Detection of Trace Drugs of Abuse in Baby Formula using Solid Phase Microextraction Direct Analysis in Real Time Mass Spectrometry (SPME-DART-MS)

Abstract

The intentional or unintentional adulteration of baby formula with drugs of abuse is one of the many increasingly complex samples forensic chemists may have to analyze. This sample type presents a challenge because of a complex matrix that can mask the detection of trace drug residues. To enable screening of baby formula for trace levels of drugs the use of solid phase microextraction (SPME) coupled with direct analysis in real time mass spectrometry (DART-MS) was investigated. A suite of five drugs were used as adulterants and spiked into baby formula. Samples were then extracted using SPME fibers which were analyzed by DART-MS. Development of a proof-of-concept method was completed by investigating the effects of the DART gas stream temperature and the linear speed of the sample holder. Optimal values of 350 °C and 0.2 mm/s were found. Once the method was established, representative responses and sensitivities for the five drugs were measured and found to be in the range of single ng/mL to hundreds ng/mL. Additional studies found that the presence of the baby formula matrix increased analyte signal (relative to methanolic solutions) by greater than 200 %. Comparison of the SPME-DART-MS method to a traditional DART-MS method for trace drug detection found at least a factor of 13 improvement in signal for the drugs investigated. This work demonstrates that SPME-DART-MS is a viable technique for the screening of complex matrices, such as baby formula, for trace drug residues and that development of a comprehensive method is warranted.

Drug chemists are regularly faced with increasingly complex samples for analysis. These cases can involve new or difficult to identify compounds such as novel psychoactive substances or can be challenging matrices such as foods, oils, and waxes. For the complex matrices, the use of multi-step extractions and sample cleanup procedures can require substantial amounts of time from a drug chemist just to determine whether a drug of abuse is present in the sample. The development of tools and screening processes that allow chemists to clean-up and analyze their samples in a less time-consuming manner and can aid in identifying if a more thorough workup is warranted, would be beneficial.

One such complex matrix that may be submitted for forensic analysis is baby formula. Cases containing baby formula may be submitted because of suspected intentional or unintentional exposure or overdosing of infants to drugs of abuse. Infant exposure to drugs of abuse through formula can occur through a variety of means that have been documented in the literature. Intentional adulteration of baby formula is often associated with trying to get infants to fall asleep, which is highlighted in a 2011 poll of 26,000 mothers, where one in twelve mothers reported giving their children sleep medication(1). This practice has led to overdoses and, in some cases, infant deaths(2). In one such instance a two-month-old was given over the counter medication via a baby bottle that contained lethal doses of pseudoephedrine, brompheniramine, and dextromethorphan(3). Accidental poisonings with illicit drugs such as fentanyl have also been reported due to contamination of baby bottles(4). These accidental poisonings are of concern since approximately 7.5 million children, 17 years and younger, live with at least one parent who abuses drugs(5), which leads to increased risk of exposure. In a study that investigated opioid related poisonings between 1997 and 2012, a three-fold increase in poisonings of children and adolescents aged one to four was found(6).

The need to accurately and sensitively detect drugs of abuse in suspected adulterated baby formula is an important analytical challenge for which methods need to be developed. The complex matrix of baby formula, which contains a protein base (such as cow’s milk or soy), vitamins, salts, minerals, nucleotides, fats, carbohydrates, and emulsifiers, is not amenable to a simple dilute and shoot approach that is often used for drug analysis. Instead it requires some form of sample extraction or cleanup. Extraction and cleanup steps are especially critical in instances when only trace amounts of material remain, or low concentrations of drugs are expected. While there is minimal published literature focusing on the detection of drugs of abuse in baby formula, the detection of other adulterants, specifically melamine, has been extensively studied. Much of the research in this area has used chromatographic-based techniques, such as gas chromatography mass spectrometry (GC-MS), liquid chromatography mass spectrometry (LC-MS), and liquid chromatography with UV detection (LC-UV) coupled with extraction procedures(7). Liquid/liquid or solid phase extraction are commonly used, and methods developed for baby formula or milk have limits of detection in the range of 0.01 mg/kg to 1 mg/kg(7).

With the continued push for more rapid, and simpler, sample analysis approaches other techniques have also been evaluated. Direct analysis in real time mass spectrometry (DART-MS) is one of the ambient ionization mass spectrometry techniques that has been shown to readily detect melamine in powdered milk and milk(8,9). DART-MS has also been widely researched for detection of drugs of abuse, and is being increasingly adopted by forensics laboratories (10–12). For grossly adulterated baby formula, detection of drugs of abuse could likely be completed using the traditional direct analysis of the sample. However, for low concentration or trace residue samples the use of a pre-separation or extraction step, such as solid phase microextraction (SPME), to remove a portion of the chemical background and thereby lower competitive ionization effects, would likely have to be deployed. SPME-DART-MS has been demonstrated for forensic applications, including screening of toxicological samples(13,14) and the analysis of psychoactive plant species(15) by introducing the SPME fibers directly into liquid samples or by sampling the headspace surrounding a sample. SPME-DART-MS has also been demonstrated for the analysis of food-based matrices(16). The approach provides simple sample preparation with minimal user intervention allowing for faster analysis times and decreased solvent usage compared to other, more traditional, extractions. SPME has also been shown to provide improved detection limits compared to traditional analyses(17).

In this work, a proof-of-concept study was completed to identify whether or not screening for trace amounts of drugs in baby formula is possible using SPME coupled with DART-MS. Fiber-based SPME tips were used to extract a panel of five drugs (cocaine, dextromethorphan, fentanyl, heroin, and lorazepam) from baby formula. Optimization of the method, establishment of method sensitivities, and comparison to traditional DART-MS analysis were also completed. The results of this work highlight that this a potentially valuable approach for screening that warrants further investigation and development of a validated method.

Materials and Methods

Materials and Chemicals

The five drugs examined in this study (cocaine, dextromethorphan, fentanyl, heroin, and lorazepam) were purchased as 1.0 mg/mL solutions in methanol or acetonitrile. Cocaine and heroin were purchased from Cayman Chemical (Ann Arbor, MI) while the remaining were purchased from Cerilliant (Round Rock, TX). Working solutions were created from the stock solutions through volumetric dilution with methanol (Honeywell, Charlotte, NC), to concentrations of 10 μg/mL, 1 μg/mL, and 100 ng/mL. Aliquots of these solutions were spiked into the baby formula or diluted further in methanol. Similac Advance, 0–12 months, Infant Formula with Iron (Similac, Columbus, OH) was used for this study and was diluted with an equal amount of water, as per the manufacturer’s instructions, prior to use. A solid phase extraction (SPE-it) kit from IonSense (Saugus, MA) with C18 extraction fibers was used for sample extraction. Chromasolv grade water, (Millipore-Sigma, Burlington, MA) methanol (Honeywell), and acetonitrile (Millipore) were used for fiber conditioning.

Sample Creation and Extraction

Prior to use, the C18 fibers were conditioned by placing them into 2 mL amber glass vials (Restek, Bellefonte, PA) containing 1.0 mL of a 50:50 methanol : water mixture. The vials were loaded into the foam insert of the microplate shaker from the SPE-it kit, and the fibers were conditioned for 30 min while shaking at 52.4 rad/s (500 RPM). After conditioning, the fibers were placed in vials containing 1.0 mL of LC-MS Grade water and shaken for approximately 5 s at 52.4 rad/s (500 RPM). They were then placed into a 1.0 mL aliquot of baby formula spiked with one of the five drugs at the desired concentration and allowed to extract for 60 min while shaking at 52.4 rad/s (500 RPM). For comparison purposes, samples containing just methanol spiked with the drug were also analyzed. Finally, the fibers were again placed into vials containing 1.0 mL of LC-MS Grade water and shaken for approximately 5 s at 52.4 rad/s (500 RPM). Fibers were then loaded onto the SPE-it insert for the linear rail of the DART source for analysis. The fibers in his study were not reused.

Instrumentation and Sample Analysis

The fibers were analyzed using the SPME holder attached to the linear rail that was mounted to the DART source. A DART-SVP ion source (IonSense) was used for analysis and was coupled with a JMS-T100LP time-of-flight 4G LC-Plus mass spectrometer (JEOL, Peabody, MA). DART parameters that were kept constant throughout the study included the use of helium as the DART ionization gas and an exit grid voltage of +50 V. Unless otherwise stated a gas stream temperature of 350 °C and a linear rail speed of 0.2 mm/s were used. The “12 DIP-it” method within the DART-SVP software was used for controlling the linear rail and had constant parameters of 5 s for the heater wait time and 1 s for the contact closure delay. Mass spectrometer parameters used throughout the study included operation in positive ionization mode, an orifice temperature of 125 °C, an orifice 1 voltage of +20 V, a rings lens voltage of +5 V, an orifice 2 voltage of +5 V, an ion guide RF voltage of 400 V, and a detector voltage of 2300 V. Full scan mass spectra, from m/z 60 to m/z 700 were collected at 1 scan/s.

Results & Discussion

Method Optimization

The first step of this study was optimizing a method for detection of the drugs in both methanol and baby formula using three of the drugs of interest: cocaine, heroin, and lorazepam. These drugs were chosen because they have either high (cocaine) or low (heroin and lorazepam) ionization efficiencies, and therefore are easier (cocaine) or more difficult (heroin and lorazepam) to detect. Considering that there is a significant body of existing work on drug detection by DART-MS(10,18) it was necessary to optimize only two parameters that may be affected by the implementation of SPME tips and sampling procedures – the DART gas stream temperature and the linear rail speed. The linear rail speed needed to be optimized as it is not a parameter used in traditional DART-MS analysis. The DART gas stream temperature was optimized because use of the linear rail causes the DART to be placed further from the mass spectrometer inlet than with traditional DART analyses, which could affect the temperature experienced by the sample, and therefore effect desorption of the analyte.

The DART gas stream temperature was optimized by evaluating six temperatures (250 °C, 300 °C, 350 °C, 400 °C, 450°C, and 500 °C) at a linear rail speed of 0.2 mm/s. Three replicates of SPME fibers exposed to a 10 μg/mL solution of the individual drugs, in both methanol and baby formula, were analyzed at each temperature. The integrated peak areas from the extracted ion chronographs (EICs) for the base peak of each drug were used for comparison, the results of which are shown in Figure 1.

FIGURE 1.

Integrated peak areas of cocaine (A), heroin (B), and lorazepam (C) in methanol (blue circle) and baby formula (orange diamond) as a function of DART gas stream temperature. Error bars show the standard deviation of 3 replicate measurements. Note the secondary y-axis for lorazepam in methanol (C).

Optimization of the DART gas stream temperature for drug solutions in both the methanol and baby formula gave varying results. The methanol solutions, which were examined to understand the response independent of the complex matrix baby formula present, produced an optimal response at 450 °C for cocaine and heroin and 300 °C for lorazepam. Interestingly, however, an optimal DART gas stream temperature of 350 °C was found for all three drugs in the presence of the baby formula matrix. Differences in the optimal desorption temperatures for the two matrices are potentially due to compounds in the baby formula that alter the desorption temperature of the analyte of interest or the co-desorption of other compounds that will either enhance or competitively ionize with the drug. It also appears that lorazepam is less thermally stable than the other two drugs, as a significant decrease in signal was observed at temperatures above 350 °C for methanol and baby formula solutions. Flash desorption or thermal degradation likely also drives the decrease in cocaine and heroin signal above 400 °C.

The second factor that was optimized was the linear rail speed, which controls the amount of time a sample is interrogated by the DART gas stream. Both methanol and baby formula solutions were studied, using a DART gas stream temperature of 450 °C for the methanol solutions and 350 °C for the baby formula solutions. Three rail speeds, representing the lower (0.2 mm/s), middle (1.5 mm/s), and upper (10 mm/s) bounds of the system were investigated and the integrated peak areas from the EICs of three replicate measurements were calculated. Replicate measurements were individual tips analyzed in series. The results of this experiment are shown in Figure 2.

FIGURE 2.

Integrated peak areas of cocaine (A), heroin (B), and lorazepam (C) in methanol (blue) and baby formula (orange striped) as a function of linear rail speed. Error bars show the standard deviation of 3 replicate measurements. Note that peak areas are plotted on a log scale.

For all three drugs studied, regardless of their presence in methanol or baby formula, optimal results were obtained using the slowest rail speed (0.2 mm/s). This was expected as the slower rail speed allows for a factor of six greater time for the SPME fibers to interact with the DART gas stream and therefore allows for more thorough desorption of the analytes. The approximate times that the fibers were in the DART gas stream for the three linear rail speeds were 18 s, 3 s, and 1.8 s, for the 0.2 mm/s, 1.5 mm/s, and 10 mm/s speeds, respectively. At faster linear rail speeds, responses from the adjacent SPME fibers began to overlap, which made integration of the peaks and differentiation of spectra from individual fibers difficult. Because of this, for all further experiments a DART gas stream temperature of 350 °C and a linear rail speed of 0.2 mm/s were used. Under these parameters, the total time for analysis was approximately 5 min, coupled with a sample preparation time of approximately 100 min, of which approximately 10 min required manual intervention.

Representative Responses & Limits of Detection

After development of the optimized method, all five drugs of interest were investigated to establish their representative responses. Regardless of solution type (methanol or baby formula), all drugs formed protonated molecules, [M+H]⁺, as expected. No adducts with other cationic species, such as ammonium or sodium, were observed. Comparing the spectral response from a blank SPME filter and one from the pure baby formula (Figures 3A, 3B) highlights that the SPME extract was able to remove some of the matrix components but, as expected, was not selective to just the drug. The C18 SPME fibers used in this work were able to readily extract all the drugs of interest, as shown in Figure 3C and 3D as well as Supplement Figures S1 and S2. Without background subtraction, visual identification of the drug signature was difficult at low concentrations, especially for lorazepam (Supplemental Figure S2). Detection at these lower concentrations can be aided by using mass spectral search software or, if possible, spectral subtraction of a clean formula sample. Given that the analytes were present in a complex matrix, and the method was being investigated for screening purposes, only the low orifice voltage, which provides minimal in-source collisionally induced dissociation (is-CID), was used. Utilization of higher is-CID voltages, especially those exceeding +40 V, produced noisier spectra due to fragmentation of the drug and matrix components, making detection more difficult.

FIGURE 3.

Representative mass spectra of a blank SPME fiber (A), baby formula (B), and baby formula doped with 10 μg/mL solution of fentanyl that has been background-subtracted (C) and has not been background-subtracted (D).

Once the representative mass spectral response of the baby formula matrix was established, the approximate sensitives of the method for all five drugs were established. Approximate sensitivities were determined by analyzing the drugs in decreasing concentrations, in triplicate, until a signal to noise ratio (S/N) close to but above 3:1 was obtained. The S/N ratios were calculated using the peak to peak method in the MassCenterMain software, native to the JEOL system. The EIC of the protonated molecule was used for S/N calculation. Peaks were automatically identified by the software and the background (noise) range was selected by manually choosing the interval where a negative control sample (SPME fiber exposed to baby formula without any drugs) was analyzed. The approximate sensitivities and their respective S/N ratios are shown in Table 1. Cocaine was found to be the most sensitive drug investigated, which was expected given its high ionization efficiency, and is in line with previous drug detection research showing lowest detection limits for cocaine(11,19). Dextromethorphan, which can be added to baby formula through cough syrup, also showed excellent sensitivity. It is important to note that dextromethorphan cannot be distinguished from levomethorphan by DART-MS. Heroin and lorazepam were the least sensitive compounds, due to either poor extraction, poor desorption, or poor ionization efficiencies. Lower sensitivity for these drugs, compared to the others, has been previously reported(19). Representative mass spectra of the five drugs, at their respective approximate sensitivities, are shown in Figure S3.

Table 1.

Base peaks, approximate sensitivities and corresponding signal-to-noise ratios for the five drugs studied as well as estimated lethal doses. Uncertainties represent the standard deviation for triplicate measurements. All estimated lethal doses assume a starting volume of baby formula of 240 mL. These values also assume the lethal dose for adults, since values for infants were not obtainable.

Drug	Base Peak (m/z)	Approximate Sensitivity (ng/mL)	Average S/N	Estimated Lethal Dose (ng/mL)	Adult Lethal Dose Reference
Cocaine	304.155 [M+H]+	2.5	7.3 (±1.1)	5.0×106	(20)
Dextromethorphan	272.201 [M+H]+	5	6.4 (±0.6)	6.2×105	(21)
Fentanyl	337.228 [M+H]+	10	13.7 (±1.4)	8.3×104	(22)
Heroin	370.165 [M+H]+	250	14.2 (±1.4)	2.0×105	(20)
Lorazepam	322.020 [M+H]+	100	6.1 (±1.5)	4.1×104	(23)

While sensitive detection of the drugs is desired, the method is only practical if the analytes can be detected at, or below, levels of toxicological relevance. Table 1 also displays estimated lethal doses for the five drugs in baby formula using published lethal dose data(20–23). The limitations of the availability of data lead to two assumptions for these estimates. First, the estimated lethal dose was calculated using available data for adults. Second, the concentrations reported assume a bottle size of 240 mL as this represents the largest bottle used for feeding infants [16]. Using these estimates, the lethal doses ranged from 41 μg/mL to 5 mg/L, which are two to four orders of magnitude higher than the lowest detectable concentrations. As Table 1 shows, the estimated lethal doses are significantly higher than the approximate sensitivities for the five drugs. The sensitivity of the method shows that detection of non-lethal doses and residues are readily achievable using this technique.

Signal Suppression/Enhancement

With ambient ionization techniques, like DART–MS, competitive ionization is a common occurrence. Since all analytes are desorbed nearly simultaneously, high ionization efficiency compounds can, at times, complicate detection of analytes of interest that may not be as easily ionized (17). In some instances, ionization of the analyte can be enhanced by other compounds in the matrix. While complete competitive ionization, in which detection of an analyte is not possible, was not observed in this work, the effect of the matrix was still investigated. Aliquots of the methanol and baby formula solutions containing 10 μg/mL of cocaine, heroin, and lorazepam were analyzed in triplicate and the peak areas from the EICs were calculated (Table 2). For all three drugs, a significant enhancement in signal was observed for the formula–containing solutions. This enhancement may be driven by the presence of additional compounds that promote ionization of the drug species.

Table 2.

Average peak areas for cocaine, heroin, and lorazepam in baby formula and in methanol. Uncertainties represent the standard deviations of triplicate measurements.

Drug	Peak Area in Baby Formula (Counts)	Peak Area in Methanol (Counts)	Percent Change in Baby Formula
Cocaine	3.0×107 (±2.6×106)	9.7×106 (±6.9×105)	209 %
Heroin	1.4×107 (±7.7×105)	2.7×106 (±9.4×105)	418 %
Lorazepam	3.0×105 (±3.2×104)	9.4×103 (±3.0×103)	3091 %

Comparison to Traditional DART-MS

The final component of this study compared the SPME-DART-MS to traditional DART-MS analysis to understand the magnitude of enhanced sensitivity offered by the SPME-DART-MS configuration. To evaluate this, analysis of cocaine, heroin, and lorazepam using both the SPME-DART-MS method and an identical DART-MS method was completed. The only difference between the two methods was sample prep-treatment (extraction or not) and sample introduction – where glass microcapillary sampling rods were used for traditional DART-MS. A concentration of 10 μg/mL of the drug in baby formula was analyzed. Both the integrated peak areas (Figure 4 A. ) and mass spectral responses were compared. For all three drugs, the response of the drug using SPME-DART-MS compared to conventional DART-MS showed improvements in signal ranging from a factor of 13 for lorazepam to a factor of 118 for heroin. Interpretation of the mass spectra was also simplified greatly using SPME-DART-MS, as highlighted in Figure 4 B and 4 C. The use of the SPME fiber was shown to greatly eliminate matrix signal, resulting in the base peak associated with the drug for all spectra.

FIGURE 4.

Integrated peak areas of cocaine, heroin, and lorazepam for SPME-DART-MS (blue circle) and traditional DART-MS (orange diamond) (A). Error bars show the standard deviation of 3 replicate measurements. Note that peak areas are plotted on a log scale. Also shown in a mass spectral comparison of the 10 μg/mL cocaine in baby formula solution when analyzed by SPME-DART-MS (B) and traditional DART-MS (C) which have not been background subtracted.

Conclusions

SPME-DART-MS provides a sensitive method for the screening of cocaine, dextromethorphan, fentanyl, heroin, and lorazepam in baby formula. Using an optimized method, consisting of a DART gas stream temperature of 350 °C and a linear rail speed of 0.2 mm/s, sensitivities in the range of single to hundreds of ng/mL were obtained. It was observed that the baby formula matrix caused ion enhancement when compared to a methanolic solution containing the same drug. It was also demonstrated that the SPME-DART-MS provided an enhancement in signal intensity and a more easily interpretable mass spectrum than traditional DART-MS. While this has been shown for the baby formula examined, additional work is investigating whether these trends are observed across multiple types of baby formula. Current efforts are also focused on expanding the list of drugs that could be used to adulterate the baby formula and investigating the potential to use this tool for quantitative measurements. Utilizing a method such as this would allow chemists to sensitively detect drugs in baby formula in cases where there is suspicion of intentional or unintentional addition of drugs into the formula.