Microplastics are now recognized as pervasive in the environment, as well as in stormwater, wastewater, ocean water, coastal sediments, drinking water, and even animal tissue. The mismanagement of plastic waste throughout the world is a matter of urgent concern that has led to product bans for small plastics, such as microbeads used in cosmetic products, to larger plastic items, such as bags and straws, that can degrade into microplastics. This concern has also resulted in new and planned requirements to monitor microplastics in the environment. The implementation of monitoring programmes requires reliable standardized methods and best practice guidelines, yet environmental measurement and monitoring practitioners still do not have established practices and methods for collecting and analyzing plastic pollution. This deficiency has resulted in significant challenges in the interpretation of research findings for use in corporate or government policy. A major issue with measurement based upon sampling and analysis of microplastics is their unique behaviour; environmental microplastics occur as non-homogeneously distributed particulates and fibres composed of various polymeric materials. This article provides an overview of the environmental and analytical challenges caused by microplastics; it proposes an analytical measurement approach that enables accurate qualitative and quantitative analysis of microplastics, and demonstrates some of the advantages of pyrolysis–gas chromatography–mass spectrometry (Py–GC–MS) for the analysis of environmental samples as compared to the other leading analytical methods.

Plastic waste has proven very difficult to measure. In 2019, 368 million tons of plastic were produced (1), an increase of more than 600% over the last four decades (2) (Figure 1). Global collection rates of plastic waste over the same timeframe remained stagnant, not surpassing 10% of production worldwide (3). Collection and sorting of plastic waste into recovered material is not a guarantee that it will be recycled. This low recycling rate has been exacerbated by implementation of China’s National Sword Policy on 1 January 2018, with its rejection of imported recovered plastic waste from Western countries.

Examination of the sources of these problems can also be misleading. For example, a leading economic model uses surrogate factors for plastic waste mismanagement and depicts non-Western countries as being primarily responsible for the inputs of plastic pollution into the environment (4). These models miss the point that countries like the United States have high per capita plastic consumption with sporadic and inconsistent waste management infrastructure, often exporting large portions of recovered plastic waste to non-Western countries (5). Final disposition of plastic waste exports from Western countries is now a bigger unknown than it was prior to 2018, with likely more plastic ending up in landfills, burned, or accumulating in outdoor heaps and piles and migrating to watersheds, creeks, streams, rivers, and the ocean during wind and rain events. The problem of plastic pollution does not belong to “other” countries but is very much one of every nation. As plastic debris enters roadways and watersheds, it threatens our natural environment (Figure 2), pollutes and clogs stormwater conveyances, and can pose significant sanitation and public health issues, such as an increase in standing water and mosquito vector growth. As debris enters streams and rivers, micro- and nanoparticles are generated at rates dependent on the forces acting upon them and their polymer type and plastic form. We anticipate that microplastic monitoring will become more prevalent as a means for measuring plastic pollution in the environment.

Plastic debris in the environment is not inert. Plastic debris ultimately fragments into smaller particles over time as a result of the effects of weathering, including UV radiation and wind and wave action. As debris breaks down into smaller and smaller particles, polymer surface areas increase relative to their volume, and these surfaces can host pathogens, such as bacteria and viruses. These smaller plastic particles, with their relatively higher surface area, have a greater propensity to sorb persistent organic pollutants (POPs) such as polychlorinated biphenyls (PCBs) and pesticides from surrounding air, dust, and water.

Additionally, the manufacturing process of plastic products typically adds chemical constituents, or plasticizers, to augment polymer characteristics and performance, and some of these additives are toxins or carcinogens. It is currently unknown to what extent plasticizers from within microplastics and any sorbed contaminants on the surface of particles are able to desorb and, upon ingestion by an organism, be taken up by the organism’s tissue and organs (Figure 3). These contaminated microplastics form dust pools both indoors and outdoors. Plastic particles have been documented in every environmental media, and reports in 2019 and 2020 present evidence that plastic is even raining down on us. This pervasiveness and persistence of microplastics and nanoplastics cause some to speculate that scientists of the future will use these plastics as a tool for dating abandoned habitations and geologic strata.

The behaviour of microplastics in environmental media is expected to be somewhat more predictive than that of larger debris, which is subject to chaotic ebbs and flows of wind and water. Although plastic fragments may initially float on the water’s surface, they soon become suspended in the water column as they become covered in biological growth; biofouling increases the specific gravity, which decreases buoyancy. It has been assumed that, ultimately, most waterborne microplastics are destined for the floor of an ocean, lake, or river, perhaps being ingested along the way or resuspended through various mechanisms on their way to, and once on, the bottom. However, colloidal suspension and electrostatic interaction between the smallest of these plastic particles and the resulting effect on transport and occurrence in the environment is an area of study that has yet to be explored.

Until recently, calculations of density and distribution of plastic in the environment focused almost exclusively on the visible size range above 300 µm, and employed visual techniques as a means of identification. In addition, these samples were collected on the ocean surface. The majority of microplastics in the environment, both water and air, are now believed to be smaller than 300 µm.

You Can’t Manage What You Can’t Measure

Rigorous and reproducible identification of plastic particles is critical to pollution monitoring. Considering the mantra “you can’t manage what you can’t measure”, effective upstream policy, including corporate and government action to reduce the occurrence of plastic particles, must rely on robust monitoring regimens. Efficacy of mitigation actions will not be measurable without such monitoring, and measurement standards therefore must be agreed on and widely adopted.

The risk of microplastics to our natural environment, and to public health, can be described as the likelihood of harm based on both toxicity and exposure, yet this understanding of risk has been stymied by the lack of rigour, standardization, and reproducibility of measurement practices and methods. While spectroscopy techniques measure plastic particle count, the properties of smaller plastic particles create a sort of fingerprint more easily traced with mass-based concentrations. Mass-based concentrations are much more conducive to benchmark comparisons such as ambient conditions or regulatory thresholds. Measuring concentrations of these tiny plastic particles and their effects on humans and animals necessitates the global standardization of collection, preparation, and identification procedures and methods.

Standardized methodologies for nondestructive collection, handling, separation, sample preparation, and positive identification of microplastics are in the early stages of development. ASTM International is leading the way with recently published standards for collection (D8332) and preparation (D8333) of microplastic particles. Some of the leading technologies currently available that yield scientifically defensible and reproducible results for positive microplastic identification include focal plane array detection, thermogravimetric analysis–pyrolysis–gas chromatography–mass spectrometry (Py–GC–MS), and infrared (IR) and Raman spectroscopy.

Presently, most researchers attempting to positively identify polymers are employing IR or Raman spectroscopy for analysis of particle type and count. Information on particle polymer type, count (per unit volume), and size are germane to specific research contexts, and some advanced spectroscopy technologies have reduced operational procedures from upwards of a week to only hours; in most instances, spectroscopy remains time-consuming and costly, and requires highly trained and skilled technicians with sophisticated laboratory equipment. Although a high degree of skill is also needed for Py–GC–MS, this technique offers a solution to speed up sample preparation and analysis time, while yielding perhaps more practical information on polymer types and mass per volume.

For water quality monitoring activities such as discharge management, plastic particle identification by mass per unit volume may be more useful for both point source and non-point source measurement. We believe that the use of Py–GC–MS for polymer identification has broader applicability than IR or Raman spectroscopy because it is less constrained by particle size (large or small), requires less preparation, has fewer interferences, has less potential for analytical error, enables straightforward calibration, provides consistent and highly reproducible results, much lower equipment entry costs, and readily available polymer identification libraries. It is also conducive to varied matrices, such as water, soil, and biota, enabling more easily accomplished comparisons.

Here, we outline the development of the Py–GC–MS method, and explain its practical application to understanding the occurrence and prevalence of microplastics, and potentially even nanoplastics, in the environment.

Recent studies have discovered the unique capability of the microfurnace pyrolysis–GC–MS technique in the identification of polymer types and quantities in microplastic particles. This article describes the study of microplastics in water samples with high and low suspended solids using a microfurnace pyrolyzer directly interfaced with a benchtop GC–MS instrument.

The samples were collected from drinking water, wastewater influent and effluent, surface water, and marine water. The developed method using the microfurnace Py–GC–MS confirms that the analysis of microplastics in such water samples is independent of polymer type, and not limited to any particular water matrices. The method proves higher analytical sensitivity for both qualitative and quantitative analyses, including better detection limits. The inorganic contaminants in the sample matrices do not impede the identification and specification of the polymers and associated additives. Detailed information about the chemical nature of the polymer and contained organic additives are also obtained simultaneously.

Most of the initial research on analyzing microplastics was developed using spectroscopy techniques such as Fourier-transform infrared (FT-IR) and Raman spectroscopy. These traditional techniques are nondestructive and are commonly used in analytical laboratories. They provide information for functional group identification and library comparison. Detailed information about the contained organic additives is usually not obtained because contaminants (organic or inorganic) and additives can overlap polymer bands. In other words, FT-IR can suffer from peak masking if two or more compounds in a mixture with the same dipole moments absorb IR rays at the same or nearby wavelength. In this case, identifying the overlaid peaks will be extremely difficult. With Raman spectroscopy, measurement of samples that contain fluorescent dyes is very challenging.

The need for sample preparation, such as matrix isolation, solvent extraction, or pretreatment, is another challenge with these traditional techniques. Typically, it could take several hours to perform the analysis, and quantitation is challenging. The limit on sample particle size with these techniques must also be considered. Micro-Raman spectroscopy and an optical microscope are used for particle size analysis that is larger than 1 µm, whereas FT-IR can be used for particle size larger than 10 µm.

In contrast, microfurnace Py–GC–MS can be used for both qualitative and quantitative analysis of low microplastics concentrations because of its high sensitivity capability. Although there is no practical lower limit in particle size, total polymer mass in a sample constrains the lower limit of the measurement (see Figure 4). In addition to microplastics, GC–MS adds the benefit of detecting and quantifying both additives and contaminants in the sample using interpretation libraries. Sample preparation is easy and straightforward; micrograms of a sample (including solids taken “as is”) are placed into an inert pyrolysis cup, and the analysis can be performed. The recent developments in F-Search software for microplastics data processing, which is an automated analysis (quantitative and qualitative), has also significantly shortened the time needed for the analysis.

The microfurnace utilizes a low thermal mass ceramic heater to quickly heat and cool with thermal uniformity. This technology can be used for multiple analytical techniques in addition to flash pyrolysis, such as evolved gas analysis (EGA), thermal desorption (TD), double‑shot (TD followed by flash pyrolysis), and heart-cutting. All the surfaces are quartz or temperature stable deactivated, and there is no cross-contamination.

The microfurnace performs continuous pyrolysis in a single step by directly depositing all the pyrolyzates on column as they are generated. The direct connection of the microfurnace to the GC injection port allows this single-step pyrolysis where reproducibility and temperature accuracy are obtained. The sample is held near ambient temperature and in an inert atmosphere prior to pyrolysis. Therefore, there is no chance for evaporation or degradation of the sample before the analysis, maintaining sample integrity and contributing to accuracy, precision, and reproducibility of the data set. The microfurnace is preheated to the desired temperature and precisely measured with a thermal couple sensor. The sample cup then freefalls into the quartz pyrolysis tube, where the sample is rapidly and reproducibly pyrolyzed. The pyrolyzates are directly swept onto the GC column for separation and detection by MS. A critical factor with the microfurnace is its ability to control the actual sample temperature to within ±0.1 °C set point of the pyrolysis temperature. This feature guarantees reproducibility and precision, as well as true flash pyrolysis, where the set temperature in the furnace by the analyst reflects the pyrolysis temperature on the sample.

There are other pyrolysis technologies: filament, which uses platinum coil wire, and Curie-point, which utilizes a ferromagnetic foil. These technologies can suffer from poor temperature accuracy and poor reproducibility because of non-uniform heating. In both systems, to avoid condensation of the pyrolyzates, the sample must be inserted into a heated manifold before pyrolysis. This preheating step can cause degradation or thermosetting of the samples prior to analysis. With filament pyrolyzers, sample placement and heat-transfer variations also affect the temperature of the sample, depending on where the sample sits in relation to platinum wire coils and on sample contact. The filament temperature is estimated from the electrical properties of the wire, but the temperature of the sample is not actually measured as it changes over time. In other words, the platinum wire can only control the temperature of its heating element, not the actual sample pyrolysis temperature, requiring the user to program the system at a much higher temperature than the desired true pyrolysis temperature on the sample.

In a study of microplastics in surface water using a filament pyrolyzer, the pyrolysis temperature was set at 750 °C, with the sample placed in a narrow quartz tube inside the platinum wire (6). Based on actual bond energy strengths, polymers and other organic material will typically pyrolyze prior to a temperature of approximately 700 °C. Therefore, the only reason to set an initial pyrolysis temperature in excess of 700 °C is that the actual sample temperature in the pyrolyzer is not accurate. Literature indicating pyrolysis at 1200 or 1300 °C is either unnecessary, over pyrolysis, or shows poor actual sample temperature control in the particular pyrolysis device.

It is critical to note that some filament pyrolyzers perform two-step pyrolysis. First, some of the pyrolyzates are collected on a trap, and then a switching valve sends the pyrolyzates from the trap through a long transfer line into the GC column. In this configuration, the pyrolyzates are indirectly placed on column. During this two-step process, polar and heavy compounds may cause cross-contamination, or become irreversibly stuck in the trap or the long transfer line, and are not transferred for analysis because of the potential of cold spots in the system.

Figure 5 shows the system configuration of the microfurnace pyrolysis–GC–MS used for microplastic analysis in the water samples. Experiments were performed with a microfurnace pyrolyzer (EGA/PY-3030D, Frontier Laboratories Ltd.) with Auto-Shot Sampler (AS-1020E, Frontier Laboratories Ltd.) coupled to a GC–MS system (5977B MDS, Agilent Technologies, or GCMS QP-2020NX, Shimadzu Corporation). Chromatographic separation was achieved using a metal capillary column (UA-5, 0.5 μm film of 5% diphenyl-95% dimethylpolysiloxane, 30 m × 0.25 mm, Frontier Laboratories Ltd.) with helium (He) as the carrier gas (1 mL/min). Reduced analysis time can be obtained by utilizing a backflush system, especially when dealing with high-boiling point matrix interference or high-molecular-weight pyrolysis products. This system consists of a precolumn (UA-50, 1 μm film 50% diphenyl 50% dimethylpolysiloxane, 2 m × 0.25 mm, Frontier Laboratories, Ltd.) connecting the injector and the separation column with a three-way split (7).

Using the flash pyrolysis mode of operation (the single-shot mode) of the microfurnace, the chemical composition analysis was performed on microplastic samples. The microfurnace was programmed at 600 °C to break down the organic bonds of the samples into smaller and more stable molecules, referred to as 

. The pyrolyzates are directly swept from the pyrolysis tube into a GC column, via a needle fixed in a split injection port. The column was interfaced directly to an MS instrument. The chemical species were ionized and sorted, based on their mass-to-charge ratio. Compounds eluted from the GC column were identified by comparing the mass spectra at a predetermined retention time to reference spectra and retention index (RI) to those contained in a database.

Table 1 shows some of the polymers that can be identified using microfurnace Py–GC–MS. The validation is performed using a pyrolyzate primary quantitative ion and identification ion(s) representative of a specific polymer. It is important to note that the developed method is not limited to the polymers listed in the table.

Figure 6 demonstrates a typical pyrogram of the 11 target pyrolyzates and their associated polymer (in parenthesis), using the developed method described above.

The concentration of each identified polymer is determined using either internal or external standard calibration. Using a multi-component standard (8), the calibration curves can be obtained. Table 1 lists the data precision of each polymer for 

As mentioned earlier, F-Search, an automated search program, has been augmented for the fast identification and quantitation (in the range of seconds) of microplastics. The search algorithm is based on the identification of the selected number of characteristic pyrolyzates of each polymer in the sample. Each selected pyrolyzate is only obtained from the pyrolysis of a polymer, and not from the pyrolysis of all other investigated polymers. For each characteristic pyrolyzate, a set of specific 

 signal intensities is also selected. A normalized summated mass spectra (SMS) is obtained from the selected 

 signal intensities so that the mass spectra are cleared from any interfering signals due to coeluted species. The similarity of the obtained SMS with the one in the library is calculated and expressed as a percentage. A signal-to-noise ratio (S/N) is calculated for each polymer by extracting a characteristic 

 signal chosen from one of the main pyrolyzates. If the S/N is lower than a set threshold, the presence of the corresponding polymer is excluded, regardless of the similarity index obtained by the algorithm.

Figure 7 demonstrates an example of the identification process of polypropylene (PP) in the reference mixture using the automated F-Search program. The ions at 

 41, 43, 55, and 70 were selected as the representative four pyrolyzates of PP. The peaks of each characteristic pyrolyzate identified in the extracted ion chromatograms (EICs) are marked with dots. The SMS of these peaks was then compared with that obtained for reference PP. Because a similarity of 98.4% was obtained, the presence of PP in the reference mixture was confirmed. The results obtained with the algorithm for the two samples are summarized in Table 2 (5), along with the calculated S/N values and the limit of detection of the algorithm. Both a high match and S/N greater than the set threshold were obtained for all polymers in the reference mixture.

The microfurnace Py–GC–MS technique provides detailed information about polymers, additives, and even contaminants. The identification of polymers is rapid and automated, easy to use in different laboratory setups, and less costly compared to traditional spectroscopy techniques. Neither the presence of biofouling nor that of contaminants or additives impedes the identification and quantitation of polymer types.

The sample preparation is easy and straightforward; the sample is placed into an inert sample cup and held in the ambient atmosphere before the analysis starts. Once the furnace achieves the desired temperature as programmed by the user, the sample cup freefalls into the microfurnace, where rapid and reproducible pyrolysis happens. All the generated pyrolyzates are directly placed on column in a single step, foregoing the need for a transfer line or chemical trap.

The F-Search automated program significantly increases analytical efficiency by providing polymer identification in a matter of seconds. Microplastics are often a mixture of polymers and additives, and in most cases, they are coeluted, so identification using spectroscopy techniques is challenging. However, the coelution of compounds and contamination in the microplastic mixture samples does not affect the ability of microfurnace Py–GC–MS to provide simultaneous identification and specification of polymers and additives.

Another advantage of this technique is the elimination of solvent extraction or sample pretreatment before analysis. The microfurnace can be used for thermal desorption analysis where the additives and light compounds are thermally extracted from a mixture. Laboratories save money over time by using fewer solvents, while also protecting analysts and the environment from solvent exposure and hazards.

The microfurnace produces reproducible data, and controls the actual sample temperature to within ±0.1 °C of the pyrolysis temperature. Other types of pyrolyzer devices can only control the temperature of the heating element, and not the actual sample temperature. Also, the ability of the microfurnace to protect the sample from any heat prior to the analysis and the absence of the transfer line makes it possible to analyze a wide range of compounds, including additives, volatiles, light, heavy, polymeric, and polar compounds.

The collection approach and sample number must always be considered in the analysis of microplastics. Often, water samples are filtered to collect the microplastic mixture. With Py–GC–MS, only a small amount of sample (microgram quantities) is needed, and multiple analyses can be performed with one-time sampling. This new Py–GC–MS method is currently under development as an official ASTM method by an ASTM International workgroup.

A concerted effort must be made to standardize and make widely available these types of sampling and analytical methodologies, as well as streamline methodologies and reduce their costs. As the demand for microplastics measurement and monitoring increases, practitioners must have access to reliable and repeatable methods. Ideally, these methods would provide both qualitative and quantitative information and be comparable to other standardized and commonly used environmental measurement techniques.

The ability to adapt these technologies for field deployment is essential to gain a better understanding of the quantity, distribution, and behaviour of plastic particles in all environmental matrices. Common reporting of microplastic concentration data will support both government and industry responses to the growing crisis of plastic pollution. As described, microplastic measurement provides advantages over traditional debris monitoring. Mass-based analysis is directly comparable to other environmental contaminant analysis methods, and has an advantage over spectroscopy techniques because it avoids complicated preparation and the need to count individual particles after identification.

New frontiers in microplastic measurement and monitoring require robust and repeatable analytical techniques. Analytical information should be the cornerstone of corporate and government pollution mitigation efforts, and Py–GC–MS will be critical in the evaluation of efforts to curb plastic pollution around the world.

The authors wish to thank the scientists at Frontier Laboratories whose work made this article possible: T. Ishimura, I. Iwai, M. Matsueda, A. Shiono, K. Matsui, N. Teramane, and C. Watanabe.

The views expressed in this article are those of the authors and do not necessarily reflect the views or policies of the U.S. Environmental Protection Agency.

https://www.darrinqualman.com/global-plastics-production/

https://committee.iso.org/files/live/sites/tc61/files/The%20Plastic%20Industry%20Berlin%20Aug%202016%20-%20Copy.pdf

https://www.draper.com/sites/default/files/2018-11/EPAMicroplasticsFactSheetOctober2018.pdf

J.R. Jambeck, A. Andrady, R. Geyer, R. Narayan, M. Perryman, T. Siegler, C. Wilcox, and K. Lavender Law, 

, 768–771 (2015). https://science.sciencemag.org/content/347/6223/768.full

B. Ravit, K. Cooper, G. Moreno, B. Buckley, I. Yang, A. Deshpande, S. Meola, D. Jones, and A. Hsieh, 

(6), 809–826 (2017). https://www.aimspress.com/article/10.3934/environsci.2017.6.809

K. Matsui, T. Ishimura, M. Mattonai, I. Iwai, A. Watanabe, N. Teramae, H. Ohtani, and C. Watanabe, 

, 104834 (2020). https://www.sciencedirect.com/science/article/abs/pii/S0165237020300899

M. Matsueda, M. Mattonai, I. Iwai, A. Watanabe, N. Teramae, W. Robberson, H. Ohtani, Y-M. Kim, and C. Watanabe,

, 104993 (2021). https://www.sciencedirect.com/science/article/abs/pii/S0165237020308081

 is the President of ATRq, in Orem, Utah, USA.

 is the Technical Director of Frontier Americas, in Houston, Texas USA.

 is the founder and Principal Engineer at Ocean P3 Systems, in San Francisco, California, USA, and a former engineer for the U.S. Environmental Protection Agency (USEPA), Region 9, in San Francisco, California.

 is an environmental scientist with EPA Region 9 Superfund Division and is based in Los Angeles, California, USA.

 is Co-founder & CEO of Kamilo, Inc. in Oakland, California, USA, and a former engineer for USEPA Region 9.

 is the Vice President and Manager of Global Product Marketing at Frontier Laboratories, in Koriyama, Japan.

Articles

Peer Reviewed

Py–GC–MS has advantages for the analysis of environmental microplastic samples compared to other leading analytical methods, including spectroscopic techniques.

Identification of Microplastics in Environmental Monitoring Using Pyrolysis Gas Chromatography Mass Spectrometry Analysis

Although symmetric peaks with Gaussian shapes are predicted by models of the chromatographic process, “perfect peaks” are not observed very often outside of textbooks. Tailing peaks

the most common type of asymmetric peak shape

can negatively affect both the qualitative and quantitative performance of liquid chromatography (LC) methods. In this first instalment of a multi-part series focused on the causes of peak asymmetry, I will discuss basic concepts in peak shape, and the potential for poor fluidic connections to cause peak tailing in a separation where the peak shape would otherwise be excellent.

I’d like to take a few words here to remind readers about the electronic version of John Dolan’s “LC Troubleshooting Bible”, a website (https://lctsbible.com/) with a compilation of all 390 of John’s “LC Troubleshooting” articles. I personally find this to be an incredibly rich resource for a lot of practical high performance liquid chromatography (HPLC) knowledge, and obviously a great resource for troubleshooting details. Pro tip: If you click on “Find posts by tag”, a word cloud will appear with all of the keywords associated with all of John’s “LC Troubleshooting” articles, and the size of the word is related to the frequency with which that keyword has appeared over time. This is a great way to quickly locate the articles that are likely to be relevant to a particular troubleshooting task; sometimes keywords will catch my eye that I had not been thinking about. I really encourage all readers to bookmark this resource, and explore all that it has to offer.

Well established models of the chromatographic process predict peak shapes that can be very closely approximated by the Gaussian distribution (1). There is a long list of reasons why the Gaussian peak shape might not be observed, but the Gaussian is very convenient, particularly for use in chromatography simulation tools, and development of theory that relates important parameters (such as, for example, peak width to the retention characteristics of a given molecule). In this context, one important property of the Gaussian is that it is a perfectly symmetric function. In practice, however, perfectly symmetric peaks are not observed very often, and the degree of deviation from perfect symmetry can vary widely, depending on a large number of physical and chemical factors specific to particular instruments, applications, and operating conditions. These deviations from symmetric peak shapes (that is, 

) are important in practice because the asymmetry can negatively impact both the qualitative character of a chromatogram and the quantitative performance of a method. The problems are not unique to liquid chromatography, nor are they new. Indeed, these issues have been discussed many times by many authors in 

 in the past. For example, I would point readers interested in learning much more about this area to an excellent series of articles by John Hinshaw (2–4), several articles by John Dolan (5), and two recent entries in the “

 Blog” by Tony Taylor (6,7). Nevertheless, given the importance and potential impact of peak asymmetry on the quality of separations and chromatographic results in day-to-day work, I think it is useful to draw several relevant ideas together in one place and discuss them in some detail, all with the aim of providing a robust resource to troubleshoot problems with peak asymmetry when they do arise. In this first instalment on the topic, I will review some basic concepts in peak shape and asymmetry, and then go on to start discussing some of the root causes of asymmetry, as well as potential solutions.

The equation used to model a chromatographic peak using the Gaussian function is shown in equation 1, where 

 is the analyte concentration at the detector at a particular time, 

 is the standard deviation of the Gaussian in time units, 

Many different models can be used to simulate tailed chromatographic peaks (1). One of the most commonly used models is the exponentially-modified Gaussian (EMG) function, shown in equation 2, where τ is the relaxation time associated with the exponential, tailing part of the peak, and 

Figure 1 shows an overlay of four peaks. The black peak is perfectly symmetric, simulated using the Gaussian function in equation 1. The other three peaks were simulated using the EMG function in equation 2, with increasing τ values.

A number of metrics to quantify the extent of peak asymmetry have been developed over the years. Two of the more commonly used metrics are the peak asymmetry factor (

). Figure 2 shows the measurements needed to calculate these values using equations 3 (

). Both metrics rely on measures of parts of the peak width, with 

 reliant on the widths at 10% of the peak height, and 

 reliant on widths measured at 5% of the peak height. Generally speaking, 

 values below 2 are preferred whenever possible. I’ll also note here that recent work by Wahab and associates provides other approaches to quantifying characteristics of peak shapes that take the entire peak shape into account, as opposed to measures of the peak width at specific heights (8).

Asymmetric peaks can negatively impact both qualitative and quantitative aspects of HPLC results. Figure 3 shows an example of how peak tailing can obscure the presence of a small peak in the tail of a much larger peak. When both peaks are symmetric, it is easy to see the small second peak on the right side of the larger peak. However, as the extent of tailing increases, it becomes more and more difficult to see the small second peak, and eventually it disappears into the tail of the large peak altogether.

Peak asymmetry can also affect both the accuracy and the precision of quantitative results. Figure 4 shows how more resolution is needed to obtain accurate estimates of peak area than that needed (typically a resolution of 1.5 is sufficient with highly symmetric peaks) with symmetric peaks. In the example shown here in the left panel, the peak area for peak 2 is overestimated because the tail of the larger peak 1 runs into the integration area for peak 2 when the drop method of integration is used. Of course, this can be improved in some cases using other more sophisticated integration methods, but other methods may not be as robust as the drop method. The magnitude of this type of error can range from less than 1% to more than 10%, and generally increases as resolution decreases, the ratio of peak heights increases, and the extent of tailing increases.

Common Causes of Peak Asymmetry in Liquid Chromatography

The common causes of peak asymmetry in liquid chromatography can be roughly grouped by whether they have a primarily physical or chemical basis.

Poorly packed particle bed at the time of manufacture

Rearrangement of the particle bed during use due to physical stress (for example, repeated pressure cycles)

Contamination of the particle bed or column inlet/outlet frits with physical debris (for example, particles shed from valve seals)

Mismatch between the solvent composition of the injected sample and the mobile phase

Extracolumn peak broadening and distortion outside the column due to poor connections or inappropriately large connecting tubing.

Rearrangement of the particle bed during use due to chemical stress (for example, dissolution of silica particles at high pH)

Overloading of analyte adsorption sites on the stationary phase when there are multiple types of sites with different populations

Slow kinetics of analyte desorption from the stationary phase.

This list is obviously too long to talk about each case in any detail; these will be the focus of future instalments of “LC Troubleshooting”.

Impact of Poor Fluidic Connections Outside the Column on Peak Shape

Of all the possible causes of peak tailing listed above, I think one of the most common—but easy to solve—causes is related to poor fluidic connections outside of the column. Tony Taylor has written about this recently in the ”

 Blog” as it relates to gas chromatography (6). In that context, there are multiple ways a “bad connection” can arise, including improper positioning of a column (for example, in a mass spectrometer interface), a poor cut on the end of a column (that is, a rough cut, or one that is not perpendicular to the long axis of the column), or use of the wrong ferrules. In HPLC, similar pitfalls exist. From the point of view of effects on peak shape, the only connections that matter are those between the sample injector and the detector. Typically, this flow path will involve four to eight connections, but could include more, depending on the complexity of the system (for example, injector -> capillary, capillary -> heat exchanger, and so on). Many laboratories use “pre-cut” connecting capillaries, where the ends of the tubing are prepared by the tubing manufacturer to give clean, square ends. Although I cut my own stainless steel tubing in graduate school, I am not aware of any laboratories that cut their own metal tubing these days. However, in my laboratory we routinely cut polyetheretherketone (PEEK) tubing to length for specific purposes. One can cut PEEK tubing using something as simple as a sharp razor blade, but this is not really recommended because it is very difficult to obtain a very square end on the tube with this approach. Most chromatography suppliers sell cutters for both plastic and metal (and even PEEK-clad fused silica) tubing that are specifically designed to give clean, square cuts on the ends of these tubes. Using a tube with an end that is not square will lead to a significant void space at the bottom of a capillary connection port like that shown in Figure 5(b), except that the shape of the void will look like a trapezoid rather than a rectangle. This extra void space will, in turn, lead to mixing of the analyte band as it moves across the gap, which manifests in chromatograms as peak tailing like that shown in Figure 6 (9). The good news is that this particular problem can be avoided easily by either using pre-cut capillaries, or using the proper tools to cut your own tubing and ensure clean, square cuts.

The other major problem with connections that can lead to peak tailing in HPLC is clearly illustrated in Figure 5(b). Here, the fitting has been tightened (referred to as 

) when the end of the capillary was not seated properly at the bottom of the connection port. With a compression-style connection that uses a metal ferrule, once the ferrule position is set, it cannot be moved, and every time this capillary is connected to a fitting port, there will be a void space like that highlighted in green, which again will lead to peak tailing like that shown in Figure 6. There are multiple potential solutions to this problem. If a situation like that shown in Figure 5(b) exists with a metal ferrule, then the only solution is to throw the capillary away and start over. When making the new connection for the first time, be careful to make sure the tubing end is seated at the bottom of the fitting port before swaging the ferrule. Alternatively, one can use a fitting that does not involve a metal ferrule (usually PEEK, or a graphite-filled polymer) so that the position of the ferrule relative to the capillary end can be adjusted each time it is connected to a fitting port. Several manufacturers now sell very high quality fittings of this type. Readers interested in learning more about these options are referred to a previous instalment of “LC Troubleshooting” focused on this topic (10).

In this instalment of “LC Troubleshooting”, we have reviewed some basic concepts in peak asymmetry and how asymmetric peaks can affect the quality of HPLC results. There are many physical and chemical causes of peak tailing in particular—too many to discuss here. In this instalment, we’ve discussed how poor fluidic connections in the flow path between the sample injector and detector can lead to peak tailing. The extent of tailing resulting from this particular cause can be serious, but the good news is that it can be addressed quite easily by fixing poor connections with new capillaries or fittings. In future instalments, we will continue discussing some of the other causes of peak asymmetry in detail, all with the aim of adding to your personal knowledge base that can help you troubleshoot poor peak shapes when they arise.

Data Analysis and Signal Processing in Chromatography

 (Elsevier, Amsterdam,
The Netherlands, 1998), Chapter 3, pp. 43–96.

T. Taylor, “GC Diagnostic Skills I – Peak Tailing”, 

 (2020). https://www.chromatographyonline.com/view/lcgc-blog-gc-diagnostic-skills-i-peak-tailing.

T. Taylor, “HPLC Diagnostic Skills II – Tailing Peaks”, The

Blog (2019). https://www.chromatographyonline.com/view/lcgc-blog-hplc-diagnostic-skills-ii-tailing-peaks.

M.F. Wahab, D.W. Armstrong, and D.C.
Patel, 

T. Li, D. Zeko and M. Fuehrer, “Optimizing HPLC and UHPLC Performance with Agilent A-Line Quick Connect Fittings,” presented at Analytica 2014, Shanghai, China, 2014.

is the editor of “LC Troubleshooting”. Stoll is a professor and the co-chair of chemistry at Gustavus Adolphus College in St. Peter, Minnesota, USA. His primary research focus is on the development of 2D-LC for both targeted and untargeted analyses. He has authored or coauthored more than 75 peer‑reviewed publications and four book chapters in separation science. He is also a member of 

’s editorial advisory board. Direct correspondence to: amatheson@mjhlifesciences.com

LC Troubleshooting

Tailing peaks—the most common type of asymmetric peak shape—can negatively affect both the qualitative and quantitative performance of LC methods. This instalment will discuss basic concepts in peak shape, and the potential for poor fluidic connections to cause peak tailing in a separation where the peak shape would otherwise be excellent.

But My Peaks Are Not Gaussian!  Part 1: Basic Concepts in Peak Shape, and the Effect of Extracolumn Connections

It is often a good idea to look at our methods and see how we can make them more efficient and effective without doing a major overhaul. In this month’s column, I examine simple steps we can take to optimize a separation without changing the column. I discuss adjusting temperature, flow, and sampling with an eye towards the classical method optimization goals of high resolution, high speed, high sample capacity, and ease of use. I examine optimizing as much as possible without getting the wrenches out to install a different dimension or stationary phase column.

One of the first and most important lessons I learned about optimizing and troubleshooting chromatographic problems is to try the simple fixes first and turn wrenches last. Changing a column in a gas chromatography (GC) instrument usually requires additional downtime to cool the instrument, install the new column, heat it back up, condition or bake out the new column, and then think about a change in peak elution order. Because these steps can be very time-consuming, I always taught my students to get as far as they can with optimizing the separation on the current column before changing it. Let us consider some optimizations that can be attempted without changing column dimensions or the stationary phase.

Before considering optimization steps, it is important to consider the goals of the optimization process. Many short course instructors (including me) use a simple triangle approach to assist students in thinking about optimization. A classical optimization triangle is shown in Figure 1. Each corner of the triangle represents an optimization goal—high resolution, high speed, or high capacity (ease of use). In method development and optimization, we can generally obtain any two of the three goals simultaneously. Obtaining all three often requires difficult compromises. A high-resolution, high-speed method will be more difficult to run and will likely have a lower sample capacity or require a small sample. A high capacity (easy) method that is also fast will likely be low resolution. Finally, a high-resolution method that has high capacity and is easy to run will generally be slow.

Because the goal of a chromatographic separation is resolution, the classical resolution equation, shown in equation 1, can also be helpful in thinking about the chromatographic challenges in method optimization.

Resolution is seen as a combination of retention factor (the “

” term), selectivity (the “α” term), and efficiency (the “

” term). In these examples, we look at the terms in this equation broadly. Resolution can be thought of as involving long retention, as seen in the term involving 

 is also small, so retention does not contribute to resolution. If 

 is too large (approximately 20), then there is little resolution gained by increasing 

 by lowering the temperature. Selectivity, α, is a measure of the relative strength of intermolecular interactions between an analyte pair and the stationary phase. We see several cases where a lack of selectivity hinders our ability to achieve resolution while maintaining our goal in this exercise of not turning wrenches. Finally, resolution is dependent on 

, the number of theoretical plates, or efficiency. We would like the sharpest possible peaks, and because retention and efficiency are most affected by temperature and carrier gas flow rate, these are the variables that most influence resolution without turning wrenches. In our discussion, we use these three ideas of retention, selectivity, and efficiency in a broad framework, rather than dissecting the equation, as our examples are temperature programmed and equation 1 assumes constant temperature.

With the first step being to set the goals of the optimization, we now examine each of the three scenarios and see how each can be optimized without changing the column. We assume temperature‑programmed separations and discuss optimizing for each case. In my own laboratory, we do most of our initial method development work on 15 or 30 m × 0.25 mm × 0.25 μm 5% polydimethylsiloxane columns, so these dimensions will be used as examples throughout the discussion. Chromatograms were simulated using Pro-EZGC online software from Restek (1). To get started, Figure 2 shows a sample chromatogram of 

-alkanes separated on a 15 m column as described above with a 10 ºC/min temperature program, starting at 40 ºC and a column flow rate of 1.5 mL/min. These simple conditions are typical of scouting runs that might be performed at the outset of method development for a new separation problem. For a simple comparison between runs, note the retention time and elution temperature of the last peak (dodecane) are 7 min and 110 ºC, respectively, and the resolution between the last two eluted peaks is 37.

The combination of high resolution and high speed is often the most challenging scenario. In Figure 2, there is already high resolution, so we can optimize for speed by increasing the temperature program rate. For most temperature‑programmed separations, increasing the speed of separation, effectively reducing retention (

) by increasing the temperature programming rate, will reduce the resolution. In this case, this is fine because baseline resolution is 1.5 and our current resolution is 37. The injection technique and speed of injection can also impact the maximum speed of the separation. In fast GC, a split injection with a high split ratio is used to ensure the narrowest bands entering the column, ensuring a high number of theoretical plates. However, a split injection with a high split ratio increases the limit of detection (LOD) as more of the sample is ejected from the split vent instead of entering the column. Additionally, the column flow rate can be increased to drive analytes in the vapour phase more rapidly through the column, causing a potential reduction in the number of theoretical plates that is mitigated by temperature programming, as discussed in a previous column (2).

Figure 3 shows two further optimized chromatograms of the alkane mixture. Figure 3(a) shows the temperature programming rate increased to 20 ºC/min and Figure 3(b) shows the separation in Figure 3(a) with the flow rate increased to 3 mL/min. In Figure 3(a), note that the retention time for dodecane has reduced to approximately 4.4 min, reducing the run time by approximately 35%. The resolution of the last two peaks is slightly lower, still approximately 35, because of the peak widths becoming smaller as the retention time is decreased. However, the elution temperature of dodecane increases to 128 ºC, an increase of nearly 20 ºC. A higher elution temperature increases downtime between runs, reducing the speed benefit. Extremely high elution temperatures that approach the maximum column temperature can reduce column lifetime over time. In Figure 3(b), the added increase to the flow rate further reduces the retention time of the dodecane peak to approximately 3.7 min and its elution temperature to 114 ºC, which is almost back to the original. Resolution of the last two peaks is still quite high (approximately 30). By increasing both the temperature program rate and the flow rate, we were able to reduce analysis time by nearly 50% with a similar elution temperature and a small cost in resolution. This separation could be optimized even further for speed because there is still excessive resolution between the peaks; however, for today, we will be satisfied with a 50% reduction in analysis time in two simple steps, with little loss of resolution.

A requirement for both high speed and ease of use or high capacity also presents special challenges. This combination usually requires that the problem involve few analytes (and few interferences!), so that there is not a great need for high resolution. High speed usually means low retention, high selectivity, and high efficiency. Resolution is provided by high efficiency or selectivity and not by long retention. Figure 4 shows the separation of the five components with further optimizations for speed. The initial temperature is raised to 80 ºC, the temperature programming rate is increased to 30 ºC/min, and the flow rate to 5 mL/min, with a column head pressure of about 26 psig and a gas hold-up time (

) of about 15 s. The retention time of the final peak is now about 1.4 min with an elution temperature about 120 ºC. The faster programming rate and higher initial temperature have combined to raise the elution temperature. However, this final temperature is only approximately 40 ºC higher than the initial temperature, so downtime between runs would be significantly reduced when compared to the separations in Figures 2 and 3. Resolution between the last two peaks has reduced to approximately 25 but is still very high. This separation could be further optimized for speed, but the differences seen between Figure 4 and Figures 2 and 3 are dramatic. Further optimization can be tried using this online simulator or by method translation (3). With three simple changes, increased temperature programming rate, increased initial temperature, and increased flow rate, we have reduced the run time by 80% and the downtime between runs by approximately 50%, with a resolution loss of about 30%.

Ease of use or high capacity and high resolution require a slower analysis; these are at the apex of the triangle in Figure 1 farthest from speed. In this case, high resolution comes from selectivity. The column with the most separating power will also likely have high retention and require less efficiency. Classical separations of petroleum-related samples on nonpolar stationary phases such as the one shown in Figure 5 illustrate this problem. This sample of brown mousse from the surface of the Gulf of Mexico during the Deepwater Horizon oil spill in 2010 was prepared by placing about 100 mg of the brown mousse in a vial and diluting with pentane, so the sample preparation was very simple. The conditions, shown in the caption, are very common and simple. As seen from this figure, while the separation efficiency is quite high and the retention is strong, there is not enough selectivity to achieve a full separation of the many components of this mixture in a simple analysis. This sample could be better separated by a lower temperature programming rate, but this 15 m, thin‑film column alone lacks the separation power to fully separate this mixture.

Although this is a much slower analysis than those presented in the simulated chromatograms in Figures 2–4, it is not slow enough to fully separate this highly complex mixture. Samples like this one led to the rapid development of multidimensional techniques, such as comprehensive two-dimensional gas chromatography (GC×GC), over the past decade (4). In analyses like this, sample preparation and injection are often quite simple, with high split ratio split injection employed for both simplicity and to ensure narrow peak width.

There is much method optimization that can be done without turning any wrenches. Looking back at the analysis goals, broadly at equation 1 and thinking about capillary columns in general, we see that even relatively short columns can have high plate numbers, with the 15 m columns discussed here having plate numbers around 50,000. Retention is controlled using temperature, with lower temperatures and lower temperature program rates leading to longer retention. Flow rate can be used as a final optimization to speed up the analysis with a higher flow rate.
In temperature programming, higher flow rates lead to lower elution temperatures with surprisingly small effect on resolution; lower elution temperatures can increase column lifetime. In my own laboratory, we start with a column like the 15 m column described here and optimize the separation as far as we can before getting out the wrenches and changing the column.

Finally, I close with some sad news. My research, career, and life mentor, Professor Harold M. McNair, passed away on 27 June 2021, at peace and surrounded by his family, in Blacksburg, Virginia, USA. He was 88 years old. Harold was a man of the world whose work in research, teaching, and being a great mentor and friend touched everyone who practices GC. His career as a chromatographer spanned over 60 years, including publications in every decade from the 1950s to the 2010s. In 1959, he authored the first PhD thesis on GC in the United States. His most recent paper was published in 2019. If you took a webinar or short course online over the past year, or ever took a short course, you have Harold McNair to thank for the idea of short courses in chromatography, which he started in the 1960s. His classic book, 

, has been published all over the world since the first edition in 1965 and set the original standard for training books and manuals (5). 

 brought GC into the mainstream and made GC accessible to all scientists, not just chemists and chromatography specialists. I plan on writing more about Harold, lessons from his work, and his impact on chromatography and analytical science in an upcoming column.

“Pro-EZGC Chromatogram Modeler” Restek Corporation, 2021. https://ez.restek.com/proezgc (Accessed July 2021).

“GC Method Translation Software,” Agilent Technologies, 2013. https://www.agilent.com/en/support/gas-chromatography/gcmethodtranslation (Accessed July 2021).

Basic Multidimensional Gas Chromatography

 (Elsevier, Amsterdam, The Netherlands, 2019).

H.M. McNair, J.M. Miller, and N.H. Snow, 

 (John Wiley and Sons, New York, New York, USA, 2019).

 is the Founding Endowed Professor in the Department of Chemistry and Biochemistry at Seton Hall University, and an Adjunct Professor of Medical Science. During his 30 years as a chromatographer, he has published more than 70 refereed articles and book chapters and has given more than 200 presentations and short courses. He is interested in the fundamentals and applications of separation science, especially GC, sampling, and sample preparation for chemical analysis.His research group has ongoing projects using GC, GC–MS, two‑dimensional GC, and extraction methods. Direct correspondence to: amatheson@mjhlifesciences.com

GC Connections

This column will examine simple steps that can be taken to optimize a separation without changing the column. Adjusting temperature, flow, and sampling with an eye towards the classical method optimization goals of high resolution, high speed, high sample capacity, and ease of use will be discussed. 

Capillary Gas Chromatography: Getting the Best Separation Without Turning (Too Many) Wrenches

There is a hidden factory in plain sight in many laboratories. Not many people know of its existence, yet it stares back at you when you work. What is the output from this hidden factory? Paper.

Ask any analyst what the function of a laboratory is and why they work there and you will get one of two answers. The first may seem frivolous but it’s true, and that is they to get to play with some amazing toys—sorry, analytical instruments. The second is to analyze samples, interpret the data, and generate results for people to make decisions—as well as play with some amazing toys. What nobody will ever say is that the function of a laboratory is to generate reams and reams of paper records, and that they just love clicking on a print button. Welcome to the hidden factory.

In many laboratories, paper constitutes most analytical records. Whilst there are electronic records generated and stored in instrument data systems, many laboratories still print paper out of custom and practice, or following an edict from the sadists in quality assurance (QA). Welcome to the nightmare world of paper and hybrid systems that make up the hidden factory.

Warning! Although this column focuses on paper printouts, don’t forget the electronic records sitting in computerized systems throughout your laboratory—protect and preserve them please.

This column will look at the advantages and disadvantages of both paper records and hybrid systems (there are not many of the former but plenty of the latter), with suggestions to reduce the amount of paper generated by the hidden factory. We will finish by discussing approaches to reduce paper storage volume and costs by scanning and destroying paper records that you have just created. I could not make this up, even if I tried.

This is a very short section: what are the advantages of paper records?

A paper record is easy to establish and is infinitely flexible. If you modify a process, just redesign the associated paper form. No messing around with software configuration and no computer validation to worry about.

Errors are easy to correct and a reviewer does not have to wade through all those annoying audit trail entries to see them. Corrections are in plain sight, although you may have to squint to read the reason for change with some individual’s handwriting.

Paper is a relatively robust record medium with less likelihood of data loss, except when you attempt undocumented chromatography by spilling solvent on the pages. Paper has well understood archive properties as long as the environmental conditions are maintained and monitored.

That was quick—now let’s look at the problems with paper.

Unsurprisingly, there are several disadvantages of paper records.

Blank forms can be at the heart of data falsification and therefore they need to be controlled by uniquely numbering each form, allocating a form to an individual or group for use, and reconciling when the work is complete as required by data integrity guidance (1–4). For more information on how to implement controlled blank forms see the article by Burgess and McDowall (5).

Blank forms are expensive to administer, but the cost is hidden from management as it comes from general staff payroll, rather than a specific budget item. The cost of long-term record storage of records is expensive, either with on-site buildings and associated staff or off-site storage rental costs.

Think about your laboratory notebook—anybody can read it when placed on the top of your desk, but nobody can if it is locked in a desk drawer. Look further to QC batch records from product analysis. There is only one copy of the paper record and it is difficult to share all the information with more than one person. Of course, your laboratory has never lost a batch record.

Although there is a requirement for trending some analytical data in EU GMP Chapter 6.9 (6), this is very difficult to achieve with paper. It requires that information be abstracted from paper and entered into an informatics application over time to trend results. This is a slow, tedious, and error-prone process. Searching through paper records to extract data for annual product reviews (7) or product quality reviews (8) is the same.

 It is obvious that paper is a physical record medium that, in large quantities, has bulk and weight, requiring much storage space within a controlled environment. While the days of loading copies of drug submissions into the back of a lorry with a forklift truck are long gone, the same can’t be said for analytical batch records. Typically, records are stored on-site for up to two years and then transferred to an off-site records management facility until the end of the retention period when they should be formally destroyed. If required for inspection, audit, or complaint, investigation records should be retrieved rapidly under the service agreement with the records management company.

Back to the hidden factory and its only product: printed paper records.

We need to examine which tasks in an analytical process generate most of the hidden factory’s paper. Figure 1 shows a typical analytical process involving chromatographic analysis in green and highlighting the paper records that are generated during the hidden factory’s operation in yellow. This figure focuses on the paper outputs from both manual processes, such as sampling and sample preparation, as well as the chromatographic analysis and post analysis calculation of results from sample aliquots and generation of the reportable result.

You can see that paper records are generated by virtually all tasks throughout this process. Typically, the second person review may also generate some paper by using a review checklist to confirm that work has been performed correctly. If there is an out of specification (OOS) investigation, the hidden factory can be working overtime or double shifts, producing even more paper records. The scope of any hybrid systems, for example, the chromatography data system (CDS), as well as any spreadsheet use, is shown at the bottom of Figure 1 in blue. This is where we will focus our discussion to reduce some of the hidden factory’s paper production, as well as simplifying the second person review process.

Welcome to your worst nightmare: hybrid systems. We have signed paper printouts generated from electronic records in the CDS coupled with the regulatory requirement in 21 CFR 11.70 for record signature linking (9). Most hybrid systems print paper as if it is going out of fashion, helping the hidden factory production quotas. At a minimum, the paper printout needs to be initialled and dated at the beginning and end of the printout by both the analyst and reviewer. If the sadists in QA insist, each page must be initialled by the tester and again by the reviewer, which are yet two more error-prone and tedious tasks. The printouts must also be checked against the electronic records, including those that were created or used in the analysis. Never mind, just think of playing with the toys when you have finished moving paper around the laboratory. But wait!

There is a bonus from the hidden factory. If a reviewer finds that there is an error in any data in the CDS, the paper records are returned to the analyst to correct. After the correction, guess what? Some or all the corrected record set are printed again and added to the batch records. The hidden factory is the gift that keeps on giving.

Let me ask you a very simple question: are you lazy? Those answering no will be asked if they are masochists and condemned to an eternity of initialling and dating paper records in purgatory. Lazy people, please follow me to the sunlit uplands of simpler regulatory compliance—even with a hybrid system.

A simple way to reduce the amount of paper printed from a hybrid system is outlined in appendix 1 of the WHO data integrity guidance (10). Figure 2 shows a controlled review form for documenting both the creation and review of the record set created during an analysis with a hybrid CDS system. The review form is linked to the analytical procedure as well as the standard operating procedure (SOP) for review of laboratory records and the audit trail review work instructions for the computerized system that has generated the electronic records.

At the start of the analysis, a uniquely numbered version of the form is issued to the tester. This person documents the start and end dates and times of the analysis (to help the reviewer search audit trail entries), as well as listing all records created, modified, and where they are stored. The only printout from the CDS is a summary of the reportable results and a list of the contextual metadata used, for example, the instrument control file, data acquisition, and processing methods. Nothing else is printed from the system. The tester signs both the results printout and review form.

A much lighter set of paper records is passed to a supervisor or peer who will conduct the second person review as follows:

All data files are reviewed on screen. This requires a large enough screen to have two windows open—the first to see the data and the second to review other files, such as the acquisition method or the audit trail. This approach avoids the check of the bulk of the paper printouts and makes the task simpler and quicker. It is also an opportunity to ask the boss for a high definition 55-inch screen for reviewing data; any other uses I’ll leave to your imagination.

The reviewer checks the CDS data and contextual metadata files generated by the performer of the test and documents these on the review form.

The applicable audit trail entries are also reviewed and documented. Two windows are very useful for correlation of data modifications in the data and contextual metadata with specific entries in the audit trail.

Checks for falsification of data are included in the form and will be checked during data integrity audits and, if applicable, data integrity investigations.

If changes are needed, these are documented and sent back to the tester to update.

When the review is complete, the reviewer then signs the form.

If required by laboratory procedures, there may be space on the form for an approval or QA signature.

Regardless as to whether you implement the paper reduction process outlined above or not, all hybrid systems need the electronic records backed up. Many hybrid systems are standalone and the process owner is responsible for electronic record backup as required under EU GMP Annex 11 clause 7.2 (11) and 21 CFR 211.68(b) (7). Do not use USB devices for backup, as discussed in a recent 

 column (12). Instead, connect the workstation to the organization’s network, either using the existing network socket on the PC or by installing a network card, and then use an agent from the IT department’s backup system to regularly backup the system. Alternatively, IT should write and verify a script to perform the same function automatically after initiation by a user. Ensure that the backup process works by performing test restores on a regular basis.

Having spent time and effort generating a humongous amount of paper we come to the pièce de la résistance as we say in Britain. The increasing storage costs for the paper may be spotted by an eagle-eyed senior manager. This may generate a mastermind light bulb moment: “Let’s cut the costs of external storage by scanning all these records and then destroying the paper. Look at the cost savings!” At this point the initials MBA after their name simply mean Minus Brain Activity.

The more sensible option is investing in informatics solutions such as laboratory execution systems (LES), electronic lab notebooks (ELN), or laboratory information management systems (LIMS) to capture data electronically at point of origin and eliminate paper printing in the first place. Why spend a huge amount of money scanning and destroying paper? How dumb and stupid is this?

The UK regulatory authority, MHRA, published advice for scanning and destroying good manufacturing practice (GMP) records on their old website that, unfortunately, is no longer available publicly. I have a word version available that I have used as a basis to write this section. Outside of the GMP arena there is good laboratory practice (GLP) guidance from the AGIT, a Swiss regulator and industry collaboration (13), as well as Question 6 of an FDA draft good clinical practice (GCP) guidance that discusses scan and destroy (14) and Section 5 in an EMA GCP guideline (15). Even the German Federal Office for Information Security gets in on the act with an English version of guidance for replacement scanning (16).

The first task is to perform a risk assessment of the proposed process before starting work. We will assume that this risk assessment does not veto the approach. The overall process is described in Figure 3, which is divided into three phases as follows:

The first phase, shown in red in Figure 3, is to define how you will scan the paper records and the equipment to be used. This must have high capacity automatic document feeders for the defined make(s) and model(s) of the scanners. For the latter what resolution is adequate for the job? Does the scan need to be coloured or just black and white? Single or duplex scanning? Can you search the resulting file (typically a PDF file)? Will any software be used or will the embedded software in the scanners be used instead? How will scanned copies be verified? Will a separate application be used to electronically sign to verify that the scan is a true copy or will the scanning application have electronic signature capability?

Once the system is defined, the requirements can be written and tested. To complete this phase, a procedure for scanning and verifying the copies must be written and staff trained. How will the records be named? (A file naming convention must be agreed from the beginning.) Where and how will the scanned records be stored? If this is not done, how will you retrieve the records in a hurry?

2. Identify Records to be Scanned and How to Destroy Paper after Scanning

The paper records in the process are shown in yellow in Figure 3. Records for scanning must be checked for completeness before starting the process. After scanning, the records need to be assessed to see if they can be destroyed or if a legal hold should be placed on them due to any pending legal action. If there is no legal impediment the formal process for destruction should be signed off.

adequate Quality Control (QC) checks are performed and documented during the transfer process (to provide assurance that unacceptable errors have not occurred during the transfer process).

In contrast, the AGIT guidance proposes a 100% check of all pages (13). I agree. You must know that if 100 pages went into the document feeder then 100 pages of scan came out. That is not the end of the story. Each and every page must be checked versus the original to ensure the whole page was scanned, there were no misfeeds, orientation is correct, the content is legible, characters have not been misinterpreted by the software, colour reproduction is correct, and that the resolution is adequate. Only then can an individual verify that the electronic file is a true copy by electronically signing the file. The act of signing must lock the file to prevent any post-scanning tampering and ensure the integrity of the contents. The file must be stored in a secure location for long-term retention.

Backups of the system storing the scanned records must be performed along with periodic checks to ensure that the data are still readable.

What a pain—all because there was not the foresight to implement an electronic solution in the laboratory in the first place. Instead, the laboratory creates Great Mountains of Paper (GMP) only to then scan and destroy it. You couldn’t make it up! Welcome to your laboratory’s hidden factory.

A laboratory’s hidden factory produces paper records in abundance. A simple way of reducing the amount of paper printed from a hybrid system is described to help both the analyst and reviewer and make both jobs easier. Laboratories may also consider scanning and destroying paper to reduce storage cost; this is only a sticking plaster and fails to address the problem of a lack of foresight and investment. Tackle the problem at source and invest in informatics solution(s) to stop printing paper. Then close the hidden factory in your laboratory.

FDA Guidance for Industry Data Integrity and Compliance With Drug CGMP Questions and Answers

August 2021 | LCGC Europe

Identification of Microplastics in Environmental Monitoring Using Pyrolysis Gas Chromatography Mass Spectrometry Analysis

But My Peaks Are Not Gaussian! Part 1: Basic Concepts in Peak Shape, and the Effect of Extracolumn Connections

Capillary Gas Chromatography: Getting the Best Separation Without Turning (Too Many) Wrenches

The Hidden Factory in Your Laboratory?