Advanced in Applications of GC-MS and LC-MS 

Advances in Clinical and Forensic Analysis: Focus on Cannabis 

Moving Toward Green and Sustainable Sample Preparation

The Potential of Automated Strategies in Microextraction Procedures 

Determination of Fipronil and Its Metabolites in Eggs and Environmental Matrices by LC–MS/MS

Intact Mass Analysis of Biopharmaceuticals as its Own Unique Application

Part III: Physicochemical Causes of Peak Tailing

 But My Peaks Are Not Gaussian! 

Bis(pinacolato)diboron (BPD) and tetrahydroxydiboron (BBA) are commonly used reagents in Suzuki-Miyaura coupling for the synthesis of many intermediates and active pharmaceutical ingredients (APIs). These reagents have tested Ames positive and are classified as mutagenic by the International Council for Harmonization (ICH), under the “Assessment and Control of DNA Reactive (Mutagenic) Impurities in Pharmaceuticals to Limit Potential Carcinogenic Risk,” M7 (

) as Class 2 impurities. Because BPD and BBA are mutagenic, they need to be controlled and subsequently detected down to very low levels. Current methods using inductively coupled plasma–mass spectrometry (ICP-MS) give only the total boron content, which is problematic if the API also contains a boron molecule. In this paper, we describe a novel gas chromatography (GC)–MS approach for the low-level analysis of BPD and BBA in drug substances, even in the presence of boron from other sources. The analysis of BBA required derivatization of BPA to BPD, resulting in quantitation limits of 30 ppm of BBA (back calculated from the derivatized compound). BPD could be analyzed directly with quantitation limits of 2.5 ppm.

The Suzuki-Miyaura reaction forms a carbon–carbon bond, typically using a palladium catalyst, and has become routine in synthesizing pharmaceutical products (1). Bis(pinacolato)diboron (BPD) and tetrahydroxydiboron (BBA) are commonly used reagents in Suzuki-Miyaura coupling for synthesizing many intermediates and active pharmaceutical ingredients (APIs) (2). A recent publication by Hansen and others (3) reported that BPD and BBA tested positive in the Ames assay and are considered Class 2 impurities based on the International Conference on Harmonization (ICH) M7 “Assessment and Control of DNA Reactive (Mutagenic) Impurities in Pharmaceuticals to Limit Potential Carcinogenic Risk” (4). As such, they should be closely monitored and controlled at or below acceptable limits (acceptable threshold of toxicological concern, or TTC) (4). As outlined in the guidance, acceptable total daily intakes for individual impurities for the duration of treatment greater than 10 years and between 1–10 years are less than 1.5 μg/day and 10 μg/day respectively, for an estimated maximum daily dose of 0.2 g (as is the case with the API studied). The corresponding acceptable TTC limits for BPD or BBA are 7.5 ppm and 50 ppm, respectively.

Previously, colorimetric methods have been used to determine low levels of boron (5), and Sah and Brown (6) have reviewed more modern techniques for determining boron content and the different boron isotopes. Atomic emission or absorption spectrometry have shown poor sensitivity. Prompt-γ spectrometry can calculate the boron concentration in solid samples, which makes it useful; however, it is not suitable for low levels of boron. Inductively coupled plasma–optical emission spectrometry (ICP-OES) and ICP–mass spectrometry (ICP-MS) have been used to analyze total boron content in APIs. ICP-MS is the more sensitive technique; however, it only quantifies the total boron in an API and cannot differentiate between a boron molecule in the structure of the API (if one exists) and impurities from the Suzuki-Miyaura coupling or any other boron-related impurities in the API (7). To have a more generic methodology, chromatographic quantitation, which can separate the boronic species, is desirable.

A high throughput analysis of boronic acids with aromatic groups has been reported by Pandiiyan (8) and used for reaction monitoring samples of Suzuki-Miyaura coupling. A lack of chromophores and poor ionization make it difficult to analyze BBA by high performance liquid chromatography (HPLC) or LC–MS, whereas a lack of solution stability under aqueous conditions makes it difficult to analyze BPD by LC–MS. However, synthetic chemists working with boronic compounds have been using a combination of gas chromatography (GC) and GC–MS to monitor chemical reactions (9–11).

To determine the feasibility of detecting the individual boron-containing compounds in the absence of any matrix or API, BBA and BPD were analyzed by GC with a flame ionization detector (FID) and GC–MS. For this portion of the method development, BBA and BPD were dissolved in an aprotic solvent (carbon tetrachloride) to avoid any decomposition of the components at low concentrations in the solution. The precision of direct injection GC improves when using an internal standard because of high flow rates and split ratios (12). A commercially available boronic compound with a similar structure and fragmentation pattern to BPD, bis(hexyleneglycolato)diboron (BHGD), proved to be a useful internal standard for quantifying BPD down to 0.09 μg/mL. Although BBA was not detected when analyzed by direct injection GC, its derivatization to BPD allows it to be detected by both FID and MS detectors, whereas BBA could not.

BPD was employed for the synthesis of API-A. The method development work in Part 1 was successful in developing a quantitative method for the BPD solution, which used carbon tetrachloride (CCl4) as the diluent. However, for this study, dimethyl sulfoxide (DMSO) was selected as the diluent because API-A was not soluble in CCl

. Using the results from Part 1, a method was developed for the quantification of BPD in the API-A matrix also using BHGD as an internal standard. The method detection limit achieved was 0.16 μg/mL of BPD.

In contrast, API-B was synthesized using BBA. To improve the detection of BBA using GC–MS, BBA was derivatized to BPD using pinacol (Scheme 1). The amount of BBA in the API-B matrix was analyzed by GC–MS after BBA was derivatized to BPD and could be quantified down to a limit of 2.5 ppm. Because the final product (BPD) could be purchased, the efficiency of derivatization was able to be monitored.

SCHEME 1: Structures of compounds and studied fragments.

All compounds (BBA, BPD, BHGD, and pinacol) for this study were sourced from Merck KGaA, with the exception of carbon tetrachloride (CCl

), which was purchased from Acros Organics (part of Thermo Fisher Scientific). GC–MS analyses were performed on an Agilent 7890B GC system coupled with a 5977A mass selective detector with an internal splitter (~1:10) to both a mass spectrometer (MS) and a flame ionization detector (FID) and Mass Hunter software. The GC separation was achieved on an Agilent HP-5MS having the dimensions 30 m x 0.25 mm and a film thickness of 0.25 μm.

For the GC program, high purity helium (Airgas) was used as a carrier gas with a constant flow of 2 mL/min. The column oven temperature was initially set at 80 

C, and held for 1 min, then raised at a rate of 2.5 

C and held for 1 min. The sample was injected using a 7683B auto sampler. The inlet temperature was held at 240 oC and operated in a split mode with a 10:1 split ratio. An ultra-inert deactivated inlet liner with glass wool for split injections (Agilent Part No. 5190-3163) was used. The retention times (

) of the components studied are as follows: CCl

: 1.65 minutes, BPD: 5.49 min, and BHGD: 5.58 min.

The Agilent 7890N was interfaced with the mass spectrometer (operating in electron ionization, or EI, mode) by a direct transfer line maintained at 250 oC with the source temperature set at 230 oC. The mass spectrometer was operated in both total ion count (TIC) and single ion monitoring (SIM) modes. The SIM was set to scan for two ions at 

 69.1 and 239.2 as these ions are common to both BPD and BHGD electron ionization fragmentation (see Scheme 1). It is noted that the fragments of interest are identical; however, because the compounds can be separated by chromatography, each of these compounds can be used as an internal standard for the other. By collecting the ion count chromatograms for the same ions, internal standard calculations were made. The ion at 

 84.1 is also a strong ion for both of these compounds; however, CCl

 interferes with the signal so that ion could not be used in Part 1 of the study.

Part 2: Detection of BPD in API-DMSO Matrix by GC–MS

Analysis of BPD was conducted by GC–MS with BHGD as the internal standard. The chemicals used were anhydrous dimethyl sulfoxide (DMSO) (Merck KGaA), BPD (Aldrich, 473294-25G), and BHGD (Aldrich, 525685-1G). Anhydrous DMSO was further treated with magnesium sulfate as a drying agent in a ratio of 60–65 mg magnesium sulfate to 1 mL DMSO.

GC–MS analyses were performed on an Agilent 7890B GC system coupled with a 5977A mass selective detector. The GC separation used an identical column to the one used in Part 1.

For the GC program, high purity helium was used as a carrier gas with a constant flow of 2 mL/min. The column oven temperature was initially set at 80 

C, and held for 3 min, then raised at a rate of 15 oC/min to a final temperature of 250 

C and held for 5 min. The sample was injected using a 7683B auto sampler. The inlet temperature was held at 250 oC and operated in a split mode with 100:1 split ratio. An ultra-inert deactivated inlet liner with a glass wool injection and single taper (Agilent Part No. 5190-3163) was used.

The Agilent 7890N was interfaced with the mass spectrometer (operating in electron ionization, or EI, mode) by a direct transfer line maintained at 220 

C, and the source temperature was set at 230 

 was at 84.1 and 239.2, it was used to detect BPD samples. For this study, 

 84.1 was selected to monitor BPD because it has better response than 

 239.2. To detect the BHGD internal standard samples, 

 69.1 was selected to monitor BHGD because it is an unique ion to BHGD. The mass spectrometer was operated in single-ion-monitoring (SIM) mode and set to scan for the following four ions: 

 69.1, 83.1, 84.1, and 239.2. Because DMSO was used as a diluent in this study, the 

 84.1 peak (which is the most intense peak for both the sample and internal standard) could be used as there was no longer an interfering peak when using this diluent.

Because BBA was only used in the synthesis of API-B, the derivatization procedure was only used with API-B. The BBA in API-B was derivatized into BPD using pinacol by mixing approximately 40 mg of API-B, 200 mg of pinacol, and 120 mg of magnesium sulfate in 2 mL of anhydrous DMSO. The mixture was prepared in a scintillation vial and stirred periodically using a magnetic stirrer. Aliquots were taken between 2.5 and 19 h, centrifuged for 5 min, and then the supernatant was removed. The supernatant was diluted at a 1:1 ratio with anhydrous DMSO and analyzed by GC–MS using instrument parameters described above in “Materials and Methods: Part 2.” A blank sample was run to ensure no carryover. Detailed sample information and preparation details are listed in Table I.

To evaluate the derivatization time required for conversion of BBA to BPD, neat solutions of different BBA concentrations in DMSO were prepared: 10 mg/mL, 10 μg/mL, 1 μg/mL, and 0.1 μg/mL, and these solutions were derivatized with an excess of pinacol. The mass-to-charge ratio (

) 84.1 was used to monitor BPD formation (as it is the strongest ion signal and under these conditions there no interfering signal from the solvent). Table II summarizes the conversion data noting that the optimal time for conversion at 1 μg/mL concentration is dependent on BBA concentration.

Based on the synthetic process, the endogenous BBA was likely negligible. A concentration of 1 μg/mL was tested to show feasibility of the derivatization of BBA in the API-B matrix. It took between 3 and 6 hours for about 80% of the 1 μg/mL BBA to be converted to BPD, which was tested against an authentic BPD sample. For verification of the method limit, the derivatized 1 μg/mL BBA sample was run as a process or system suitability sample prior to analyzing API-B samples.

Part 1: Results Determining the Limit of Detection of BPD With Internal Standard BHGD

The standard ionization technique for GC–MS is electronic ionization (or formerly electron impact), which has the advantage of always fragmenting molecules into the exact fragments every time. This process makes it difficult to detect parent compounds, but by monitoring only the fragments, GC–MS affords a sensitive technique for detecting small amounts of material in a matrix, which in this case, would be any number of active pharmaceutical ingredients (APIs). Scheme 1 shows the compounds investigated and the fragments monitored.

Reproducibility with direct injection with GC has improved with the development of new instrumentation, but since these studies are looking to detect minute amounts of materials, an internal standard (BHGD) was used at a constant concentration of 4 μg/mL to minimize variability in injection.

The sensitivity of using four different detection parameters (FID, MS, and SIM at 

 69 and 239) was evaluated during this study. Because the carrier gas flow of the instrument post column was split to both the MS and the FID (split ~1:10), detection at both detectors was recorded almost simultaneously on the same sample for each data point. Figure 1 shows the concentration range of BPD studied and the resulting total ion count (TIC) and FID signals using BHGD as the internal standard. The resulting area ratios are plotted (BPD/BHGD). Interestingly, the slopes of the detectors are similar; however, the limit of detection for these two detectors using the split column flow was determined to be approximately 0.2 μg/mL. Figure 2 shows the analysis of similar samples, collecting the data by single ion monitoring (SIM) ions at 

 69 or 239, to increase sensitivity. By monitoring only the extracted ions, the limit of detection was 0.09 μg/mL. The slope of the 

 69 fragment curve is shallower than that of the TIC and FID, and the slope of the m/z 239 fragment curve is twice that of the TIC and FID. The fragment curve being this way is because of the response of the internal standard. When the total signal is used (for example in the FID and TIC) the response of the entire internal standard is similar to the entire sample and the slopes are similar. However, when using the SIM, only a portion of the signal is collected. The intensity of the 

 69 signal is stronger in the internal standard than in the sample, making for very low ratios that change only slightly over the concentration range. The intensity of the 

 239 signal in the internal standard is weaker but similar to that of the sample, which generated higher ratios and an overall stronger response at different concentrations, thereby generating a steeper slope. Here, the single ion scan that showed a similar intensity for both the internal standard and the sample, was the ion that gave a more discriminating response. Therefore, when choosing an ion to monitor, not only the intensity of that ion in the analyte needs to be considered, but also that of the internal standard as the optimum ion may not be the most intense one.

FIGURE 1: Area ratio response for FID and MS total ion count (TIC). BPD/BHGD area ratios showing linearity with detection by both MS TIC and flame ionization detector (FID).

 84 in DMSO. BPD/BHGD area ratios showing linearity using both 

Reproducibility with direct injection can be a problem, therefore use of internal standards is common. Reproducibility was tested at multiple concentrations of BPD, while holding the internal standard (BHGD) concentration at approximately 4.0 μg/mL. Reproducibility was tested using MS TIC, FID detectors, and single ion monitoring (SIM) at 

 69 and 239. At the lower levels, the relative standard deviations (RSDs) are less than 10% for the SIMs, which is consistent with typical assay methods for samples at the limit of detection (LOD). At 0.9 μg/mL, the RSDs for the SIMs and the FID are less than 3%. Table III summarizes, the reproducibility of the method at various concentrations with all four detection techniques.

Because these samples were not in an API matrix, recovery was not tested. The GC–MS approach using SIM for detection was shown to be feasible and sensitive. As a next step detection in matrix was investigated. Because of the nature of the matrix, some modifications to the method and derivatization were necessary, such as a change in diluent to one in which the APIs studied are soluble.

It was found that there was little injection-to-injection and day-to-day variability in the signal both from the analyte and the internal standard. However, during the course of this study, there were cases where a series of injections showed an overall higher or lower signal for the analyte than the majority of the samples for that run. In each case, the signal from the internal standard varied in the same manner, which resulted in a constant ratio in line with other samples. It is proposed that eventually the need for the internal standard may be eliminated by having well-maintained equipment and tight system suitability controls. However, for the initial method setup, using the internal standard reduces the possibility of variability from injection to injection, allowing the focus of development to be on the sensitivity of the method.

Part 2a: Results of BPD Injection with Internal Standard BHGD in API-A

For the in-matrix study, the internal standard (BHGD) concentration was set to 3 μg/mL, and the monitoring ion used was 

 was eliminated because of the use of DMSO as the diluent. The linearity of BPD was evaluated from 0.16 μg/mL to 3.2 μg/mL, resulting in a 

 value of 0.9998. Figure 2 shows a typical linearity curve plotting the area response ratios of BPD compared to BHGD. The LOD was found to be 0.16 μg/mL, and the LOQ was 0.48 μg/mL. The RSD of the BPD LOQ of six injections was 4.3%, showing good reproducibility at low levels. Figure 3a shows a typical chromatogram of BPD (with the corresponding internal standard) at the LOD concentration.

FIGURE 3: Counts vs. acquisition time (in min) for (a) Chromatogram of LOD standard using SIM 

 84.1 (BPD concentration 0.15 μg/mL vs. internal standard), and (b) Chromatogram of API-A sample spiked with internal standard BHGD at RT 9.42 min, with no BPD detected at 

To evaluate the recovery of BPD from the matrix at the LOQ level, 0.48 μg/mL of BPD was spiked into API-A. The average recovery was 119%, and the RSD of the three spiked samples was 4.7%. A LOD of BPD at 0.16 μg/mL corresponds to 40 ppm of BPD in API-A.

The following parameters were evaluated: specificity; system precision; accuracy; method precision; linearity; LOQ; and LOD. This method is considered fit for the purpose to determine BPD in API-A.

Seven batches of API-A have been analyzed using this method and no BPD has been detected in any of the batches. A typical chromatogram is shown in Figure 3b (LOD <0.16 μg/mL, corresponding to <40 ppm in API-A).

Part 2b: Results of BBA Detection Using a Derivatization Method

Previous studies have shown that in the absence of a matrix (Part 1) and in a drug substance matrix (Part 2a), and using an internal standard, direct injection, and MS detection, BPD can be detected down to low levels, such as 2.5 ppm.

A second matrix, API-B, was tested for both BPD and BBA. The range of BPD (prior to derivatization of the BBA in the matrix) in API-B was evaluated from 0.0625 μg/mL to 1 μg/mL, yielding a 

 value of 0.9993 using polynomial regression. The LOD of the method is 0.0625 μg/mL, and the LOQ of the method is 0.25 μg/mL. The RSDs of the BPD LOQ of the six injections was 9.8%. The LOD of 0.0625 μg/mL corresponds to 2.5 ppm BPD in API-B.

Using the derivatization method presented to convert BBA to BPD, the amount of BBA in matrix API-B was analyzed. The BBA level of 1 μg/mL corresponding to approximately 32 ppm of BBA in API-B was selected as the limit for the samples, since the acceptable daily intake for API-B is less than 50 ppm, which is the calculated TTC limit. A limit of 0.1 μg/mL of BBA corresponding to 3.2 ppm of BBA in API-B would be needed if the acceptable daily intake of BBA is required to be lower than 5 ppm. This level was shown to be achievable.

This test is considered sufficient to determine BBA to a limit of 32 ppm and BPD to 2.5 ppm in API-B.

This article summarizes the results of several iterations developing a method to determine the levels of several boronic species resulting from Suzuki-Miyaura coupling. Linearity and limits of detection were established for both neat compounds and compounds in matrix. Direct analysis of BPD using GC–MS and an internal standard, BGHD, was successful both in and out of an API matrix. BPD could be detected in matrix down to 2.5 ppm. The boronic compound, BBA, however, required derivatization prior to similar analysis by GC–MS. Detection of derivatized BBA in matrix was successful down to levels less than 32 ppm.

As more scrutiny is placed upon genotoxic impurities, tests for these boronic impurities will become more common. Because derivatization is sometimes required, it would be worth investigating the automation of the derivatization to further streamline these analyses.

The authors would like to thank Jonathon Truong for providing the derivatization conditions.

(3) M.M. Hansen, R.A. Jolly, and R.J. Linder, 

(4) International Conference on Harmonization,

 ICH M7(R1), Assessment and Control of DNA Reactive (Mutagenic) Impurities in Pharmaceuticals to Limit Potential Carcinogenic Risk

(5) F.J. Lopez, E. Gimenez, and F. Hernandez, 

(7) I. Patel, C.J. Venkatramani, A. Stumpf, L. Wigman, and P. Yehl, 

(8) P.J. Pandiyan, R. Appadurai, and S. Ramesh, 

(9) A. Bonet, C. Sole, H. Gulyás, and E. Fernándex, 

(10) C. Pubill-Ulldemolins, A. Bonet, H. Gulyás, C. Bo, and E. Fernándex, 

Chromatographic Method Development: Chapter 12, Considerations for Using GC in the Pharmaceutical Setting

 is with Millennium Pharmaceuticals, Inc., in Cambridge, Massachusetts. 

 is with Ikena Oncology in Boston, Massachusetts. 

 is with Xenon Pharmaceuticals Inc. in Burnaby, British Columbia, Canada. Direct correspondence to: 

Articles

This novel gas chromatography (GC)–MS approach enables analysis of the suspected mutagens bis(pinacolato)diboron (BPD) and tetrahydroxydiboron (BBA) found at low-levels in certain intermediates and active pharmaceutical ingredients (APIs), even in the presence of boron from other sources.

Analysis of Boronic Compounds as Potential Mutagenic Impurities in Drug Substances By GC–MS 

We live in a world with ever-increasing automation, from the simple (hello, automatic coffee maker!) to the complex (I think I’ll just let the car do the driving...or at least the parking). The pervasive, almost intelligent machines have seemingly always been a part of an analytical chemist’s world. Are we ready to let programs and instruments take over the art of separations?

When I joined Bristol Myers Squibb (BMS), I was introduced to some very cool software programs to aid in liquid chromatography (LC) method development, mainly in the achiral reversed-phase LC domain, such as Molnar Institute’s DryLab, S-Matrix’s Fusion, and ACD’s LC Simulator, to name a few. With only a couple of empirical LC runs, these modeling programs could map out predicted separations in silico across a wide design space. With a click of a button, we could potentially let the program “choose the optimum” separation, which we would then quickly discard. Why? Mainly because the programs at the time did not know what we humans really wanted. For example, it was difficult to specify our needs, such as “Only make gradient changes in this region,” or “Allow a step gradient here,” or “Keep the retention factors between 1 and 20,” all while maintaining baseline resolution in an under 5 min runtime. . .with re-equilibration. Usually, we were faster at defining these boundary conditions in our heads, mapping them into the program’s gradient, and seeing what popped out visually, followed by empirical verification. However, these programs are making significant and rapid evolutions, with better optimal design optionality, more dimensionality by increasing from one dimension (for example, gradient), to two (such as add-on temperature), to three (for example, add in pH), more separation modes to be able to model non-reversed-phase separations with the same facility as reversed-phase LC, and back-end analysis of robustness, sometimes with pre-templated reports. Regardless, the human has traditionally always been the conductor of the in silico express train, picking the input packages, keeping them on the rails, and deciding on the final destination. Sometimes wholly iterative method development will arrive at the same place as the modeling, albeit at a slower pace. However, sometimes the modeling program can reveal better-operating spaces, while your iterative work only finds a locally ideal space that is not globally optimal.

At BMS, we have a comprehensive reversed-phase LC system capable of automatically running well over 100 conditions. In these cases, automation results in single-button data acquisition with an overnight run. Although processing can be automated to some degree, at this stage data interpretation can be the primary bottleneck. Some efforts have been made in current commercial chromatography data systems (frequently as an add-on purchase) to facilitate this evaluation, but it is still up to the human to decide what “optimal” actually means. Do we prioritize peak count? Efficiency? Resolution? Tailing factors? Does the answer change if I tell you the sample has 20 components, but only 10 are relevant? The scientist is accountable for picking the right samples to inject, determining the results’ relevancy, and usually defining the ultimate final separation conditions. Also, in the absence of human–data engagement, potential critical situations may be missed, such as a clean sample turning up a new unknown during screening (those pesky peak shoulders!). My fellow scientists and I are extremely grateful that we don’t have to serially churn through setting up all these conditions, giving us the time back to delve deeper into the data or work on other, more rewarding activities.

A more recent LC method automation system available today is feedback optimized screening, such as by ACD’s Autochrom. In this implementation, we let the software choose the best screening condition, but also allow it to do its own final optimization on that condition through an automated modeling exercise. Will this perfect marriage of screening and modeling result in the bench scientists hanging up their lab coats? I hope the previous paragraphs have already convinced you otherwise. Beyond the above arguments, there are just too many separation problems that require serious intervention beyond what current systems can handle. If my automated system relies on mass spectrometry to track peaks, what will it do when I need to do low UV detection? What if the optimal condition is missed due to chelation effects that I could mitigate with a mobile-phase additive? Buffer concentration? Ionic strength? Alternative ion pairs? Uncommon columns? On-column degradation? New technology? Lest we forget, the human is also in the driver’s seat for proper sample preparation (it’s more than just a separation!). We are both blessed and cursed to have so many selectivity levers in LC separations. There’s still plenty of opportunity for artistry by those who study the field. Let automation give you the freedom to explore the many less-common ways of doing your separations. You may just find the next big thing.

 is an Associate Scientific Director at Bristol Myers Squibb. Direct correspondence to: 

Are we ready to let software and instruments take over the art of separations?

Don’t Fear the Automation

The absorption of migrating substances, such as bisphenols, from food contact materials has been associated with adverse health effects. Thus, increasing concern about food contamination and human exposure has necessitated the development of analytical methodologies to monitor the migration of organic substances from food contact materials to packaged food. Following the modern trends of green analytical chemistry (GAC) toward the automation, miniaturization, simplification, and minimization of organic solvents and sample volumes, green microextraction techniques, such as molecular imprinted polymer solid-phase extraction (MISPE), are witnessing rapid growth in the field of food quality and safety. This short review summarizes recent trends in the development of MISPE analytical methodologies applied to foodstuffs for the determination of bisphenols from packaging materials and to ensure the quality and safety of food products.

Bisphenol A (BPA) is widely used as a starting material in the production of epoxy resins, polysulfonates, and polycarbonates. Furthermore, BPA-based polymer materials are widely used for the production of electronic products, medical equipment, thermal paper, bottles, and food containers. BPA leaches from containers to the content, causing several health hazards. Different factors, such as the food type, temperature, pH, contact time, and the thickness of the packaging material, affect the migration amount of BPA from the container to the product during industrial processing and storage. Considering that BPA is a contaminant with endocrine-disrupting properties, and that the general population is exposed to it not only from the environment, but also from food sources, the European Union has set a migration limit of 0.05 mg/kg (1).

The endocrine-disrupting properties of BPA have favored the use of bisphenol analogs in the industry. The most commonly used bisphenol analogs used are bisphenol AF (BPAF), bisphenol AP (BPAP), tetrabromobisphenol A (TBBPA), bisphenol B (BPB), bisphenol C (BPC), bisphenol E (BPE), bisphenol F (BPF), bisphenol S (BPS), and bisphenol Z (BPZ) (Table I). Even though studies have shown that bisphenol analogs have proven less harmful than BPA, they are still under investigation. All bisphenols have revealed estrogenic activity, with BPAF having the highest level and BPS having the lowest. Therefore, it is evident that there is an emerging need to develop sensitive and accurate analytical methodologies for the determination of these compounds (2).

Even though analytical methodologies have progressed throughout the years and made quantification of substances at very low concentrations possible, sample pretreatment still remains the most critical and time-consuming part of any analytical methodology (3). Traditional sample extraction techniques, such as Soxhlet extraction and liquid–liquid extraction (LLE), require the use of large solvent volumes and a substantial amount of time. In accordance with the recent trends of green analytical chemistry (GAC), sample preparation methodologies aim at the simplification and miniaturization of sample preparation steps of the laborious and time-consuming traditional LLE and solid-phase extraction (SPE). Currently, new objectives have been established in sample preparation, including the use of smaller sample sizes and volumes to maximize automation and minimize the aforementioned drawbacks of traditional techniques. In this context, a variety of different approaches toward the development of microextraction techniques have been proposed (4).

The long-lasting search for universal multi-purpose sorbents in SPE has led to the quick development of novel tailor-made sorbents that can be used for selective and efficient sample clean-up and preconcentration of the target analytes. Molecularly imprinted polymers (MIPs) are used in SPE as synthetic alternatives to immunosorbents because of their high selectivity and affinity for the target analytes. The MISPE technique describes the use of SPE with molecularly imprinted polymers, and is considered one of the most advanced preconcentration techniques with respect to the adoption of MIP-based technologies by the scientific community. A MIP is produced by the polymerization of a solution containing a functional monomer, a cross-linker, and template (5). Prior to the polymerization, the functional monomer interacts with the template. After polymerization, the template is removed and the final material contains cavities that can selectively bind to compounds with a similar chemical structure.

MIPs are tailor-made materials that possess specific recognition ability for the target analytes (6,7). In particular, MIPs offer several advantages, such as a moderately inexpensive synthesis, near-infinite self-life, and the ability to be used several times without considerable loss in molecular recognition, while offering ease of production in bulk (8). MIPs are valued for their capability to enrich the analytes, reduce matrix effects, and clean up the extract. The fabrication of MIPs includes the template molecules, one or more functional monomers or precursors, a cross-linking monomer, a porogenic solvent, and a catalyst or a reaction initiator. Their presence in the reaction container in appropriate molar ratio and optimum reaction conditions results in self-assembling of the functional monomers around the template and subsequent cross-linking that yields a solid mass with the entrapped templates in its core.

The successful removal of the template from this solid matrix creates particles possessing nanocavities that are complementary to the size, shape, and functional composition of the template molecules (4). Because of their increasing demand and use, different synthesis approaches have been developed (9).

In most cases, MIPs are used in a cartridge or column format to clean up and enrich the samples prior to chromatographic analysis (10). The general MISPE procedure involves non selective adsorption of the sample onto the cartridge, followed by the selective elution of the analytes of interest in a small volume after choosing the appropriate solvent. Optionally, a washing step can take place between sample loading and elution, with a weak solvent to remove the undesired matrix interferences, without displacing the analyte.

Taking into consideration that foods and beverages are complex matrices owing to the presence of several interfering molecules and that BPA exists in trace levels, MISPE has already proved its excellent selective adsorption ability and applicability to bisphenol analysis, while at the same time increasing the efficiency of the extraction process for the isolation of bisphenols (11).

Monitoring the residual concentrations of bisphenols in foods is the key step to minimize potential hazards. MIPs enable the extraction of bisphenols at trace and ultratrace levels (12).

The typical protocol of MISPE extraction involves using a small amount (15–500 mg) of imprinted polymer painted into cartridges. The MIP is conditioned with the appropriate solvent and then the sample is loaded and remains for a specific holding time. A washing step is necessary to maximize the specific interactions between the target analyte and the imprinted polymer, while at the same time nonspecifically bound interfering compounds will be eluted. After the washing step to remove other bound matrix components, the analytes are eluted with the appropriate solvent (13). A typical MIPSE protocol is illustrated in Figure 1.

FIGURE 1: Typical steps of MISPE extraction.

There are several commercial MIP cartridges dedicated to bisphenol extraction, such as SupelMIP from Sigma-Aldrich and Amfimip from AFFINISEP. BPA, BPB, BPF, BPA diglycidyl ether (BADGE), and bisphenol F diglycidyl ether (BFDGE) were determined by Gallo and coworkers using Amfimip and high performance liquid chromatography (HPLC) with fluorescence detector (FLD), and achieved relatively low limits of detection (LOD). Despite the fact that MIP cartridges are available on the market, several authors have proposed in-house synthesis of MIPs and their application for bisphenol extraction. The most common types of MIPs used for BPA extraction from food samples are presented in Table II.

Dummy template molecularly imprinted polymers (DMIPs) use structurally related analogs as alternatives to a template molecule. DMIPs offer the advantage that a possible template leakage from the polymer will not influence quantification results in complex matrices (14).

Magnetic MIPs (MMIPs) have additional features, such as magnetic properties, compared to conventional MIPs. The incorporation of a magnetic component in a MIP can build a controllable rebinding process, allowing magnetic separation. When MIP particles contain magnetic components, adsorption can be achieved by dispersing them in solution, and then they can be easily separated from the matrix using an external magnet (14). Magnetic separation enables convenient and highly efficient enrichment and evades the requirement to pack the SPE column and the time-consuming process of loading a large-volume sample. Dong and associates (15) established a thermosensitive MIP (T-MIP) on porous carriers for the sensitive extraction of BPA from yogurt and investigated the effects of temperature on the adsorption capacity.

Between different synthesis protocols, the sol-gel technology for molecular imprinting has managed to overcome several shortcomings off the organic synthesis approach, providing a versatile new synthesis approach that can be operated at room temperature and ambient pressure. Sol-gel MIPs offer several advantages over organically synthesized MIPs, such as high imprinting factor, controllable pore size, superior molecular recognition, significantly high surface area, high concentration of imprinted cavities, and superior structural integrity (16). The use of sol-gel MIPs in sample preparation may incorporate several factors to be optimized, depending on the matrix. These include the optimization of the sorbent mass in the cartridge, the selection of the sample volume and the extraction solvent, the holding time and the elution volume, and the extraction flow rate. After the optimization of the extraction, selectivity studies are critical to evaluate the performance of the MIPs. For this reason, a non-imprinted polymer (sol-gel NIP) is synthesized corresponding to sol–gel MIP, but in the absence of the template to evaluate the selectivity of a sol–gel MIP for the target analyte. The same extraction protocol is applied for compounds that have a chemical structure and properties similar to those of the target analyte (17). In the case of BPA, the selectivity of the MIP is evaluated by comparing the imprinting factor of the template (BPA) and its close structural analogs BPB, BPF, BPS, and so on. The imprinting factor (IF) expresses the tendency of the MIP to selectively recognize and bind the template and is calculated by:

 represent the adsorption capacity of the MIP and NIP, respectively.

The selectivity coefficient (SC) presents the selectivity of the MIP sorbent towards the target analyte in comparison to other homologues is calculated by:

Chromatographic analysis is commonly used for the determination of bisphenols. The analytical methodologies that have been applied to the determination of BPA in foodstuffs include high-performance liquid chromatography (HPLC) coupled to UV (18), diode array detection (DAD) (11) or FLD (19), liquid chromatography coupled to mass spectrometry (LC–MS) (20), and gas chromatography coupled to mass spectrometry (GC–MS) (21). LC enables the analysis of the extracts without the extra derivatization step that is necessary in GC. LC analysis is usually carried out using reversed-phase C18 columns with water and methanol or water and acetonitrile as mobile phases, in isocratic or gradient mode (3). UV and FLD detectors have been used for the detection of bisphenols, but MS is the most commonly used, providing very low LODs (3).

The fabrication and use of MIPs in sample preparation as selective sorbent materials has attracted great attention in recent years, leading to the development of analytical protocols that enable the selective extraction of specific analytes. The migration of BPA is an emerging issue and has necessitated the development of rapid, sensitive, selective and efficient microextraction techniques for their determination. MIPs are versatile and highly selective tools that enable the sensitive determination of bisphenols in complex matrices like food. Their inherent advantages include selectivity, sensitivity, robustness, and low cost. MIPs are subject to extensive commercialization in the future. MIP separation efficiency is achieved by the simultaneous realization of improved molecular recognition, enhanced mass transfer kinetics, reduced site heterogeneity, and better access to the site. Even though MISPE is a robust sample preparation technique, improvements concerning the selection of polymerization mixtures, the development of water-compatible sorbents. and industrial scale-up are necessary.

(3) A. Tsalbouris, N.P. Kalogiouri, V.F. Samanidou, 

. (2020). doi:10.2174/1 573411016999200930115626

(6) N.P. Kalogiouri, A. Tsalbouris, A. Kabir, K.G. Furton, and V.F. Samanidou, 

(7) J.O. Mahony, K. Nolan, M.R. Smyth, and B. Mizaikoff, 

J. Chromatogr. B: Anal. Technol. Biomed. Life Sci.

(11) A. Tsalbouris, N.P. Kalogiouri, A. Kabir, K.G. Furton, and V.F. Samanidou, 

(12) A. Kabir, K.G. Furton, and A. Malik, 

(17) V. Samanidou, M. Kehagia, A. Kabir, and K.G. Furton, 

(18) C. Peng Diao, X. Yang, A. Ling Sun, and R. Min Liu, 

(20) N. Jalal, A.R. Surendranath, J.L. Pathak, S. Yu, and C.Y. Chung, 

(21) C. Alexander, L. Davidson, and W. Hayes, 

(22) D.K. Alexiadou, N.C. Maragou, N.S. Thomaidis, G.A. Theodoridis, and M.A. Koupparis, 

(23) E. Herrero-Hernández, R. Carabias-Martínez, and E. Rodríguez-Gonzalo, 

(24) N.P. Kalogiouri, A. Pritsa, A. Kabir, K.G. Furton, and V.F. Samanidou, 

 are with the School of Chemistry at Aristotle University of Thessaloniki, in Thessaloniki, Greece. Direct correspondence to: 

We examine the rapid growth of green microextraction techniques, such as molecular imprinted polymer solid-phase extraction (MISPE), in the field of food quality and safety, and how this technique is used for bisphenol A (BPA) analysis.

The Use of Molecular Imprinted Polymers Prior to Chromatographic Methods for the Analysis of Bisphenols in Packaged Foods

We present a new automated active pharmaceutical ingredient (API) extraction approach with integrated analysis by online liquid chromatography (LC) to assess drug product potency. The platform consists of an extraction module for repeated dynamic pressurized solvent extractions of tablets. Extractions rely on temperature and pressurized solvents (50:50 0.1% TFA:acetonitrile, 100 bar, 3.0 mL/min) to dissolve the tablet coating (if present), disintegrate the tablet, and disperse and solubilize material. Upon exiting the extraction vessel, solvent carrying extracted material is sampled every 15 s using a dual loop LC injection valve. The valve allows direct injection of drug product extract onto a short LC column for online monitoring of the extraction process and generation of an “extract-o-gram.” Autoprocessing of API peaks in the extract-o-gram facilitated direct calculation of tablet potency by integrating the area under the resulting API concentration-time curve. To assess the performance of the platform, a series of experiments was conducted with four pharmaceutical products spanning a wide API dosage range: 200 mg ibuprofen, 81 mg aspirin, and 12 mg and 1 mg tablets of Lilly development materials (Lilly A and Lilly B). Results were encouraging; using a generic set of extraction conditions, reproducible quantitative recoveries were obtained for ibuprofen (99.3% label claim, 0.5% RSD), aspirin (103.2%, 1.1% RSD), Lilly A (98.5%, 1.1% RSD), and coated and core Lilly B tablets (100.8%, 1.3% RSD and 100.9%, 2.4% RSD, respectively).

Prior to release of pharmaceutical products to the market, materials must meet a set of acceptance criteria that ensure product quality and consistency. Common drug product release tests rely on extraction of the active pharmaceutical ingredient (API) from the dosage unit and subsequent liquid chromatography (LC) analysis, including identity, assay, purity, and uniformity of dosage units. Sample preparation and extraction of an API from a solid oral dosage form is a tedious and labor-intensive activity for many projects in pharmaceutical development. In a typical tablet analysis, the sample preparation step serves to isolate the API from the pharmacologically inactive excipients that comprise the bulk of the sample matrix, while also preparing a solution phase sample for analysis. Traditional sample preparation methods involve transferring the tablet into a volumetric flask, adding large volumes of organic solvent, and either mechanical shaking or sonication of the sample to disintegrate the tablet and extract the API (1,2). Following extraction, sample clarification removes insoluble material in preparation for analysis. Although modern LC methods are relatively fast, sample preparation is often the most time-consuming element of the analysis, comprising up to 80% of the total time (3). Therefore, with the continued trend toward shorter pharmaceutical development timelines using fewer resources, there is always a driver for more efficient sample preparation procedures (4).

To facilitate rapid sample preparation for solid oral dosage forms, several non-traditional sample preparation techniques have been developed. Two common techniques involve the use of elevated temperature, elevated pressure, or particle-size-reduction approaches to increase the efficiency of API solubilization (5). The techniques developed to leverage elevated temperature conditions include accelerated solvent extraction (ASE, also known as a 

), and microwave assisted extraction (MAE). Elevated temperature and pressure conditions have also been shown to dramatically improve the recovery of extractions while also reducing extraction time (6). Since its development, ASE techniques have been applied effectively to extract several APIs from crushed and whole tablets for both immediate and sustained release solid oral dosage forms (5, 7–12).

Supercritical fluid extractions (SFE) are another ASE-based approach where a super- critical CO

-based fluid, modified with a small amount (<20%) of organic solvent, extracts API from tablet formulations using equipment similar to ASE. SFE has been used extensively in the research setting with tablet dosage forms to improve the environmental impact of the sample preparation process (13–19). The primary limitation to SFE-based extractions is the poor recoveries obtained with polar pharmaceuticals. For example, in the extraction of pseudoephedrine-HCl from tablets up to 20% (v/v) methanol and an ion-pairing reagent (1% v/v, 1-hepatensulfonic acid), modifiers were required to achieve quantitative recoveries (20–22). Subcritical water extractions (SWE) are a subset of PLEs where only aqueous extraction solvents are used. Richter and associates advanced this approach to extract nifedipine from tablet formulations (23), and the Thurbide group more thoroughly characterized the approach using acetaminophen, vitamin C, and fluoxetine-HCl from tablet and capsule formulations at extraction temperatures up to 225 °C (24,25). Most notably, quantitative recoveries were observed without thermal decomposition of the API at such temperatures. Microwave-assisted extractions (MAE) have also garnered interest as an alternative extraction technique, because of the relative simplicity of the instrument and the highly parallel extraction potential of the systems.

The two most used particle-size-reduction approaches to extraction are ball milling and homogenization. Both techniques act to reduce sample particle size, increasing sample surface area and leading to increased extraction efficiency (5). Ball milling pulverizes samples in the presence of the extraction solvent using stainless steel balls. Following extraction, centrifugation is used to separate insoluble material prior to analysis. Kok and Debets (26) showed that milling and extraction times on the order of minutes could be performed for a variety of high- and low-dose tablets. While relatively fast, ball milling requires several manual manipulations, and is limited in the amount of extraction solvent that can be used. Homogenization uses a set of rotating blades combined with wet grinding, shredding, or shearing to reduce sample particle size and increase surface area for rapid sample disintegration (27). The primary advantage to the homogenization approach for tablet formulations is its current integration into an automated robotic sample preparation system known as the 

 (Sotax) (5,28–30). Shamrock and associates (31) used the automated sample preparation work- flow to transfer a manual sample workflow for roxifiban tablets to an automated whole tablet workflow, where samples were automatically weighed, disintegrated by homogenization, extracted, automatically diluted, and transferred to LC autosampler vials. When paired with effective automation systems, homogenization facilitates rapid sample preparation for solid oral dosage formulations, but the cost and complexity of this instrumentation is a limitation.

In this work, we describe a new automated API extraction platform with integrated analysis by online LC to assess drug product potency for solid oral dosage forms. The platform includes an extraction module for repeated dynamic pressurized liquid extractions of drug product tablets. Extractions rely on a generic set of extraction conditions to dissolve the tablet coating (if present), disintegrate the tablet, and then disperse and solubilize the API. Upon exiting the vessel, extraction solvent travels to a dual loop LC injection valve for online, quantitative monitoring of the extraction process. To assess the performance of the assay, we conducted a series of proof-of-concept experiments with four drug product samples—two over-the-counter products and two Eli Lilly and Company development materials.

Figure 1 shows a diagram of the instrument used to automate tablet potency analysis. A 30 x 10 mm i.d., 2-μm frit porosity, modular high-performance LC (HPLC) column hardware (IDEX H&S) served as the extraction vessel, allowing tablet disintegration and extraction while preventing insoluble material from traveling to downstream fluidic components. Total volume for the extraction vessels was 2.35 mL. Extraction vessels were inside a Waters MV-10 analytical supercritical fluid extraction system oven (Waters Corporation). An Agilent 1260 analytical scale binary pump (G1312B, Agilent Technologies) delivered extraction solvent to each vessel. Extraction fluid temperature was equilibrated at the extraction temperature while passing through a 5 mL sample loop (Rheodyne). Residence time in the preheating coil was 100 s. Detector I was an in-line UV absorbance detector (Agilent 1260, G1314B); the detector was connected directly to the outlet of the extraction oven. To prevent bursting the inline flow cell at extraction pressures above 100 bar, the detector was fitted with a high-pressure flow cell (G1314-60082; 10 mm pathlength; V = 14 μL; Pmax = 400 bar). The outlet of the flow cell was connected to an Agilent 8-port two-position duo valve (P/N 5067-4244) using a 5-foot coil of 0.040” i.d. stainless steel tubing (IDEX). The valve had two 1.2- μL injection loops (L1 and L2) for direct injection of the material exiting the extraction vessels onto the chromatographic column. The column was in an Agilent 1290 thermostated column compartment (G1316). The waste port of the valve was connected to the MV-10 active back pressure regulator (ABPR) using a second 5-foot coil of 0.040” i.d. stainless steel tubing (IDEX). Pump II, an Agilent 1290 qua- ternary pump (G4204A), delivered mobile phase to the separation column. The outlet of the column was connected to Detector II, an Agilent 1290 DAD (G4212B) equipped with a 0.4 μL flow cell (G4212-60032; 3.7 mm pathlength). The instrument was controlled using software from both vendors Agilent ChemStation (OpenLab C.01.07) and Waters ChromScope (Version 1.2).

FIGURE 1: Instrument configuration schematic for automated drug product extractions. The instrument consisted of two independent LC pumps; Pump I performed the dynamic PLE; Pump II was connected to the LC column. Two 10-port column selection valves (S1 and S2) placed inside the extraction oven directed flow to up to nine individual extraction vessels. Port 10 in both valves provided an extraction vessel bypass (purple). An eight-port two-position valve was equipped with dual 1.2 μL sample loops (L1 and L2) to interface the extraction with the online LC. Detectors I and II were both ultraviolet (UV) absorbance detectors.

Aspirin and ibuprofen were from Sigma Aldrich; Lilly’s Corporate Reference Standards Organization provided standard samples for the two Lilly development materials studied—development material A (Lilly A) and development material B (Lilly B). Acetonitrile (LC–MS Optima grade) and premade 0.1% trifluoroacetic acid (TFA) in water were from Fisher Scientific and EMD Millipore, respectively. Method development and calibration standards for each solute were made in 50:50 acetonitrile:0.1% TFA in water.

Drug product extractions utilized a dynamic extraction mode with a genericized set of conditions for all proof-of-concept work. Extraction fluid was pumped through the extraction vessels for 25 min at 3.0 mL/min. Extraction solvent consisted of 50:50 acetonitrile:0.1% TFA in water; extraction oven temperature was 30 °C. The ABPR pressure setting was 100 bar.

All separations used an Agilent Zorbax SB-C18 (50 mm x 3.0 mm i.d., 1.8 μm dp) column and an acetonitrile:0.1% TFA in water mobile phase. Acetonitrile composition for each solute was set to achieve a retention time of about 12 s. Acetonitrile compositions for ibuprofen, aspirin, Lilly A, and Lilly B were 0.65, 0.25, 0.275, and 0.30, respectively. Injection-to-injection cycle time, (valve switching time or chromatographic time) was 15 s. The extract-o-gram run time was 25 min. In total, 100 LC injections were performed for each extraction experiment. Injection volume, column temperature and mobile phase flow rate for all experiments were 1.2 μL, 85 °C, and 2.5 mL/min, respectively; Interstitial linear velocity was about 1.0 cm/s. Detection at 80 Hz used absorbance at 220, 254, and 280 nm. Quantitation used the wavelength with the best signal for the specific sample. Peaks were autointegrated in the chromatographic software, and quantified by response factor from a single point calibration. Standard concentration was selected to fall within the linear range of the detector, and be relevant to the tablet dosage for each product evaluated.

Automation of Drug Product Potency Workflow by Fast, Online LC 

Figure 1 shows a block diagram for the instrument designed to automate the tablet potency assay for up to nine replicate drug product samples. Briefly, the platform consisted of an extraction module for repeated PLE of tablets in 2.35 mL extraction vessels. Extractions utilized temperature and pressurized acetonitrile:water mixtures to dissolve the tablet coating (if present), disintegrate the tablet, and disperse and solubilize material during the 25 min extraction time. Extraction solvent composition was selected to have good solubility for the APIs tested. Online LC monitored the dynamic extraction process using the dual loop injection valve with high temporal resolution sampling, 15 s sampling cycle, and chromatographic analysis. High frequency sampling and analysis were key to quantitative performance.

Figure 2 shows representative data from an extraction experiment, highlighting platform performance and utilization of online LC for quantitation. Drug product content determination began by manually loading the extraction vessels with the desired sample and placing extraction vessels in the extraction oven. Each PLE was initiated when, as shown in Figure 1, S1 and S2 valves rotated to place the extraction vessel into the flow path of the extraction solvent. Online LC analysis start time was aligned with this extraction start time beginning the repeated actuation of the dual loop injection valve every 15 s. Figure 2a shows the pressure profile obtained at the extraction pump (Pump 1). Each PLE can be broken down into three distinct phases based on this profile, a filling phase, shown in red, where the empty extraction vessel filled with solvent and was pressurized by the ABPR to the desired extraction pressure (100 bar); note that the instrument contributed approximately 50 bar to the extraction pump, accounting for the offset in pressure readings between the extraction pump and ABPR setpoint. Following pressurization, the extraction entered a disintegration phase, shown in blue, where the tablet coating was penetrated by the solvent breaking up the tablet, and the solvent solubilized the API. Completion of the disintegration phase was determined by detection of the first API peak in the online LC chromatogram of Figure 2b, 3.75 minutes into the extract-o-gram. The first several minutes of the extract-o-gram correspond to the filling and disintegration phases. During these phases, the two-position valve switched every 15 s injecting extraction solvent onto the online LC column. Each 15 s cycle represents a single chromatographic analysis; conditions were optimized so API peaks were eluted with a retention time of ~12 s to enable the subsequent injection. Following the disintegration phase, the extraction entered the steady-state phase, shown in green in Figure 2a. During the steady-state phase, API concentration in the extraction solvent increased, and a commensurate increase in API peak size (red arrows in Figure 2b and profile in 2c) in the chromatograms of the extract-o-gram was observed. Figure 2c shows the maximum concentration of API in the extraction solvent was observed at 8 min, and then fell off as the extraction completed its quantitative removal of API from the drug product matrix.

FIGURE 2: Example experimental data stream used in the automated tablet potency workflow. (a) Pump I pressure profile collected during the extraction. During the filling phase, pump I filled the vessel with solvent, and in combination with the ABPR, pressurized the vessel to the desired extraction pressure. Following pressurization, the vessel entered the disintegration phase where extraction pressure stabilized, and solvent penetrated the tablet coating (if present). The steady-state phase represents the time where the vessel maintains set pressure and temperature for the duration the extraction; (b) Chromatograms observed at detector II during the initial five minutes of the drug product extraction. Red and blue dashed lines indicate sampling valve actuations, that is, every 15 s; color highlights the dual injection loop-sampling scheme used; (c) Representative 25-min extract-o-gram generated when using online LC to monitor the extraction of a drug product with 15 s time resolution. Elution conditions were optimized for each API peak retention time to be ~12 s.

Auto-integration and comparison to the external standard of the 100 individual online LC injections collected during the extraction resulted in an API concentration value, Ci, for each sample in the extract-o-gram. Equation 1 was used to determine the mass of API, 

, recovered from each drug product extraction:

 is the sample concentration in each online LC chromatogram, 

 is the online LC sampling time (15 s). Multiplication of the sample concentration 

 (3.0 mL/min) resulted in a value for API in mass per time for each LC injection. Multiplication of the mass per time in each online LC chromatogram by the 15 s sampling time resulted in the mass of API, 

, in each extract-o-gram chromatogram. Integration of the mass values during the entirety of the extraction yielded a total mass of API recovered during the extraction.

Performance of the extraction platform was assessed using proof-of-concept extractions with two over-the-counter (OTC) materials. Figure 2c shows a representative extraction profile for one of the two OTC materials tested: 200 mg ibuprofen immediate release tablets and 81 mg delayed release aspirin tablets. Results for both OTC materials were encouraging. Table I shows the total mass recovered, and percent label claim for eight ibuprofen and six aspirin extractions. For ibuprofen, an average of 198.6 mg or 99.3% of label claim was recovered, with a 0.5% RSD and 83.6 mg or 103.2% of label claim for aspirin was recovered with a 1.1% RSD. Recovery and reproducibility were quite good for both materials, conforming to the USP requirements for drug product potency (90–110% of labeled claim on content, 

Application to Lilly Development Materials

Figure 3 shows representative extract-o-grams for two Lilly development materials, Lilly A, and Lilly B (coated and uncoated tablets), and the quantitative results from the three series of six replicate extractions reported as percent label claim. Figure 3a shows an example extract-o-gram for 12 mg coated immediate release tablets of Lilly A. The inset chromatogram was pulled from 9.75–10.00 min into the extraction, a time corresponding to that required to extract greater than 95% of Lilly A from the tablet. An average of 98.5% of label claim material was recovered from Lilly A tablets with an RSD of 1.2%. Figures 3b and 3c show example extract-o-grams for low dose 1 mg uncoated tablet cores and 1 mg coated tablets of Lilly B. Extraction results for the two low dose materials were the same, with the average API mass recovered from both materials being 1.01 mg with RSD of 1.3% and 2.4% for the core and coated tablets, respectively. A qualitative comparison of the core and coated extract-o-grams showed similar extraction profiles for the two samples. Close inspection highlights that the concentration of Lilly B for the tablet core extraction is marginally higher than the coated tablets early in the extraction, as the tablet coating present must be penetrated by the solvent prior to observation of Lilly B in the extraction solvent.

FIGURE 3: Representative extraction profiles for three development materials evaluated. (a) Extraction profile for a 12 mg tablet of Lilly A. The inset chromatogram was from 9.75-10.00 minutes into the extraction, corresponding to the time required to extract greater than 95% of the API from the tablet matrix; (b) Example profile from the extraction of a 1 mg uncoated tablet core of Lilly B; (c) Extraction profile from a coated 1 mg tablet of Lilly B; (d) Quantitative extraction results from Lilly A and Lilly B as a percent recovery based off of label claim.

A new automated API extraction platform with integrated analysis by online liquid chromatography was developed to determine drug product potency. Using a generic set of extraction conditions, proof-of-concept experiments were performed on four different drug products: 200 mg ibuprofen, 83 mg aspirin tablets, and two Lilly development materials. Quantitative recoveries of 98.5% to 103.2% were obtained for tablets containing 1 to 200 mg API. Reproducibility was also quite good, with extraction RSD for all materials evaluated being less than 2.5%.

The authors declare no competing financial interest.

Funding for this work was provided by a Lilly Research Labs Process Research and Development Post-Doctoral Fellowship. We would like to thank Waters Corporation for the loan of instrumentation.

Sample Preparation of Pharmaceutical Dosage Forms

 (Springer, New York, New York, 2011). https://doi.org/10.1007/978-1-4419-9631-2.

, 123–144 (2005). https://doi.org/10.1016/S0149-6395(05)80049-2.

(4), 232–244 (2003). https://doi.org/10.1016/S0165-9936(03)00402-3.

(4) A.M. Pimenta, M.C.B.S.M. Montenegro, A.N. Araújo, and J.M. Calatayud, 

(1), 16–34 (2006). https://doi.org/10.1016/j.jpba.2005.10.006.

(5) B. Nickerson, W.B. Arikpo, M.R. Berry, V.J. Bobin, T.L. Houck, H.L. Mansour, and J. Warzeka, 

(2), 268–278 (2008). https://doi.org/10.1016/j.jpba.2008.01.006.

(6) B.E. Richter, B.A. Jones, J.L. Ezzell, N.L. Porter, N. Avdalovic, and C. Pohl, 

(6), 1033–1039 (1996). https://doi.org/10.1021/ac9508199.

(7) A. Blanchard, C. Lee, B. Nickerson, L.L. Lohr, A.J. Jensen, K.M. Alsante, T.R. Sharp, D.P. Santafianos, R. Morris, and K.D. Snyder, 

(2), 265–275 (2004). https://doi.org/10.1016/j.jpba.2004.05.012.

(8) A.M. Abend, L. Chung, D.G. McCol- lum, and W.P. Wuelfing, 

(6), 1177–1183 (2003). https://doi.org/10.1016/S0731-7085(03)00020-7.

(9) E. Björklund, M. Järemo, L. Mathiasson, L. Karlsson, J.T. Strode, J. Eriksson, and A. Torstensson, 

(4), 535–549 (1998). https://doi.org/10.1080/10826079808001238.

(10) T.H. Hoang, R. Farkas, C. Wells, S. McClintock, and M. Di Maso, 

(1), 257–261 (2002). https://doi.org/10.1016/S0021-9673(02)00956-1.

(11) C. Lee, J. Gallo, W. Arikpo, and V. Bobin, 

(4), 565–571 (2007). https://doi.org/10.1002/jps.2600831104.

(4), 1509–1516 (2015). https://doi.org/10.1039/C4AY02876G.

(13) A.L. Howard, M.C. Shah, D.P. Ip, M.A. Brooks, J. Thompson Strode, and L.T. Taylor, 

(11), 1537–1542 (1994). https://doi.org/10.1002/jps.2600831104.

(14) B.R. Simmons, O. Chukwumerije, and J.T. Stewart, 

(3), 395-403 (1997). https://doi.org/10.1016/S0731-7085(97)00093-9.

(15) J.K. Lawrence, A.K. Larsen, and I.R. Tebbett, 

(1), 123–130 (1994). https://doi.org/10.1016/0003-2670(94)85121-2.

(16) N. Bahramifar, Y. Yamini, and M. Shamsipur, 

(3), 205–211 (2005). https://doi.org/10.1016/j.supflu.2005.01.005.

(7), 747–751 (1993). https://doi.org/10.1039/AN9931800747.

(18) S. Scalia, G. Ruberto, and F. Bonina, 

Here, a new automated active pharmaceutical ingredient (API) extraction platform is integrated with online liquid chromatography (LC) to assess drug product potency. The platform dissolves the tablet coating (if present), disintegrates the tablet, and disperses and solubilizes the tablet for analysis.