Optimization of a method for collecting infant and toddler urine for non-target analysis using cotton pads and commercially available disposable diapers

Abstract

Background:

Urine is an abundant and useful medium for measuring biomarkers related to chemical exposures in infants and children. Identification of novel biomarkers is greatly enhanced with non-targeted analysis (NTA), a powerful methodology for broad chemical analysis of environmental and biological specimens. However, collecting urine in non-toilet trained children presents many challenges, and contamination from specimen collection can impact NTA results.

Objectives:

We optimized a caregiver-driven method for collecting urine from infants and children using cotton pads and commercially available disposable diapers for NTA and demonstrate its applicability to various children biomonitoring studies.

Methods:

Experiments were first performed to evaluate the effects of processing method (i.e., centrifuge vs. syringe), storage temperature, and diaper brand on recovery of urine absorbed to cotton pads. Caregivers of 11 children (<2 years) used and retained diapers (with cotton pads) to collect their child’s urine for 24 hours. Specimens were analyzed via a NTA method implementing an exclusion list of ions related to contamination from collection materials.

Results:

Centrifuging cotton pads through a small-pore membrane, compared to a manual syringe method, and storing diapers at 4 °C, compared to room temperature, resulted in larger volumes of recovered sample. This method was successfully implemented to recover urine from cotton pads collected in the field; between 5-9 diapers were collected per child in 24 hours, and the total mean volume of urine recovered was 44.7 (range 26.7-71.1) mL. NTA yielded a list of compounds present in urine and/or stool that may hold promise as biomarkers of chemical exposures from a variety of sources.

Significance:

We describe a simple method that eases the collection and improves the analysis of an abundant biological media (urine) to support the advancement and expansion of biomonitoring and tracer studies to characterize chemical exposures across childhood, a vulnerable and highly exposed lifestage.

INTRODUCTION

Infancy and childhood are lifestages with unique exposure profiles and increased vulnerabilities to environmental chemicals (1). Incidental ingestion of soil and dust is an important but poorly characterized pathway of exposure to chemicals for children (2). Previous tracer studies aiming to estimate rates of soil and dust ingestion have methodological limitations (i.e., mass balance concerns) (3), and these methods rely on elemental tracers that are ubiquitous in the environment, complicating interpretations of exposure sources in these studies (4, 5). Consideration of organic chemicals could revitalize tracer studies as a method for generating reliable estimates of soil and dust ingestion (6).

Urine is an abundant biological medium containing potential biomarkers of chemical exposure for infants and children (hereafter: children); compared to blood and other biological media, its collection is non-invasive and poses minimal risk. Clinicians routinely collect pediatric urine to screen for urinary tract infections and for other diagnostic applications (7). In research settings, exposure scientists and epidemiologists have used children’s urine to characterize exposure to pesticides (8–10), metals (11, 12), and other chemicals (13) . Depending on the chemicals and biomarkers of interest, repeated sampling of urine over time may be needed to fully characterize exposure. Collecting urine from children who are not yet toilet trained in non-clinical settings can be logistically difficult, especially if more than one sample is needed (14).

Several methods for collecting children’s urine exist, including urine bags, clean catch, cotton pads in combination with a disposable diaper, disposable diapers and cotton diapers (14), and each method carries with it logistical (e.g., acceptability and ease of use, length of time to collect a sample, volume of sample collected) and ethical considerations (e.g., potential discomfort experienced by the child). As each collection method has its own advantages and disadvantages, the unique research context dictates the most appropriate collection method.

Non-target analysis is a powerful methodology for the detection and identification of novel chemicals in biological media (15). An important consideration when developing any non-targeted methodology is accounting for background contamination so proper biological results interpretation can be achieved. For example, disposable diapers have previously shown to contribute contaminants to nuclear magnetic resonance (NMR)-based urine metabolomics (16, 17) and extensive data processing was required to account for these contaminant signals. Given the complex and often proprietary chemical composition of the materials used to construct disposable diapers (18) there may be unique analytical challenges to using this method for non-target analysis. Therefore, testing and characterization of contaminants arising from collection materials is necessary for successful non-targeted analyses of children’s urine collected with disposable diapers.

The aim of this study is to optimize a caregiver-driven protocol for collecting children’s urine via a combination of cotton pads and disposable diapers for analysis via liquid chromatography–high-resolution mass spectrometry (LC-HRMS). The use of cotton pads and commercially available disposable diapers can be a relatively low cost, acceptable, and caregiver-friendly method for the collection of children’s urine outside of the clinical setting. We evaluated processing methods and storage temperatures on amounts of specimen recovered from cotton pads used in combination with various widely commercially available diaper brands; these data informed our methodology for urine collection in real-world settings. Caregivers of children were recruited, and children’s urine was collected in caregivers’ homes over 24 h. Participant specimens were analyzed with an LC-HRMS method that accounted for contaminants leaching form collection materials. Finally, the applicability of this collection and analysis method to a wide range of children biomonitoring studies was demonstrated. We would like to emphasize that the presented data are used to specifically address the study objectives mentioned above. We believe, however, that this method, including sample preparation and data processing, can be tailored for researchers’ specific purposes.

METHODS

Materials, chemicals, and reagents

We purchased the hypoallergenic variety of four widely commercially available diaper brands: Luvs (Proctor & Gamble, Cincinnati, OH, USA; hereafter brand A), Pampers (Proctor & Gamble, Cincinnati, OH, USA; brand B), Seventh Generation (Seventh Generation, Burlington, VT, USA; brand C), and Huggies (Kimberly-Clark, Neenah, WI, USA; brand D). We summarize our process for testing and selecting the most effective and consistently performing diaper brand for this and future non-target analysis studies in Figure 1. Facial cotton pads were purchased from Shiseido Co. LTD (Japan). Centrifuge tubes containing a 0.45 μm cellulose acetate membrane insert were purchased from Thermo Fisher Scientific (Rockwood, TN, USA) and 50 mL glass syringes were from Popper & Sons, Inc. (New Hyde Park, NY, USA).

Figure 1. Summary of analyses for selecting brand of hypoallergenic disposable diaper for non-targeted analysis of children’s urine.

Chemical and reagent sources are provided in the Supplemental Information (SI). A mixed-analyte quality control (QC) solution was prepared at 1,000 μg/L in methanol and an internal standard solution of isotopically labelled compounds was prepared at 2,500 μg/L in methanol; solutions were stored at −20 °C until analysis. A list of compounds included in the mixtures is provided in the Supplemental Table.

Evaluating specimen recovery techniques and diaper storage temperature

We evaluated two techniques for recovering liquid specimens absorbed into cotton pads: 1) manual aspiration with a glass syringe by placing the wet pad in the barrel of the syringe and depressing the plunger to remove liquid (hereafter: syringe method) and 2) by centrifuging the wet pad in a tube containing an insert with a small-pore membrane filter to separate the liquid from the pad (hereafter: centrifuge method).

Six samples per diaper brand were prepared by placing a single cotton pad on top of the diaper’s absorbent material and slowly pouring 100 mL Milli-Q Water (MQW) directly onto the pad until absorbed. The diaper was folded and closed using its adhesive tags and stored in a resealable bag at room temperature for 24 h. The cotton pad was removed and the MQW recovered by either the syringe method or the centrifuge method (n = 3 per method, for each of four brands; total of 24 replicates). For the centrifuge method, tubes were centrifuged at 2,500 xg for 10 minutes. The specimen was collected in pre-weighed tubes and the mass of the recovered MQW determined by difference. To evaluate the effect of diaper storage temperature on the volume of liquid specimen recovered, we prepared (as described above) six additional samples per diaper brand and stored them for 48 h at either room temperature or refrigerated at 4 °C (n = 3 per temperature, for each of four brands; total of 24 replicates). The syringe method was used to recover MQW from all 48 h samples, and samples for which MQW was successfully recovered were analyzed via LC-HRMS (see below) to evaluate reproducibility of chemical leaching.

Participant recruitment, home visits, and urine collection

We recruited 11 caregivers of non-toilet-trained children aged <2 years living in the Baltimore/Washington metropolitan area to collect their child’s urine using cotton pads placed in disposable diapers for 24 h. All sample collection was completed between August and September 2022. The Johns Hopkins Bloomberg School of Public Health Institutional Review Board approved the study procedures.

A member of the study team arrived at the home of each child-caregiver pair to deliver a set of standardized diapers and collection materials and place and plug in a portable mini-refrigerator (Cooluli 20L, C20LDXBK, China) to store all soiled diapers until pick up. Twelve hypoallergenic diapers (in the size requested by the caregiver) were provided to each caregiver for use during the 24-hour collection period. The two diaper brands that performed similarly in the laboratory-based analyses described above were used to assess if differences in recovery existed after real-world use. Six diapers from each brand were included and caregivers alternated between brands after each diaper change. Members of the home visit team used a doll to demonstrate proper placement of 4 cotton pads directly on the child’s genitals prior to securing a disposable diaper. Caregivers were instructed to change their child’s diapers at the time and frequency typical for them, as dictated by the child’s voiding schedule and household routines. We instructed caregivers to retain all soiled diapers for 24 h by storing each diaper in an individual gallon-size ZipLoc Bags placed in the mini-refrigerator at 4℃. Caregivers marked a label placed on the bag to indicate if the soiled diaper contained urine only, stool only, or both. To minimize differences in chemical fingerprints attributable to use and contact with diaper wipes, we also provided caregivers with a single brand of hypoallergenic baby wipes to use exclusively during the diaper collection period. Caregivers were requested to avoid use of diaper creams or ointments during the diaper collection period, and to document use and the name of the product if use was necessary. When the study team dropped off and picked up the diapers, a member of the study team queried the caregiver about the child’s mood, whether the child was sick, received any medications (last 48 hours) or received any immunizations (last 72 hours). At diaper pick-up, we asked caregivers whether they encountered difficulties integrating our diaper collection protocol into their usual diapering routine.

Diapers were kept inside the mini refrigerator during transportation and for temporary storage in the laboratory until processing. The mini refrigerator was unplugged for no more than 1 h during transportation and was immediately plugged in and turned on upon arrival to the laboratory. Within 24 h of arrival in laboratory, all four cotton pads were removed from each soiled diaper and processed with the centrifuge method described above. Only two cotton pads could fit in the centrifuge tube insert, so two tubes were needed to process all 4 pads from a single diaper. Specimens from the two centrifuge tubes used for the same diaper were combined in a graduated cylinder and the volume measured to determine the amount of recovered urine per diaper. Then, specimens from a single participant were pooled in 50 mL centrifuge tubes by diaper brand and the caregiver’s indication of diaper contents (urine only, stool only, or both urine and stool); there were up to six possible diaper brand/diaper content pairs for each participant (e.g., brand A/urine only, brand B/urine only, brand A/stool only; brand B/stool only; brand A/both; brand B/both.). Urine specimens were refrigerated at −80 °C until LC-HRMS analysis.

Preparation of quality control samples

A pooled QC specimen was prepared from each of the individual participant samples described above. To avoid skewing the composition of the pooled QC towards any diaper brand/diaper content group, we adjusted the volume taken from each sample of a group so that the 6 groups were equally represented in the final composition of the pooled QC sample. The pooled QC was stored at −80 °C until LC-HRMS analysis.

Two “diaper blanks” were prepared to account for contaminant peaks arising from chemicals present in the collection materials, including cotton pads, diapers, and centrifuge tubes. Four cotton pads were placed on the absorbent material of one diaper from each brand given to caregivers. Then, 100 mL MQW was slowly poured on top of the pads and allowed to absorb. The diapers were folded and sealed using the adhesive tabs, placed in a resealable bag, and stored in the mini refrigerator at 4 °C for 24 h. Then, cotton pads were removed and processed as described above for soiled diapers, and the recovered MQW was stored at −80 °C until analysis.

Specimen preparation for LC-HRMS

Participant samples, pooled QC, and diaper blanks were thawed and 250 μL aliquoted to 1.5 mL microcentrifuge tubes. Fresh MQW was added to a microcentrifuge tube to serve as an extraction blank. Finally, two aliquots of reference urine (previously collected urine from one of the authors stored frozen in 1.5 mL microcentrifuge tubes) were thawed; one aliquot (hereafter: “fortified control”) was fortified with 10 μL of a mixed-analyte QC solution and the other was prepared like the other samples (“unfortified control”). Ten (10) μL of the internal standard solution were added to all samples, followed by 250 μL of 1% formic acid in MQW. Sample tubes were capped, vortex mixed for 10 s, and centrifuged at 20,000 xg for 5 min. Supernatants were transferred to amber autosampler vials containing 350 μL glass inserts.

The analytical sequence was structured based on previously published recommendations (19). The sequence started with two MQW injections, then the two diaper controls, followed by 10 injections of the unfortified control to condition the column; this was done to allow matrix to cover active sites in the system, improving the reproducibility and overall data quality of the analysis. The extraction blank was injected after the fourth injection of the unfortified control to measure systematic contamination. Then, the fortified control and the pooled QC were each injected twice, and then once each after 1/3 and 2/3 of the participant samples were injected (participant samples were analyzed in single injections); after all participant samples were injected the fortified control and the pooled QC were each injected two additional times (n = 6 each during the batch), followed by a reinjection of the extraction blank to measure possible carryover.

Instrumental analysis

Chromatography was performed with an UltiMate 3000 RSLCnano system (Thermo Scientific Waltham, MA, USA). The system was controlled by Xcalibur v4.1 software (Thermo Scientific). Separations were achieved with a Synergi Hydro-RP (150 mm x 1 mm i.d., 4 μm, Phenomenex, Torrance, CA, USA) with a KrudKatcher ULTRA HPLC in-line filter (Phenomenex). Aqueous mobile phase (A) was 1 mM ammonium fluoride and organic mobile phase (B) was methanol with the following time program: 10% B from 0-1 min, 10-98% B from 1-16 min, 98% B from 16-22 min, 98-10% B from 22-22.1 min, and 10% B from 22.1-30 min. The flow rate was 0.075 mL/min, column oven temperature was 45 °C, autosampler temperature was 10 °C, and injection volume was 5 μL. Analyte detection was accomplished with a Q Exactive HF high-resolution mass spectrometer (Thermo Scientific) with a heated electrospray ionization (HESI) source. Mass calibrations were performed before each batch. Data were acquired in Full Scan/data-dependent MS2 (dd-MS2) mode operating with polarity switching. Complete method and source settings are detailed in Table S1.

Participant specimens were analyzed twice using the method described above, with an exclusion list added to the method on the second analysis. The exclusion list contained ions detected in the diaper blanks with intensities above the threshold for triggering a dd-MS2 scan (2e5, Table S1) so that MS2 scans, important data for aiding in the structural determination of unknown compounds, would not be collected for ions resulting from specimen collection or preparation.

Data processing

The QC analytes added to the fortified control and internal standards added to all samples were evaluated for signal and retention time variability across multiple injections (n = 6 for QC analytes, n = 44 for internal standards) as a measure of the method’s performance; Quan Browser (v4.1, Thermo Scientific) was used to process data from these samples. For NTA data analysis, raw data files were processed with Compound Discoverer (v3.2, Thermo Scientific). This software incorporates a node-based workflow where each node handles a specific data processing task (e.g., retention time alignment, compound detection, compound grouping, database searching, etc.). The specific nodes that were incorporated in the workflow and their corresponding parameters are given in Table S2. Positive and negative mode data were processed separately. The software was used to process raw data from laboratory experiments evaluating MQW recovery after diaper storage for 48 h at different temperatures (see above), and participant and pooled QC specimens prepared after home visits.

The pooled QCs were incorporated in the workflow when processing participant samples to correct for variability observed during the sequence; compounds not detected in all pooled QC replicates or that did not meet reproducibility thresholds before and after data correction were filtered out of the data set (see Table S2 for settings). Integrated peak areas were exported and subjected to principal component analysis (PCA). To test whether diaper brand or diaper contents were significantly associated with PC1 or PC2 scores, a multivariate linear mixed model (MLMM) was performed with diaper brand, urine content, and stool content as fixed effects and participant as a random effect. Additional details on the statistical analysis are provided in the Supplemental Information.

Then, compounds with matches in themzCloud mass spectral database (Thermo Scientific) were selected for further review. Searching in this database uses fragmentation data, which can aid in structural identifications and produce higher confidence annotations compared to database searching with MS1 accurate mass data alone. Compounds with a database match score >80 were considered tentatively identified with confidence Level 2 based on previously published recommendations (20); at confidence Level 2 there is evidence for an unambiguous structural identification based on comparison to a library spectrum collected from a known compound. For this proof-of-concept analysis, we prioritized the list of Level 2 compounds to compounds known to be administered to children within the previous 48 h based on our questionnaire (see above), and compounds that were detected in at least one specimen from ≥9 participants (>80%); therefore, it should be noted that the subset of compounds discussed below do not represent the totality of compounds that can be identified with this method.

RESULTS

Evaluating specimen recovery techniques and diaper storage temperature

A comparison of MQW recovered from cotton pads with either the syringe or centrifuge method after storage for 24 h at room temperature is summarized in Figure 2A. For Brand D, ≤0.5 g MQW was recovered across all replicates, and no MQW was recovered from one sample processed with the syringe method and from one sample processed with the centrifuge method. Due to the poor specimen recovery this diaper brand was removed from all subsequent analyses.

Figure 2. Laboratory testing of diapers to assess specimen recovery.

A Mass of Milli-Q water (MQW) recovered from cotton pad replicates with a syringe or centrifuge method after storage for 24 h at room temperature in diaper brands A, B, and C. B Principal component analysis of MQW, recovered from cotton pad replicates, analyzed via LC-HRMS after storage for 48 h and 4 °C in diaper brands A, B, and C.

We could recover MQW from all cotton pad replicates used with Brands A, B, and C when stored for 48 h and 4 °C. However, MQW was recovered from fewer cotton pad replicates when stored for 48 h and room temperature (i.e., only 2 pads from Brand B and 1 pad from Brand C). The results of the LC-HRMS analysis of MQW recovered from cotton pads used with Brands A, B, and C after storage for 48 h at 4 °C are summarized as a principal component analysis (PCA) in Figure 2B; positive and negative mode data were merged post-processing to generate the PCA. Cotton pad replicates stored in Brand C showed the greatest variability, with replicates spread across the length of the x-axis in the PCA, while cotton pads stored in Brands A and B demonstrated lower variability. Cotton pads stored with Brand A were clearly separated from those stored with Brands B and C. Based on these laboratory results, diaper brands A and B were chosen for home visits to test specimen collection and analysis in real-world conditions.

Diapers soiled by children

We enrolled and collected soiled diapers from four females and seven males between the ages of 5 and 21 months (median =10) (Table 1). We collected a total of 70 diapers, and each child soiled between 5 and 9 (median = 6) diapers over the 24 h collection period. We recovered a mean of 44.7 (range = 26.7–71.1) mL total urine across diapers soiled by each child, and a mean of 7.3 (range = 2.4–15.2) mL per diaper; no specimen was recovered for 3/70 diapers (4%). At pickup, no caregivers reported illness or a non-normal mood in their children. No children received immunizations in the past 72 h; one child (participant 7) received ibuprofen and one child (participant 11) received penicillin in the previous 48 h. No caregivers reported using diaper creams or ointments during the 24 h collection period.

Table 1.

Summary of participant demographics and specimen collection

Participant No.	Age (months)	Sex	Number of diapers collected/soiled	Mean (range) volume of urine extracted per diaper (mL)*	Total volume of urine extracted (mL)*
1	12	male	5/5	7.5 (5.8-8.8)	37.7
2	21	female	6/5	10.0 (8.4-12.2)	50.2
3	11	male	6/6	9.0 (7.5-11.1)	54.1
4	8	male	5/5	5.3 (2.6-8.0)	26.7
5	11	male	8/7	7.2 (6.8-9.0)	50.2
6	10	female	6/5	6.3 (3.4-9.8)	31.6
7	6	male	5/5	8.6 (6.1-10.5)	42.8
8	15	male	7/7	7.7 (2.4-15.2)	53.7
9	5	male	9/9	7.9 (4.3-12.9)	71.1
10	7	female	6/6	5.6 (2.4-9.1)	33.3
11	8	female	7/7	5.8 (1.0-9.5)	40.6

^*Mean, range and total volume extracted were calculated using only soiled diapers (n=67)

Figure 3 shows the distributions of specimen volume recovered per diaper by brand (for 67/70 diapers where specimen was available). After specimen processing and pooling (see above) there were 32 total participant samples for LC-HRMS analysis: 9 samples were Brand A/urine only, 2 samples were Brand A/stool only, 4 samples were Brand A/urine and stool, 10 samples were Brand B/urine only, 3 samples were Brand B/stool only, and 4 samples were Brand B/urine and stool.

Figure 3. Distributions of volume of urine recovered from diaper brands A and B after field collection and laboratory processing.

Analysis of children’s urine

Performance data for QC compounds and internal standards are presented in the Supplemental Table. Retention times for all compounds were stable during the analysis, with CV 0.02-0.47% (n = 6 for QC compounds and n = 44 for internal standards). Peak area CV for QC compounds were 0.5-17.9%, and 30/35 compounds had peak area CV <10%. For internal standards, peak area CV were 6.4-22.7% and 5/7 internal standards had peak area CV <20%. Mass accuracies for all compounds were within ±2.16 ppm at the beginning of the batch, with mass accuracies for 37/41 compounds within ±1 ppm; by the end of the 34 h sequence mass accuracies were all within ±3.55 ppm, with mass accuracies for 22/41 compounds within ±2 ppm. When specimens were analyzed without an exclusion list, MS2 spectra were collected for 71.8% (4134/5760) of detected compounds, but when an exclusion list was added to the method the number of detected compounds with available MS2 spectra increased to 75.8% (4536/5982).

Figure 4 shows a PCA (with positive and negative mode data merged) of participant samples following the LC-HRMS analysis with samples labeled by diaper brand used (panel A) or the caregiver-designated diaper contents (panel B). Additionally, Figure S1 shows samples labeled according to participant. Results from the MLMM showed that PC1 and PC2 scores were not significantly associated with diaper brand or urine or stool content. However, it should be noted that these two principal components only accounted for ~36% of the total variation in the data. Output from running the MLMM are given in the SI. Table 2 summarizes compounds tentatively identified in participant specimens at a confidence Level 2, and comparisons of the experimental and library MS/MS spectra for each compound are given in Figures S2–S8. Participants 7 and 11 reported their child received ibuprofen and penicillin, respectively, within the last 48 h, and each compound was detected in all specimens from each participant. Desmethylcitalopram, an active metabolite of the antidepressant citalopram, was detected in all specimens from participant 10 with a high database match score of 95.6. We also detected several compounds used in various industrial applications in >80% of participants and the compounds were generally detected in multiple samples from the same participant regardless of diaper brand or diaper content. Available information on functional uses of these compounds can be found in Table S3.

Figure 4. Principal component analysis of participant specimens analyzed via LC-HRMS.

Participant samples are grouped by A diaper brand or B caregiver-designated diaper content

Table 2.

Tentatively identified compounds detected in participant specimens

Annotation	CAS-RN	Molecular Formula	Exact Mass	mzCloud Match Score	No. Participants Detected (detection rate)
Penicillin V*	87-08-1	C16H18N2O5S	350.09364	99.6	1 (9.1%)
Ibuprofen*	15687-27-1	C13H18O2	206.13068	96.7	1 (9.1%)
Desmethylcitalopram	62498-67-3	C19H19FN2O	310.14814	95.6	1 (9.1%)
Dimethyl decanedioate	106-79-6	C12H22O4	230.15181	99.1	11 (100%)
Caprolactone	502-44-3	C6H10O2	114.06808	98.6	11 (100%)
Caprolactam	105-60-2	C6H11NO	113.08406	84.5	9 (81.8%)
Hexanedioic acid	124-04-9	C6H10O4	146.05791	93.4	11 (100%)

^*caregivers indicated compound was administered to child within previous 48 h

DISCUSSION

This work describes the optimization of a caregiver-driven protocol for the collection of children’s urine for non-targeted analysis. Several urine collection methods have been employed for children who are not toilet-trained (14), including cotton balls or pads placed inside disposable diapers. Since urine collection and storage took place at caregivers’ homes in this study, one of our priorities was to minimize disruption to families’ normal routines. Therefore, we chose a method that included a high degree of familiarity for caregivers who were already accustomed to changing disposable diapers and emphasized training caregivers to effectively place and secure the cotton pads within the diapers. Most caregivers reported that collecting the diapers was not burdensome; others mentioned minimal burdens (one indicated doing more diaper changes than usual, and two noted that they didn’t use creams or ointments when they normally would have). Three caregivers reported the greatest deviation from their normal diapering routine during the collection period pertained to not using an overnight diaper with greater absorbency for nighttime use, as our protocol did not permit use of an overnight diaper.

A cited drawback when using cotton pads and disposable diapers to collect children’s urine is difficulty in aspirating the collected sample from the cotton pads (14). We tested two methods to recover sample absorbed to the cotton pads: manual aspiration with a glass syringe and centrifuging through a small-pore membrane. We were able to recover more sample with the centrifuge method than with the syringe method in our laboratory-based experiments (Figure 2A), in agreement with another published study utilizing a cotton wool plus diaper method for infant urine metabolomics (16). When the centrifuge method was used to process urine specimens collected in the field, the mean (range) volume of urine collected per diaper was 7.3 (2.4–15.2) mL. This collection and processing method fulfilled the specimen volume requirement of our non-targeted analysis (250 μL) and would likely be suitable for a range of studies, even those in which analyses can be performed with ≤1 mL specimen. Additionally, depending on the scope and objectives of the study, multiple specimens could be pooled to increase sample volume; in this study the mean (range) of total volume of urine collected in 24 h was 44.7 (26.7–71.1) mL.

Another benefit of processing samples with the centrifuge method is increased processing efficiency. The maximum number of diapers collected from any participant in 24 h in this study was 9, requiring 18 centrifuge tubes for processing. Using a centrifuge with a rotor that can accommodate 20 tubes, all specimens were processed and ready for storage in a single 10 min run. This is greatly improved over a manual syringe method that would require, in this case, either 9 different syringes to process each sample or cleaning of a single syringe between samples; processing samples in this way would likely consume considerably more time. Even if all samples were to be pooled, in which case a single syringe could be used for all samples from a single participant, physical fatigue or injury of laboratory personnel could occur if many samples need to be processed at a time. Greater amounts of contaminants leaching from cotton wool were also observed after processing via a manual squeezing technique compared to a centrifuging method (16). Disadvantages to the centrifuge method include the increased cost and amount of consumables needed to process all samples; however, we decided to prioritize maximizing the amount of volume recovered from each sample and minimizing processing time for this study.

The results from this pilot study will inform the development of a standardized specimen collection protocol to be used in a study to help identify suitable biomarkers to improve estimates of children’s soil and dust ingestion (21). To improve the standardization of the method, a single diaper brand will be used for all specimen collection, as previously recommended (17). We tested four widely commercially available diaper brands and selected each brand’s hypoallergenic line of diapers; this was done both to minimize potential adverse events (e.g., skin irritation) during the study and to minimize the number of compounds (e.g., fragrance or dye compounds) that could leach into the samples. Based on laboratory testing, diaper brands C and D were removed from further evaluation in caregivers’ homes due to large variability across replicates analyzed by LC-HRMS after storage (brand C, Figure 2B) and low sample recovery with either processing method (brand D). Therefore, we decided to test real-world specimen collection with both brands A and B to elucidate any potential differences; however, none were observed, including amount of sample recovered (Figure 3) or compounds detected (Figure 4A). Brands A and B are manufactured by the same parent company but are marketed differently and have different prices (e.g., Brand A typically retails for 0.16- 0.17 USD per diaper; Brand B retails for 0.25-0.32 USD per diaper). It is possible Brands A and B contain nearly the same, if not identical, materials, so similarities in their performance may be expected. Ultimately, a single brand, either A or B, will be chosen for all specimen collection in our larger study. Research groups conducting urine collection with diapers should undergo some testing, if possible, to ensure the diaper is suitable for their purposes and to improve protocol standardization.

In this demonstration study, we pooled participant urine for analysis rather than analyzing the urine from each diaper separately. This choice was made based on the application of the method in the aforementioned biomarker study. Different analytical choices, however, are possible using our approach; analysis of urine samples recovered from individual diapers would allow for an evaluation of the change of recovered compounds over the course of the day of diaper collection. Other factors, such as the toxicokinetics of desired compounds, could be used to make decisions about the protocol for diaper collection and sample pooling.

A consideration when refining this method was accounting for contamination arising from diapers, cotton pads, and other processing materials; diapers and cotton pads are known to leach contaminants into collected specimens (16, 17). An extraction blank is typically included in non-targeted analyses to account for contaminants introduced during the extraction procedure and instrumental analysis (19) but such a blank would not contain contaminants from collection materials. Therefore, “diaper blanks” were prepared and analyzed similar to participant samples. These diaper blanks allowed us to account for contamination arising from specimen collection and processing in two ways. First, we generated an exclusion list containing ions detected in the diaper blanks, which improved our data-dependent MS2 acquisition by preventing MS2 acquisition of contaminant ions and increasing the number of fragmentation scans for potential compounds of interest; this is considered a key step in successful data-dependent methods (22). Second, these specimens were included in our data processing workflow so background and contaminant ions could be flagged and subtracted from the results.

The results from analyzing the 32 participant samples did not show any clear differences based on caregiver-designated diaper contents (urine, stool, or both; Figure 4B). Among the 70 diapers collected, most samples (64, 91.4%) were labeled by the caregiver as containing liquid urine (either urine only or both urine and stool). For those diapers labeled as containing stool only, we were still able to recover liquid from the cotton pads after centrifuging. This could be due to the diaper changing habits of the caregivers—they may change diapers even if the wetness indicator on the diaper was not activated. For this study, caregivers were instructed by the study team to change diapers according to their typical routine to minimize disruption.

Specimens collected for this study were analyzed by non-targeted analysis. To minimize loss of analytes, specimens were prepared with a simple dilution and precipitation procedure. A fortified matrix-matched control, containing known compounds, was included in the analysis to monitor the performance of the method during the analytical batch as previously recommended (23, 24). Additionally, a pooled QC was included in the analysis to correct for changes in instrument sensitivity and to flag compounds that were reproducibly detected during the analysis (19). As shown in Figure 4, the pooled QCs were well clustered towards the center of the PCA plot, indicating reproducible data acquisition during the analysis.

While our development of this approach was intended for use in a study of children’s soil and dust exposure (21), we believe this collection and analysis method can be widely applied to biomonitoring studies of infants and children incorporating non-targeted analysis. As proof of concept, we tentatively identified compounds that could be of interest to different research areas (Table 2). Ibuprofen (administered to participant 7) and penicillin (administered to participant 11) were each detected in the respective participant’s specimens analyzed with this method. Desmethylcitalopram, an active metabolite of citalopram, was detected in all specimens from a single participant. It was not disclosed to our study team that citalopram was administered to the child. However, it is possible the child’s mother was taking citalopram at the time of specimen collection; citalopram and its metabolite are known to partition into breastmilk (25) and could represent an exposure route for the child. We did not collect breastfeeding status or data on maternal pharmaceutical use as it fell outside the scope of our study, but other studies may consider collecting such data if breastfed children are to be included. We also detected several compounds with various industrial applications in a large percentage (>80%) of participants. Publicly available data for these compounds (summarized in Table S3) show that they are used for a variety of functions (e.g., monomer, plasticizer, fragrance) and may be included in a wide range of products, making it difficult to elucidate specific exposure sources. It is important to note that these compounds were tentatively identified based on matches to a mass spectral database and identities would need to be confirmed with reference standards for further evaluations. Additionally, these data are not meant to represent the totality of chemical detections possible with this method, and this method, including sample preparation and data processing, can be tailored for researchers’ specific purposes. However, we feel these data demonstrate the applicability of this method to a wide range of children biomonitoring studies, including metabolomics and environmental contaminant exposures.

In this study, we optimized a low-burden caregiver-driven method for the collection of children’s urine for non-targeted analysis. Urine was successfully collected at caregivers’ homes over 24 h using cotton pads placed inside disposable diapers, and the specimen was recovered by centrifuging cotton pads through a small-pore membrane. Two widely commercially available hypoallergenic diaper brands performed similarly when evaluated in laboratory-based experiments and during field collection. A non-targeted method was implemented to analyze participant specimens and incorporated an exclusion list of ions representing compounds leaching from collection materials, increasing the number of fragmentation spectra for potential compounds of interest. Fortified and pooled QC specimens were analyzed during the sequence to measure performance of the method and demonstrated that data acquisition was reproducible across the analysis. Finally, several compounds were tentatively identified via spectral database matching, demonstrating the applicability of this collection and analysis method to a wide range of children biomonitoring studies.

Impact Statement:

Infant and children urine is a valuable matrix for studies of the early life exposome, in that numerous biological markers of exposure and outcome can be derived from a single analysis. Depending on the nature of the exposure study, it may be the case that a simple collection method that can be facilitated by caregivers of young children is desirable, especially when time-integrated samples or large volumes of urine are needed. We describe the process for development and results of an optimized method for urine collection and analysis using commercially available diapers and non-target analysis.

Data Availability Statement

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.