conductivity solution. red X - The system alerted of insufficient specimen volume for measuring. (Ci) Series of conductivity measurements by device vs. specimen temperature in a bath-warmed standard sample left to cool to ambient temperature (Starting Temp 39 °C, End Temp 30 °C; yellow) and a Fridge-cooled standard sample left to warm to ambient temperature (Starting Temp 17 °C, End Temp 24.7 °C; blue). Two Room Temp standard samples were included for comparison (26.4 °C; 27.6 °C, black). (Cii) Conductivity measurements of two levels of standard materials from 10 different measurement dates performed at various ambient temperatures. The X-axis represents the sample temperature as measured by the device

The device was designed for indoor use at a range of ambient temperatures, and instruction for use in the system user guide detail intended use in room temperature. The system was engineered to measure local temperature and to correct the measured conductivity to 25 °C automatically. To test the reliability of the results at a range of temperatures, the autocorrected conductivity measurements were evaluated either by monitoring the changes in specimen temperature after warming or cooling or by comparing various samples at a range of ambient temperatures; consistent results were found between 17 °C and 39 °C (Fig. 2C-D); for details, see the supplementary material.

System performance testing

The system was engineered to cover the full range of breastmilk conductivity measurements (set based on an in-house empirical dataset) [24]. Conductivity linearity was confirmed in the measuring range of 10–100 mM KCl corresponding to the physiological range expected in breastmilk throughout the breastfeeding stages (Fig. 3A; R² = 0.999). The system is designed to generate an alert when the specimen’s conductivity is out of the full expected range, as this may imply specimen dilution or external contamination.

System performance was evaluated for precision and accuracy of conductivity measurements using two levels of conductivity standard solutions (Level 1 standard material (1413 µS/cm) and Level 2 standard material (5.00 mS/cm)).

The variability and stability of the conductivity results were evaluated throughout intensive use. The evaluated dataset included 599 measurements generated by routine quality assurance tests of 5 different device units, assessed over 2 months, with 32 independent recording dates. A representative Levey–Jennings chart of the quality control data logged by one of the device units demonstrates stable reads over time (Fig. 3B).

Fig. 3.

Device measurements are stable and reliable over time. (A) Linear regression of conductivity to KCl molarity in the measuring range of 10–100 mM. (B) Levey–Jennings chart of the quality control data logged by one of the devices tested for stability, with two levels of conductivity standard solution (Level 1 (1413 µS/cm) and Level 2 (5.00 mS/cm)) collected over a 2-month period. The middle line is the mean, and the dashed line measures the distance from the mean to the standard deviation. (C). Test-retest analysis to assess the consistency of the measurement set-over devices by comparing conductivity records in a set of 28 mothers’ milk specimens

System reproducibility was evaluated with data from mother’s milk measurements, assessing agreement between different device units in a dataset of 112 mother’s milk measurements in an extended physiological range generated by comparing conductivity records of 28 mother’s milk specimens by 4 device units. By test-retest analysis, we demonstrated an almost perfect correlation (R > .99) (Fig. 3C) and excellent agreement (ICC = 0.99) between device units.

By stability testing we evaluated variability over time and demonstrated a small uncertainty and very good precision; Level 1 conductivity standard (1413µS/cm): %CV_95%= ±1.8%, Uncertainty(2SD) = 25µS/cm; Level 2 conductivity standard (5000µS/cm): %CV_95%= ±2.2%, Uncertainty(2SD) = 113µS/cm.

Good accuracy was demonstrated. Level 1- %Error = ± 0.9%; Level 2- %Error = ± 3.5%, with a small bias: Level 1–12 ± 9µS/cm, Level 2-174 ± 50 µS/cm. These can be systematically corrected to minimize error (For the full table, see Supplementary Figure S1A). The device battery indication over the testing period decreased from 80–90–10%; no bias could be attributed to the charging state (data not shown). Good inter-device units and inter-day repeatability and accuracy were demonstrated in an additional dataset that included 250 measurements recorded by 9 devices over 5 separate days. %CV_95% = 2.5 − 4.5%, %Error = 1.7 − 2.5% (for full tables, see Supplementary Figure S1B).

System reproducibility was further assessed with mother’s milk specimens by test-retest analysis. Inter-day consistency was assessed in a dataset of 160 measurements, more than 10 dates and 8 specimens (3 standards, 5 milk specimens) in two separate device units. An excellent intraclass correlation was demonstrated between the dates (ICC = 1.0). The conductivity measurements were found to be reproducible with very good inter-day precision (average %CV_95%±SD = 3.13%±1.7) (Supplementary Figure S1C).

For between-method comparison analysis, MyLee system measurements were also compared to laboratory records of the conductivity of a large set of mother’s milk specimens, as measured by an independent conductivity meter (LAQUATwin, EC-33 HORIBA). A good correlation and good linear fit (P < .001, R² = 0.920, N = 177) were observed between the methods (Supplementary Figure S1D).

Sodium is an accepted research biomarker used to evaluate the possibility of delayed lactation and to measure milk production in the early postpartum period [13, 17, 19, 21, 22]. To assess the validity of the MyLee system for the intended use of tracking lactogenesis and lactation adequacy, we tested the agreement between the system measurement results and the records of the milk biomarker sodium in a set of mother’s milk specimens. Milk sodium ion (Na+) data were collected from an independent dataset generated by the Na + ion selective electrode instrument (LAQUATwin, NA-11 HORIBA), a tool that was previously validated for quantifying biomarkers in human milk [23] on aliquots from the same milk specimen origin.

To test for agreement between the MM% score from the device and the milk sodium ion [Na+] measurement, a regression analysis was performed on a dataset of 167 mother’s milk specimens where data were available for both measurements. A strong inverse correlation and a good inverse linear fit were revealed between the milk [Na+] and Mylee MM% results (Fig. 4A; Pearson’s R=-.96, P < .001; R² = 0.922). Similarly, good correlation was observed between the two methods when plotting the device raw conductivity signal against the specimen sodium ion [Na+] concentration records (Fig. 4B; Pearson’s R = .96, P < .001).

Fig. 4.

The measurements from the Mylee device were comparable to milk sodium measurements in human milk (n = 167). Correlation plots of the device (A) milk maturation (MM%) and (B) raw conductivity against the specimen Na + concentration recorded by the ISE (LAQUATwin, HORIBA)

To assess the validity of the device for evaluating delayed lactation and milk production in the early postpartum period, a subset of this dataset was analyzed according to baby age records (two classes: days 5–10, days 11–60 after birth) and categorized into two subgroups based on milk [Na+] levels (ion selective analyzer; LAQUA Twin, NA-11 HORIBA), a previously suggested objective biomarker for evaluating the possibility of delayed lactogenesis in the early postpartum period [13, 17, 21]. The threshold of 18 mmol/L was set as the normal limit, as previously suggested [14, 15, 17, 19, 21, 22, 29, 30]. A total of 62 milk specimens were classified, and a positive condition was defined as a milk [Na+] > 18 mmol/L (days 5–10: condition negative (n = 25), condition positive (yes, N = 9); days 11–60: condition negative (n = 21), and condition positive (n = 7)). The mean MM% was significantly lower in the positive-classified records (Condition Positive) than in the normal-classified records (condition Negative) in the two periods analyzed (Fig. 5A-B); MM% Mean ± SE; Day5-10: Condition negative (No) 94%±1.1; Condition positive (Yes) 69%±4; Day11-60: Condition negative (No) 102%±1.3 Condition positive (Yes) 84%±1.9 (two-factor ANCOVA, Turkey test for the condition P < .001).

Fig. 5.

MM% reflects classification based on sodium limits in milk specimens. Records of mother’s milk specimens divided into two classes according to baby age records (A) 5–10 days, (B) 11–60 days after birth) and categorized into two subgroups based on milk Na + levels. Group differences were tested by two factor ANCOVA, followed by the Tukey test (*** P < .001)

As the MM% gradually changes in the first days after birth [24], the ability of the MM% parameter, as measured by the device, to distinguish between positive and negative results was further assessed by calculating the sensitivity, specificity, precision and accuracy of two separate cutoff values set for each day range postpartum. Based on the previously suggested threshold (set empirically as the 15th percentile of the normal predominant exclusive breastfeeding class [24]), the MM% cutoff value for days 5–10 was set as 80% or greater, and a positive condition was predicted (as assessed by a milk sodium ion [Na+] > 18 mmol/L) with 89% sensitivity, 100% specificity, 100% PPV, 96% NPV and 97% accuracy. Similarly, the cutoff value for 11–60 days was set as 90% or greater for MM and could predict a positive condition with 100% sensitivity, 95% specificity, 87% PPV, 100% NPV and 96% accuracy. Although promising, this analysis is limited by sample size and the lack of objective, clinically accepted measures of lactation conditions.

As measuring of Na + is being increasingly employed in research assessing lactation success, the dataset was further evaluated for the degree of agreement between methods using the Bland–Altman plot, a useful display of the relationship between two paired variables using the same scale. The MyLee device does not claim to estimate or measure Na + directly; however, to enable the analysis, an estimated sodium equivalent (‘Na eqv’) parameter was calculated for a subset of specimens (n = 132) according to equations derived from conductivity and Na + records measured in a separate set of breastmilk specimens from MyMilk Laboratories (n = 72 [24]). Differences are within the theoretical limits of agreement; the points are centered around a mean difference of -3 mmol/L and provide a reasonable confidence interval (± 1.96 SD; -9.9, + 3.7 mmol/L) (Supplementary Figure S2A, B).

As the system was designed for use outside the laboratory context, directly by a mother at her home, we included an example of real-world user-generated data. An expecting mother at her 3rd trimester was equipped with a Mylee Kit (Supplementary Figure S3), which was composed of a Mylee milk sensing device, two milk cells, a cleaning accessory, a charging cable, a quick starter guide and an online full user guide (link under Supplementary Figure S3 legend). The mother downloaded the Mylee iOS app (ID1533231342), registered and agreed to terms of use of the app. The baby’s birth date and time were recorded, followed by the use of app tutorial screens and the voluntary recording of user information (current breastfeeding exclusivity status, gestational week at birth, and birth weight). Device activation was performed by registering the device identification barcode in the app on the first use. The device was turned ON and connected to the app by Bluetooth.

According to the user guide, the mother was guided to use milk samples expressed before breastfeeding and not directly after a feed. Fresh milk specimens (6–7 drops) from each breast were collected and placed in the central circle of the milk well for measurement (with a separate well for each side). The measurements were performed via the app, and the results were recorded within 20 s on the screen and saved. Each measurement was recorded separately for each side. MM% results are assigned a color index where green indicates within or above the day-match reference normal limits, as set by an algorithm based on empirical benchmarks in a good predominant breastfeeding population [24], and orange reflects results below these benchmarks. The use of the device was voluntary, and at the sole discretion of the user and was not controlled. The Mother recorded measurements on 8 independent days, starting from the delivery day in the hospital and continuing at her home and during the first 12 days after birth until stable ≧100% results were reached. Progress was tracked in the chart, where the left and right measurement results are presented relative to population benchmarks (Fig. 6). Good dynamic aligned with good baby weight gain and self-reported breastfeeding confidence. The system also generated personalized insights based on measurement results, between-side comparisons, progress rates, and generated enhanced insights based on additional data logged by the mother to the app (data not shown).

Fig. 6.

Mylee App interface example depicting MyLee device measurement records for each breast at 12 days after delivery. Most recent measurements are assigned a color index according to the App proprietary algorithm relative to population benchmarks of good predominant breastfeeding benchmarks. All measurement records are presented as a chart (blue - left side, Gray - right side), where lines represent progression between days. Population benchmarks are marked as background in green and light green (15th, 50th, and 80th percentile limits)

Discussion

Principle finding

Here, we described the laboratory-tested performance of a newly engineered handheld system (Mylee) designed for the assessment of maternal secretory activation progress by sensing few milk drops. The system provided accurate and reliable results. It was robust to temperature and specimen volume changes, making it appropriate for operation in noncontrolled settings and by untrained users. By performing correlation and regression analyses of a large dataset of mother’s milk specimens between methods, we revealed that the Mylee MM% measurement and its raw conductivity signal were as reliable as quantitative analyses of the previously described [13, 17, 21] milk secretory activation biomarker milk sodium ion (Na+) measured by an ion-selective electrode instrument [23], with excellent fit between methods. When classifying the dataset according to the milk [Na+] threshold, which was previously associated with delayed lactogenesis [13, 17, 19, 21, 22], we revealed significant differences between groups and good predictive values for analyzing the device-generated MM% dataset, supporting the ability of the MM% parameter and the MyLee system to aid in the identification of lactation conditions.

Building on prior work

Lactogenesis, the process by which mammary glands develop and regulate milk secretion, is a complex physiological process [12]. The secretory activation stage (known as lactogenesis II) is tightly related to effective breastfeeding, with frequent and adequate milk removal from the breast [9, 10]. Milk indicators, mainly electrolytes, sodium ions, or their balance, are indicative of mammary gland progress toward the production of copious mature milk, with sharp dynamics occurring over the first days to weeks after delivery [15, 16, 31]. These factors have been demonstrated to serve as valuable biomarkers for successful or failed secretory activation [13, 14, 17] and as potential measures of increased risk for early breastfeeding cessation. We have previously described the development and validation of the computed MM% parameter based on milk conductivity measurements as an intuitive score designed to reflect breastfeeding progress and the gradual transition of milk from colostrum to mature milk on a linear scale from 0 to 100% [24]. As the Mylee device was designed for tracking this milk dynamic from the early phase, it was developed and validated for use with tiny milk volumes (approximately 6–8 drops), making it accessible for use from the first days after birth as well as in cases of low milk supply.

The milk conductivity and derived MM% vary depending on the electrolyte content in the milk specimen. We presented a system enabling reliable detection of small changes in electrolyte balance, as reflected in milk conductivity and MM%, in a small milk specimen, thereby facilitating reliable tracking of individual specimens and potentially following day-to-day progress, inter-breast differences, small deviations under certain conditions, such as breast pain, and population variability.

Although Na + on its own behalf has not been clinically validated as a reference diagnostic method for lactation failure, it has been increasingly implemented in research for lactation classification [14, 15, 17, 19, 21, 22, 29, 30]. By applying correlation and regression analyses, we revealed good agreement between the Mylee MM% and milk sodium ion [Na+) levels. The dataset was also evaluated for sensitivity vs. previously suggested Na + cutoffs and by a Bland–Altman plot with theoretical Na equivalent calculations based on the Mylee device measurements, with comparable results. All in all, these findings, together with our previous report [24] verify that the Mylee device and technology provides comparable feedback to Na + measurements in assessing secretory activation progress and is an effective instrument for monitoring individual dynamics within and between subjects.

Importantly, the MyLee device primarily measures milk conductivity, an indirect marker for electrolyte balance, primarily sodium concentration. While conductivity correlates with milk maturation and secretory activation, it is not meant to comprehensively assess all milk components or serve as a quality measure for breastmilk content.

In contrast to laboratory-based methods, or other technical instruments [23] designated for professional use, which are not intended for and are impractical for real-time day-to-day assessments of secretory activation progress, the Mylee device is specifically designed for the intended use of milk maturation assessment at home; it is compact, seamlessly operated, and robust, and it delivers the results immediately to the user and enables real-time action. While the MyLee system was designed for home use by unprofessional users, the current study did not directly analyze usability in home setting. Naturally, home-use by unprofessional users may introduce uncontrolled variables that were not assessed in the current study. For example, external contaminants such as ointments, creams, perfumes or sweat may in theory affect the result accuracy and was not directly addressed in this manuscript. The system user-guide (Supplementary Fig S3) specifically address non-professional use guidelines including minimizing such contamination (e.g. expressing milk with clean dry hands, and the need to thoroughly remove any cream or ointment from breast or nipple before sampling) but we cannot rule out any and all such effect.

In our earlier feasibility study, we address the use of a version of the technology, assessing data collected through prolong usage in non-controlled home setting [24]. Further studies focusing on broader usability, reliability and applicability measures of the MyLee device in home setting and in diverse population in real world setting are warranted and may surface certain additional capabilities and limitations.

Research, user education, and device refinements could address these potential limitations and enhance the MyLee device effectiveness and reliability in real world setting and drive large scale adoption of the tool. Currently, there is no diagnostic or screening tool for diagnosing insufficient milk supply nor delayed lactogenesis [11, 32, 33]. Conditions are classified by face-to-face evaluation by a professional lactation consultant and baby growth follow-ups, and extreme cases are identified by clinical signs in the baby (e.g., slow weight gain or dehydration). Other indirect measurements of milk transfer or test weight in a single feed are used for supporting evaluation. Although precluded from routine practice, previous studies have identified the potential use of milk biomarkers, mainly electrolytes and their balance, for clinicians’ per-case evaluation, early evaluation of lactation effectiveness, optimal timing for a follow-up visit, evaluation of progress, and cause of inadequate lactation [17, 21, 34], including at-risk mothers of preterm infants [29, 35, 36]. A recent study validated the use of portable instrumentation for sodium measurement directly by clinicians [37]. Our recently published retrospective real-world study demonstrated the use of smart milk conductivity sensing technology by lactation consultants during their home visits and directly by the mother at home for objective measurement of individual breastfeeding efficiency [24]. The system we described herewith offers an easy-to-operate, immediate, robust device for the intended use, designed for facilitating reliable data collection within a home setting by an untrained user. The current system is built as an informational tool, not intended to diagnose or treat a medical condition and secondary to any professional care. Future studies are warranted to validate the clinical usability and usefulness of the tool in home setting and to assess its diagnostic capabilities. Tools directly aiding breastfeeding were introduced in recent years to various programs (including medical insurance and tax benefits), making them more accessible to wide population ranges. The technology in the MyLee product was intentionally built to offer an affordable solution, potentially enabling integrating system use as part of the routine self-and remote care offering for new families. Future adoption of the MyLee by health systems could make such a system accessible for every new mother.

Notably, the current study focused primarily on the use of the system within the first 60 days after birth, and future studies may highlight system capabilities for extended breastfeeding periods, and conditions associated with milk supply, return to work, weaning and more.

Breastfeeding is largely managed directly by the mother at her home; early proactivity and correct breastfeeding practices beginning in the early days are key for breastfeeding success. The Mylee System, a mother-centric digital solution, is designed to drive her proactive breastfeeding care, accompanying the milk maturation results with alerts, tips, and personal-curated content, enhancing the milk measurement with data-tracking tools, personal insights, and educational materials tailored to the baby age and to the milk maturation status. The application also supports sharing results with a caregiver for remote data-powered lactation support. Studies have shown that mobile apps for lactation tracking, education, support [38–40] and tailored texting based on maternal characteristics, including milk electrolyte biomarkers [41], are feasible and may impact breastfeeding success. Future studies tracking maternal engagement and usability of the Mylee system will evaluate the digital solution’s usefulness as a valuable tool and a potential driving force in breastfeeding success.

Study design limitation

This study was a retrospective analysis of laboratory data records. Referring to standard material results, when possible, overcame some of the limitations. As breastmilk specimens were not collected as part of structured research but were based on reservoirs of laboratory-stored specimens, with limited availability of demographic and other metadata, the dataset may include inherent errors or biases and is not intended to represent the general population. Given that the method comparison was based on retrospective records of results obtained with methods performed in separate settings, the dataset may also include additional inherent errors or biases. However, despite these acknowledged constraints, the robustness of the system highlights its strength and reliability.

Conclusions

In summary, the good performance of the Mylee device and the agreement with the results from laboratory methods of human milk biomarkers provide evidence that the compact Mylee device can provide reliable and valid information regarding an individual’s secretory activation progress and status, in-par, or enhanced relative to available research-limited tools. Unlike laboratory tools, Mylee measurements can be performed in a simple manner anywhere and anytime without the need for a specialized laboratory tool or technical staff. A tool designed for use in a home setting that enable reliable tracking of secretory indicators can provide new insights into the breastfeeding progress of an individual to obtain further knowledge and understanding about breastfeeding care.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

SH contributed to conceptualization and methodology, data acquisition, data curation, data analysis, and data visualization and writing. AF contributed to the laboratory investigations, oversight and performance, data acquisition, data analysis and writing. DAN contributed to the data analysis and writing. RS contributed to conceptualization and methodology, data acquisition, data curation, data analysis, and data visualization and writing.

Data availability

The datasets generated during and/or analyzed during this study are not publicly available due to data privacy and intellectual property reasons but are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This retrospective study on MyMilk Laboratory dataset derived from stored milk specimen was approved by the Institutional Ethics Committee of Ariel University on December 20th 2022, renewed December 20th 2023 (approval number AU-HEA-AN-20221220). Application title: Assessment of indicators for breastfeeding establishment tracking – Biomarkers, risk factors and breastfeeding status Retrospective data analysis. The MyMilk Laboratory stored milk specimen set was derived from specimens voluntarily sent by mothers for various informational laboratory tests offered by the company. The company is the owner of the legally registered user database (database #7000655996, Israel Privacy Protection Authority database registry), and all users consented to privacy policy and terms of use and waiver allowing for storing and using the data and any remaining specimen anonymously for R&D purposes by the company.

Competing interests

SH, RS, and AF have financial disclosures related to MyMilk Laboratories Ltd. SH and RS are cofounders of MyMilk Laboratories Ltd. and own stock options in the company. AF is an employee of MyMilk Laboratories Ltd. MyMilk Laboratories, the developer of the described Mylee device, is a commercial company for breastfeeding and human milk analysis and support space. DAN has no conflicts to declare.