ften I am asked how I come up with ideas for "LC Troubleshooting" each month. Usually the problem is not finding a topic, but sorting through possible topics. One of the prime sources of fodder is the regular e-mail inquiries I get from readers. This month's topic is based upon one of these e-mails. As often is the case, the inquiry came loaded with specifics that, although not proprietary, were not intended for public consumption, so I'll generalize a bit on the exact conditions. The inquiry went something like this: 

"I've been using a C18 column to analyze phenylisothiocyanate (PITC)-derivatized amino acids from protein hydrolysates. A gradient separation is used with sodium acetate as the A-solvent and acetonitrile as B, running to 100% B over 35 min. Recently, I have been seeing a large baseline disturbance toward the end of the run that interferes with the separation and eventually renders the column unusable. I have replaced all the reagents, the guard column, and the column, but the contaminant is not eliminated. 

I do not get regular access to the liquid chromatograph, and because I'm the only person using it, it often is inactive for several days between uses. It appears that the problem peak is always largest in the first few runs of the day and decreases as successive samples are run, although it never goes away completely. The peak is much worse if a blank is injected than if the gradient is run with no injection. I flush the column with 80% B at the end of each batch of samples. I know little of the LC system's history, but I can guess that it has been a long time since any part of the system has been serviced. Can you help me find the source of this problem and eliminate it?" 

The most alarming statement in the entire scenario is in the next to last sentence, where the lack of service history is mentioned. Here, I would follow the advice I've heard given for guitar players — if you can't remember when you last changed the strings, they are overdue for a change. Similarly, if you can't remember when you last serviced your liquid chromatography (LC) system, it is overdue, also. You've read the  rule-of-one here many times: Change just one thing at a time. This rule is used to help isolate a problem, because it allows us to avoid a shotgun approach to troubleshooting that might fix the problem, but doesn't give us much information about the problem source. We'll use this rule later in the problem isolation scenario. However, here's one time that I recommend ignoring the rule-of-one and going for a thorough system service. Here are some essential service items and system checks that I recommend. 

 The mobile phase reservoirs should be cleaned or replaced. This is especially true for the A-reservoir, because acetate is such a good nutrient source for growing microorganisms. Next, I would replace the inlet frits in the reservoirs (I like ≥10 μm porosity frits for minimum flow restriction). These last a long time, and might be perfectly good, but with an unknown history, they can be a source of contaminants. Once the frits have been replaced, perform a siphon test to ensure that the tubing connecting the reservoir to the pump is unrestricted. Just follow the tubing downstream from the reservoir to the first low-pressure fitting in line, usually at the mixing manifold (low-pressure-mixing systems) or pump inlet (high-pressure mixing). Start a siphon flow. I like to see at least 10 times the required solvent flow when the reservoirs are mounted 0.5 m or more above the pump. For example, if you normally run 1–2 mL/min, you should observe at least 20 mL/min of siphon flow to ensure the pump has plenty of solvent available. If the flow is too low, find the restriction and correct the problem. Finally, take a close look at the solvent transfer tubing between the reservoirs and the pump, especially from the aqueous reservoir (A-solvent), to be sure there is no visible contamination — replace any suspect tubing. 

 Replace the piston seals. These should be replaced once a year, whether they need it or not. Worn pump seals can result in leaks, inaccurate or irregular flow rates, pressure fluctuations, and shedding of particles that can block frits downstream or interfere with check-valve operation. Check the owner's manual for the pump and determine if there are any internal frits that should be replaced. The most common place for these are as part of the check-valve assembly between the two pump chambers on an accumulator style pump or at the pump outlet on many systems. 

Consult the operator's manual for specific service recommendations for your autosampler, but there are some universal maintenance items that should be considered. Clean or replace the wash-solvent reservoir and replace the inlet frit for the delivery tubing, for the same reasons as mentioned for the solvent reservoirs previously. It is easy to forget to clean or change the wash reservoir, because so little wash solvent is used, but it is best to perform this task on a weekly or biweekly basis. Check the delivery tubing for restrictions. If the wash-solvent delivery pump is a serviceable item, replace the seal or perform other recommended maintenance activities. Thoroughly clean the interior of the autosampler. There often are drips inside that leave contaminating residues. It might be prudent to replace the needle seal, or at least to take the seal assembly apart and clean it in a sonicator.  

The brand of autosampler that was mentioned uses a traditional six-port valve with a sample loop mounted on it for either filled-loop or partially filled–loop injection. Although the injection valve rotor can last >100,000 cycles, there is no way to know its history in the current case, and a worn rotor can be the source of sample carryover. For this reason, I would replace the rotor seal. Be sure to label each of the pieces of tubing connecting to the valve and draw a diagram to guide proper reassembly. Sonicate all the valve parts in methanol to help clean them, install a new rotor, and reassemble the valve. Take particular care to ensure that all the tubing connections are seated properly. Any small gaps between the tube ends and the injector body can serve as tiny reservoirs to cause sample carryover. 

Most detectors do not require routine servicing, but it would be wise to check the operator's manual for any recommendations for your particular detector. Detector cell contamination is unlikely to be the source of the present problem, and cell-cleaning techniques (other than flushing with mobile phase at the end of a batch of samples) can be more likely to damage the detector than to correct nonexistent problems. 

 At this point, it would be a good idea to run a system performance check, such as that described in reference 1 (available in the archives section of 

). I like to run this test once or twice a year to ensure that the LC system is performing as it should. Correct any additional problems that are highlighted by these tests. 

Before you put the instrument back in service, document the service activities that you performed. I find that this is best done in a notebook that is kept with the system. I prefer a looseleaf notebook that can have sections for different maintenance and testing activities. In each section, I keep several blank copies of various record forms that I have created to speed record keeping as well as later review of the data. Another option is to use a bound notebook, such as a laboratory notebook or theme book. If the notebook is kept with the instrument, it is handy for making notes or checking on past maintenance, and it is more likely to get passed on to the next user. 

Now that you've serviced the instrument, you hopefully have corrected the problem. We can go back to the rule-of-one and make changes one at a time to help isolate the problem, if it remains. Make up fresh mobile phase and equilibrate the column. I would start with two or three blank, no-injection gradients to see if the problem peak is present or not. If it is present, it is unlikely that it comes from the autosampler, but just to be sure, bypass the autosampler by plumbing the pump directly to the column and repeat several no-injection gradients.  

If the problem persists without the autosampler in the flow path, contaminated mobile phase is a possible problem source. Usually, it is the A-solvent that is the problem. Test this by making two no-injection blanks with the normal equilibration between runs followed by a run with equilibration three times as long before the gradient is run. Compare the second and third runs. If the problem is associated with the A-solvent, the background peaks in the third (long equilibration) run should be approximately three times as large. If this is the case, try substituting each of the mobile-phase components (including water) for new ones, one at a time. Do not neglect the possibility of inadvertent contamination of the mobile phase from the pH meter. This problem, although not common, can occur if the pH probe is dipped in the bulk buffer while adjusting the pH (2). Eliminate the possibility of this contamination source by checking the pH of an aliquot of buffer, then discarding the aliquot, rather than dipping the probe into the bulk mobile phase. 

If the problem persists only when an injection is made, you need to isolate the problem source within the autosampler. I would try injecting different volumes of a blank sample to determine if the problem is associated with the sample or some other part of the autosampler. For example, you can inject  10, 20, and 30 μL of blank and see if the problem peak grows in proportion to the volume. If this occurs, it is something in the blank that is causing the problem. Once again, you'll have to isolate the problem source by substituting each of the reagents used to make the blank with fresh ones until you eliminate the problem. 

Another area to be aware of is inadvertent contamination of your samples or mobile phase from dirty glassware. I know this is a particular problem with some of the protein hydrolysis procedures and subsequent PITC derivatization. For example, fingerprints or other organic matter can be a source of contamination. Glassware that has not been thoroughly rinsed is another possible source of contamination from detergent residues. Some workers rinse all their glassware in acid during the washing process and then rinse with organic  solvent before use. A contaminated solvent filtration apparatus can be a problem. This is checked easily by preparing mobile phase without filtration. 

Finally, adjustment of your LC method can help to convert this problem to the nonproblem category. You indicated that the large baseline disturbance occurred near the end of the run. I wonder if it is possible that this represents something that is not completely eluted from the column. Thus, it builds up over time as part of it is eluted in one run and part in the next, along with the contaminant from that run. It might be possible to eliminate the problem by placing a hold at the end of the run when you are pumping 100% B. This can allow sufficient time for the problem hump to be eluted. Alternatively, try using a stronger wash cycle at the end of each run. For example, rather than stopping at 80% acetonitrile, increase the solvent strength to 90% or even 100% acetonitrile. This, of course, will require using a stronger B-solvent and adjustment of the gradient program to deliver the same gradient used to elute the sample components.  

Although I have not listed every possible cause of the reported problem, I suspect the problem will be isolated and eliminated by the time all the procedures listed earlier are performed. The key to effective problem isolation is to use a systematic procedure so that you can identify cause and effect when a change is made and the problem is eliminated. 

In Jennifer Birchett's recent contribution to "LC Troubleshooting" (3), she mentioned finding microbial growth in the tubing transporting the aqueous phase from the reservoir to the pump (see Figure 1 of reference 3). Since that time, I have had two students in my short courses complaining of similar problems.  

This is something that I have never encountered, so I wonder how widespread the problem is. And how do you correct it? Tubing replacement is an obvious answer. Flushing with dilute nitric acid (with the column removed!), as when passivating the system also should be effective. What about dilute household bleach (also without the column)? I would be interested in feedback from you, the readership, on this topic. Do you ever observe it? How do you correct the problem? E-mail me at 

"LC Troubleshooting" Editor John Dolan has been writing "LC Troubleshooting" for 

 for over 25 years. One of the industry's most respected professionals, John is currently the Vice President of and a principal instructor for LC Resources, Walnut Creek, California. He is also a member of 

's editorial advisory board. Direct correspondence about this column to "LC Troubleshooting," 

, Woodbridge Corporate Plaza, 485 Route 1 South, Building F, First Floor, Iselin, NJ 08830, e-mail 

For an ongoing discussion of LC troubleshooting with John Dolan and other chromatographers, visit the Chromatography Forum discussion group at 

Articles

LC Troubleshooting

Deterioration of the separation can result from many causes. This month's column looks at some techniques to isolate the problem source.

System Contamination

Chromatography data systems (CDSs) are used throughout regulated laboratories in the pharmaceutical and allied industries. The role of a CDS in research and development and production (good manufacturing practice [GMP]) can be for determining the impurities of raw materials and finished products, in process control and stability testing, while in good laboratory practice (GLP) development laboratories, a system can be used for the measurement of a drug and its metabolites in biological fluids from nonclinical and clinical studies to determine the absorption, distribution, metabolism, and excretion (ADME) of the compound. Regardless of the role of the regulated laboratory, there is a need to validate to show that the CDS is fit for its intended purpose as required by the GLP or GMP regulations as well as 21 CFR 11 (electronic records; electronic signatures rule).    

Although validation of computerized systems is a topic of many books and regulatory guidance documents, only a few of which I have referenced (1–3), there is only one book that deals solely with CDS validation (4). However, this book (4) only addresses the topic of CDS validation but does not discuss the timelines of a project in any great depth.  

In this article, I would like to focus on what is needed to ensure a rapid validation of a CDS while maintaining the overall quality and managing regulatory risk. I'll look at how CDS validation is typically conducted today, what we should aim to achieve in a rapid validation projects, and then list my 10 critical success factors for such a project. Before we discuss these areas, I am assuming two things that I will not discuss further. First, that the CDS is a multiuser networked system, and second, that the users will be working with electronic signatures and keeping paper output from the system to a minimum.   

Typically, many CDS validation projects take between six months (if you are lucky) and 18 months (if you are not) to validate the core system and then roll out the system to the rest of the user base in the laboratory. This means that in the worst case, when you have finished the current validation, the CDS vendor has probably released another version of the software for you to implement and validate. In the current economic environment, it could be a way of preserving your job. Because of the time and resource required to validate each release, at best, a laboratory will only implement every other release of CDS software, which makes it difficult for a vendor to offer effective support when there could be up to five supported releases in the customer base at any one time. 

In an ideal world, a CDS validation should be fast and efficient; something like this: A laboratory would have written down their user requirements and have a good reason for the purchase of a particular system. The project team would be trained and have access to the system from the first days of the project so that they could understand the nuances of the way the software works and how it will be configured (including custom calculations and custom reports). In this way the user requirements would be refined to fit the purchased system rather then a generic CDS. Risk management would work in conjunction with what the vendor has done to focus testing down to show that the system works in your environment. Work on configuring the system with input of calculations and report templates will continue after the system has completed its validation and therefore, this aspect is, in my opinion, best controlled by procedure. It allows the process to be decoupled from the validation of the application software. 

In the introduction, I mentioned that we should aim for the validation of a core CDS to have an achievable target to aim for in the validation. Therefore, from this we have an implicit statement that we will be staging the validation in a number of phases. So what constitutes the core system? To give the classic consultant's response: it depends. The factors influencing this are  

The experience that the laboratory has with the data system to be validated (impacting the ease with which decisions can be take regarding configuration of the software and writing custom calculations and reports).

Training needs of the user base. Experienced users would obviously require much less than those who were completely new to the system 

How big is the laboratory where the CDS will be implemented? Users of the new CDS will be less productive until they are fully acquainted with the system compared to when they operated the old system. As the laboratory will not be able to stop work, then committing the whole laboratory to the new system is not an option to be considered by laboratory management.

Therefore, the core system could vary in size from about five to 20 chromatographs. This gives the project team the phase 1 target to aim for but they must also consider the whole system. Reducing the size of the initial target to hit means that the user acceptance testing (UAT) or performance qualification (PQ) also can be phased and this will be discussed in more detail in the next section. 

The 10 Critical Success Factors for Fast CDS Validation

My 10 critical success factors (CSFs) for a rapid CDS validation are listed in Table I and will be discussed in more detail under each heading. There is no particular order of these CSFs in the table and the text of this article.   

Table I: 10 critical success factors for rapid CDS validation 

Laboratory management can make or break any validation project by either refusing to resource it adequately (expecting the users to still carry out their normal duties as well as working on the project) or by unrealistic demands (I want it tomorrow, sorry yesterday). In either case, the project will suffer and the laboratory will end up with a poor and rushed validation.  

Therefore, this CSF has two aspects to it. First, management must set realistic and achievable goals, deadlines and expectations for the project team to achieve and keep to. Second, management must get out of their offices and talk publicly to the chromatographers who will be the user base to ensure positive support for the system and to aid the CDS validation project team. Ideally, management should change individuals' objectives to ensure that their role working on the project team or supporting the implementation is a part of each analyst's overall performance goals. 

It is important to understand one of the reasons that CDS validation projects take more time than anticipated is that the people working on the project are trying to do two jobs: their normal one and the CDS project. Something has to give and it is usually the CDS project, especially if there are high priority samples to analyze. Therefore, it is the role of management to ensure that the project is resourced adequately. The selection of the people to work on the project is a critical success factor and requires experienced chromatographers who, in my view, should work on the project full time if an aggressive three-month time schedule is to be met. If these chromatographers have experience using the selected CDS, then so much the better, as these people can provide other project team members who have little or no experience with help. Depending upon the size of the system to be implemented and the number of users, this can vary, typically between two and seven members.  

Always remember that a CDS project requires a multidisciplinary approach: information technology (IT) and quality assurance (QA) also are involved and the laboratory on their own cannot deliver a successful CDS validation project. QA and IT individuals on the project will not be full time but need to be kept involved with the progress of the project, asked for input to key activities, and when documents will be available for comment and approval. 

CSV requires a lifecycle model and typically, the pharmaceutical industry uses a classical V model approach (5). However, the classical V model does not reflect what you will be doing on a CDS validation project as there is not extensive configuration or customization of the system. Therefore, for a typical CDS validation project you will not need to write a functional specification or a design specification as the systems do not require any custom code to perform their function most of what you need comes out of the box (for example, instrument control, data acquisition, and data processing). What you do need to document is the configuration of the software: how you want the application to work with electronic signatures, user account management (user types and access privileges), custom calculations, and reports.  

Therefore, a simplified lifecycle model is essential to reduce the number of documents needed in the validation. Figure 1 shows a simplified lifecycle model for a CDS — three of the documents shown you may not be familiar with:    

Figure 1: A lifecycle model for a chromatography data system. 

This is a simple document that is essential early in the project as it defines implementation and training needs as well as an input into the technical specification written by IT. The implementation plan will document which instruments will be interfaced in the core system, where they are located, and where the data servers will be located. By implication it needs to identify, down to the network socket level, where things will go and how they will be connected together for the laboratory side of the project. Also listed will be the users in the laboratory together with their anticipated user type. In this way, training requirements will be documented.

How the CDS software will be set up, policies for using electronic signatures, user types and their access privileges, and so forth. I have shown this as a separate document but the contents could be included in the URS and often are.

This is an IT document that shows how the computer elements of the CDS will be built: size of server, disk layout, operating system to be installed with service pack and patches, the database, and the CDS software. There will be a large vendor input to this document as they know what is needed at the design level for performance and data security and integrity of the system.

Further easing of the way a system is validated is that some tasks can be adequately controlled though procedures. Custom reports and custom calculations are a case in point. These can be procedurally controlled as they will be developed long after the whole system havsbeen validated. However, the SOP needs to include the specification and testing of each report or calculation in enough detail so that it can be tested and then integrated into the CDS so that the system retains its validation status.  

As mentioned previously, this is a critical success factor that will determine (non)progress of the whole project. We have looked at the general user base here, but I want to focus on a specific group. Project team members from the laboratory will likely become local application administrators for the laboratory and will be the first line for resolving problems with the system (for example, power or super users in some terminology). Therefore, they need to have a good technical understanding of the application to help specify, configure, and test the CDS. Even if they are experienced users at the chromatographic level, further training will be needed to understand the technical level of the CDS to ensure that they can contribute fully to the project and beyond. If there is a time problem, then outsourcing of custom calculations or laboratory reports to the CDS vendor could be one option to keep the project on time. 

The role of QA often is maligned and misunderstood. They are the guardians of quality but often are seen as ultraconservative. However, you must have QA on your side if you want to get the CDS validation project though in tight timescales. Key project documents must be reviewed and approved by QA to see that they meet company and regulatory requirements. If you want to modify the validation approach and omit one or two stages of work that would normally be done, how would you approach QA if there was an antagonistic atmosphere? 

You must get QA on-side with the project and communicate and explain why you want to do tasks in a way that is different from a typical validation project. This is especially important at the start of the project when the initial planning takes place to define both the overall timelines but also when key documents are expected to be available for review. If you don't inform QA that a document is to be reviewed around a certain date, don't be surprised if you are told the document goes to the back of the queue. If the project wants to do something different to the normal computerized system validation procedure, you need to keep QA informed. Working with QA is infinitely better than working against them. 

6. Effective and Compliant IT Participation: 

The words IT and compliant in the same sentence often can seem strange and can be met with skepticism, but this is an essential part of the CDS validation project as the software has to run on the network operated by IT. Note also that there is the word effective in the CSF. It's no good being compliant if the network is slow or the hardware is delivered late or is unreliable. Therefore, the project needs to have technical input from IT on the design phases of the project, to carry out the installation and qualification of the servers, operating system, and the database but also operate the system when the CDS is released for operational use. 

If the CDS has over 25-or-so users then the technical architecture could benefit from the use of terminal emulation. Instead of installing the CDS application software on each PC in the laboratory and qualifying it, the software is installed once on a terminal emulation server (typically running Citrix) and is qualified once. Users in the laboratory will then log onto the system via an applet running on their PCs but apart from that, there is little difference in the way that they interact with the CDS. 

The benefit from using terminal emulation from the project's perspective is that there is a single software installation followed by a single installation qualification and operational qualification. For a core system this might not have much of an impact but if the overall installation is greater that 50, then the use of terminal emulation will save much time and effort. The other side of the coin is that the IT department needs to have the expertise to install and operate a terminal emulation system. The alternative is the classical client–server installation and qualification of the CDS application on all laboratory PCs described previously.  

The vendor of the CDS software has a role to play in the project from provision of technical advice (server and disk size for the laboratory), training (both regular users but also administrators), quality system development and support of the CDS application, supply of qualification documentation, and professional services. Training has been discussed in earlier CSFs here, and technical advice is an area where there can be very useful input from the CDS vendor in the technical architecture that is needed to run their product efficiently. However, I would like to focus here on two aspects, first, the quality development and support of the CDS application and second, the supply of qualification documentation for the project. 

During the development of the software, the vendor will test the software many, many times. If these functions are not configured when the software is installed, why do you need to retest them? However, the problem is how to justify a reduced testing approach? This is where a vendor can help the validation project. If a summary of the in-house testing of the CDS could be available to the project, it would provide the documented evidence to justify a reduced testing approach in some areas. Note that this is not a carte blanche to not test anything, as the project still will have to demonstrate fitness for purpose in the laboratory environment. 

The other specific area in which a vendor can be useful is the provision of qualification documentation for the CDS. This is a double-edged sword and it really is an issue of caveat emptor (buyer beware). There are very specific regulatory requirements in U.S. and EU GMP (6,7) for review and approval of documents used in any validation and use of vendor-supplied documentation could mean that the laboratory generates noncompliances. Why, you may ask? The reason is that the laboratory must review and authorize qualification documentation before it is executed but often, there is not any space for you to do this. The vendor's engineer pops into the laboratory, executes the protocols, and leaves them for one of the project team to review and approve. This is wrong and vendors must be more responsible and professional in this area. There will be a longer and more detailed discussion on this subject in a forthcoming "Focus on Quality" column in 

To cement all the different disciplines and tasks together it is essential to have effective project planning. The project needs to be broken down into the component tasks that, when completed, will deliver the validated core system. One problem in project planning is to break the project down into enough detail but not too much or too little. For each task, allocate the individuals who will be responsible for each one — for example, for writing the URS there would be a responsible individual who would need to coordinate the writing of the document and liaison with IT and QA to ensure that the inputs from these groups are elicited. Also, if there is a document required to support the validation, the review personnel need to be known and the approximate time when the draft will be available so that individual's work can be synchronized. 

Often, a project's plans can be compared with major works of fiction, so to avoid this label, project plans need to be refined and updated to allow for change time or the addition or elimination of tasks. The timetable for the validation of the core system will be aggressive but the allocation of time for each one must be realistic. 

Once the overall system has been specified in the URS, the project team then needs to concentrate on the validation of the core system to ensure that the timescales of the project can be achieved. Otherwise, looking at the core as well as the whole system will be a major cause of delay in delivering the core system. 

Finesse and subtlety are characteristics not often associated with computerized system validation. However, it is in the design of the overall test suite for the performance qualification (PQ) or user-acceptance testing that they need to come to the fore. Shown in Figure 2 is an outline design of a test suite for a CDS. This is for the core system validation and does not represent all of the testing that will take place but it will serve to illustrate the points I wish to make. The TS numbers in the figure refer to individual test scripts.  

Figure 2: Outline design of a phase 1 performance qualification test suite for a CDS.

There are two levels of testing to note. First, there is the system level testing and second, the analytical level testing based around the project-based organization of work. 

System level testing focuses on three areas: security and access control (TS01), data server reboot (TS02), and creation of new projects (TS03). All three of these test scripts are designed with two criteria in mind: first to test the function (which will test the some of the URS requirements), and second to test to trigger entries into the system audit trail. These entries in the system audit trail will be picked up and verified in TS02. 

Similarly, at the project level, there are five analytical test scripts (TS06, TS07, TS08, TS09, and TS13) checking the different calibration methods used by the laboratory. From the arrows linking many of these test scripts, you can see that there are interrelationships between many of the them. For example, TS06 tests the calibration model of average by amount with an external standard and includes capacity testing of the largest numbers of sample numbers expected during an analytical run. This test script will be run concurrently with TS05 looking at buffering of chromatographic data by the data server. After the TS06 is started, the network connection to the data server will be disconnected. The data server will buffer the data to the end of the run and then after the network connection is restored, the data will be transferred to the server. After these test scripts are completed, the system suitability calculations will be verified to comply with 211.68(b) (8), European Pharmacopoeia chapter 2.2.46 (9), and United States Pharmacopoeia <621> (10). Furthermore, the project and data generated in TS06 is used in the test script for project, archived, and restored (TS14). To save time and effort, the data generated for TS09 (multicomponent analysis) will be used for the area percent calculations (TS13) and avoid the need to undertake a second analytical result. Similar to the system-level testing of the system audit trail in TS04, the project audit trail will be tested using audit trail entries in a number of test scripts (TS07, TS11, TS12, and TS13). 

In this article, I have asked the question: Why does it take so long to validate a CDS? In answering this, I have provided 10 critical success factors that should be considered when establishing a CDS validation project. Ideally, the core system should be validated within three months. 

 is principal of McDowall Consulting, director of R.D. McDowall Limited, and "Focus on Quality" column editor for 

s sister magazine. Address correspondence to him at 73 Murray Avenue, Bromley, Kent, BR1 3DJ, UK, or e-mail: 

(1) FDA Guidance for Industry, General Principles of Software Validation, 2002. 

Article

The author examines the process of validating a CDS and describes the trials along the way such as speed of validation, risk management, and so forth.

Why Does It Take So Long To Validate a Chromatography Data System?

The ever-increasing cost of healthcare and improvements in our understanding of the various factors that play a role in the manufacturing of biotech therapeutic drugs have fueled the push for approval of biosimilars (referred to as follow-on protein products in the U.S.). Discussion and resolution of the various scientific and regulatory factors behind the approval of biosimilars is perhaps one of the most significant events in the last decade for biotechnology. Reasons for this are many. For one, biotech processes and products are complex, and while the industry has made significant improvements in our understanding of how the process impacts the product, a lot more must occur. The use of complex feed materials that sometimes have not been fully characterized can further exacerbate this balance. Further, our understanding of how the product quality attributes impact the safety and efficacy of the product in the clinic is incomplete. This can be attributed to the complexities of the human body and is confirmed by the countless examples in the literature of observing significant changes in clinical performance (often undesirable) resulting from otherwise minor changes to the manufacturing procedures. In view of the difficult job that the regulatory agencies have in balancing the financial benefits with public safety, the European Medicinal Agency (EMEA) has done a commendable job at creating the regulatory path to facilitate approval of biosimilars. The U.S. Food and Drug Administration (FDA) is expected to do the same in near future.  

In this column, we wish to provide our readers with an introduction to the topic of biosimilars, review the available regulatory guidance, discuss the various scientific challenges that are present, and talk about the opportunity that this provides to the drug manufacturers, regulators and the patients.  

Biosimilars can be defined as biotech drugs that have been shown to have comparable quality, safety, and efficacy to the original product. There are some key differences between the production of biosimilars and that of the traditional small molecule generics: 

 Cost of manufacturing a biotech product is approximately 5–10 times more than that for a small molecule product. Capital investments as well as operating costs associated with manufacturing of biosimilars are also significantly higher than that for small molecule generics along with the associated risk of failure.  

Data on the launch of biosimilar epoteins in Germany shows that biosimilars represent approximately 35% of the total sales of epoetin (4). This is in comparison to much higher substitution rates than are seen for generic small molecule drugs (can exceed 90%). 

Biosimilars are expected to be priced at a 20–25% discount in comparison to the original products. This discount is significantly smaller than what is seen in the small molecule generics and reflects the intricacies associated with production and characterization of biotech products.  

 The impact of changes to manufacturing processes can cause significant change in efficacy or immunogenicity of the drug in the clinic. 

 Biosimilars are larger, more complex molecules with associated structural heterogeneities (21). These heterogeneities are a result of activities in the host cell that are used to express the product, the bioreactor conditions under which the cells are grown, and the purification process utilized for generating the final product. Biotech products cannot be completely characterized analytically; that is, the analytical tools alone cannot predict behavior in the clinic. 

 Feed materials used in manufacturing of biotech products tend to be complex, sometimes are not characterized, and sometimes are known to impact process consistency and product quality significantly. 

Lack of understanding of relationship between product and process:

 The correlations between the clinical safety and efficacy of a biotech product and its product quality attributes generally are not known completely. So far, the clinical data have been mandatory so as to enable evaluation of safety and efficacy of the biosimilar drug in comparison to the original drug.  

Healthcare is already one of the largest expenses for developed societies (U.S., Europe, and Japan) and is slowly becoming a concern for emerging economies as well. While the cost of biotech therapies is at present a relatively small portion of the total healthcare expense in the U.S. (which is by far the largest market for biotech drugs), this is expected to grow to about 30% (a 20-fold increase) in the next 10 years (8). This is likely to continue the pressure on the FDA to devise the regulatory pathway that would allow for review and approval of biosimilars. One study has suggested that as much as $2 billion can be saved by the European healthcare providers by just the first wave of biosimilar products (17). The regulatory authorities in Europe have already approved biosimilars for recombinant somatropin, recombinant human epoetin alfa, and recombinant filgrastim.  

In a recent article, the authors concluded that as the analytical methodology becomes increasingly sophisticated, it is likely that manufacturers will consider biosimilar monoclonal antibody products (30). However, the authors also cautioned that the comparative trials that might be necessary to establish efficacy and safety of the biosimilar product versus the innovator product could be prohibitively large, making the idea of producing biosimilar monoclonal antibody products not feasible. The launch of Reditux (biosimilar of Genentech's Rituxan) by Dr. Reddy's laboratory in India seems to suggest that the possibilities do exist (6

The European regulatory authorities already have established the framework for approval of follow-on protein products and have issued several key guidelines in this regards. The "Guideline on Similar Biological Medicinal Products" states that comparability studies are needed to establish similarity with respect to quality, safety, and efficacy of the biosimilar molecule with the innovator product (9). The "Guideline on Similar Biological Medicinal Products Containing Biotechnology-Derived Proteins as Active Substance: Quality Issues" lays down the quality requirements for a biosimilar product and addresses requirements regarding manufacturing process, comparability exercise, and choice of reference product, analytical methods, physicochemical characterization, biological activity, purity, and specifications of the biosimilar product (10). "The Guideline on Similar Biological Medicinal Products Containing Biotechnology-Derived Proteins as Active Substance: Non-Clinical and Clinical Issues" lays down the relevant requirements for a biosimilar product and addresses requirements regarding pharmacokinetic, pharmacodynamic, efficacy studies, as well as the risk management plan (11). Finally, the "Annexes to Guideline on Similar Biological Medicinal Products Containing Biotechnology-Derived Proteins as Active Substance: Non-Clinical and Clinical Issues – Guidance on Similar Medicinal Products Containing Somatropin, Recombinant Erythropoietins, Recombinant Granulocyte-Colony Stimulating Factor, Recombinant Human Soluble Insulin" provides guidance on nonclinical and clinical requirements for the specific products (1–4). "The Guideline on Immunogenicity Assessment of Biotechnology-Derived Therapeutic Proteins" states that immunogenicity evaluation should include data systematically collected from sufficient number of patients with the sampling schedule determined on a case-by-case basis, taking into account the risks associated with unwanted immune response to patient (12). "The Note for Guidance on Planning Pharmacovigilance Activities" recommends that a Safety Specification and Pharmacovigilance Plan be submitted at the time of license application that outlines the design and conduct of observational studies (16). 

The FDA's scientific reasoning and experience in the area of follow-on protein products has been published recently via examples of FDA's actions involving the evaluation of various types of follow-on and second generation protein products and within-product manufacturing changes (29). A regulatory path for approval of follow-on protein products in the U.S. has yet to be proposed with the FDA Policy on Coordinated Framework for Regulation of Biotechnology governing the review and approval process of new protein products at present (FDA Policy).  

Other organizations such as the World Health Organization (WHO) are also in the process of issuing guidelines that provide globally acceptable principles for licensing biotherapeutic products that are claimed to be similar to biotherapeutic products of assured quality, safety, and efficacy that have been licensed based on a full licensing dossier (28).  

The two major concerns with respect to approval and use of biosimilars are the efficacy and safety of the biosimilars in comparison to the original drug. In this section, we briefly present some of the key aspects that impact process and product comparability (7). 

Relationship between the process and the product:

 Biotech products are known to be very sensitive to changes in manufacturing process. It is often said that "the process defines the product." An example of how changes in manufacturing process can impact the drug's clinical performance is the case of commercialization of interferon-b-1a for multiple sclerosis (24). The clinical material for Phase III study and the commercial material that had prior marketing approval from EMEA were made using the same gene, formulation, and storage conditions. They were delivered with the same dose, route, and frequency; and they were shown to have in vitro and in vivo comparability. However, because the two materials were made at two different sites with two different manufacturing processes, the Phase III material resulted in five times higher incidences of neutralizing antibodies than the material on the market. 

Figure 1: Comparison of activity index for 11 non-innovator epoetin products available outside of US and Europe. Bioactivity was determined via an in vivo bioassay in mice. Activity index was calculated as the ratio of actual bioactivity divided by the bioactivity listed on the product. (Adapted from reference 25.) 

In view of this, the biosimilar manufacturer has to demonstrate its capability to control the process so as to manufacture product of consistent quality (5). This is followed up by a thorough comparison between the biosimilar and the innovator's product based upon thorough analytical examination (via a collection of high-resolution, orthogonal techniques), stability studies, nonclinical studies (such as receptor-binding studies and cell-based assays), and clinical studies (for pharmacokinetic, pharmacodynamic, and immunogenic behavior) (22). In all likelihood, the biosimilar product will differ from the innovator's product in a subsection of the quality attributes. As the effect of these differences on the clinical performance is not known completely, clinical safety needs to be monitored not only before giving approval but also postapproval. This is best illustrated by a comparison provided in Figure 1 that shows the activity index of 11 epoetins compared to Eprex (25). The activity index was calculated as the ratio of the bioactivity as determined by in vivo bioassay in mice and that in the specifications of the drug. A couple of observations can be made from this data. First, the activity index varies widely (0.71–2.26). Second, if we compare the variability to the conventional benchmarks (0.8–1.2), we find that two of the products (in red) seem to have less bioactivity than the lower end of the benchmark and four of the products (in dark blue) have more bioactivity than the upper end of the benchmark. It is evident from this data that for biosimilars to be approved successfully, the manufacturers will need to improve product quality. A similar comparison, also on epoetins, has been made (19) and is shown in Figure 2. The observations are similar to that in Figure 1 as some of the samples tend to fail the specifications for product concentration (80–120%, Figure 2a), pH (6.6–7.2, Figure 2b), and osmolarity (200–300 mOsm/kg, Figure 2c).  

Figure 2: Comparison of (a) measured concentration, (b) pH, and (c) osmolarity of different rHuEPO samples from Asian market. (Adapted from reference 19.) 

It can be inferred from the previous examples that to have consistent product quality, biosimilar manufacturers need to understand their processes and the interrelationships between the process parameters and product quality attributes (13,31). These principles also have been outlined in the emerging initiative of Quality by Design (15,20,21). However, biotech processes are complex and gaining this process understanding is a challenge due to the large number of process parameters and the quality attributes that can be interrelated (22). Figure 3a illustrates that as the number of factors being examined increases, the number of experiments required (to fully understand the impact of each factor and the interactions between them) increases much more rapidly, and this very soon becomes too large to conduct in a time- and resource-efficient manner. Further, Figure 3b illustrates the case of a typical process chromatography step that is used commonly in purification of protein products. Such a step can use >10 different raw materials and have >10 input parameters that can be altered during performance of the step. With >20 possible variables that can impact the performance for such a step, it is evident from Figures 3a and 3b that it is not possible to understand completely the impact of each raw material or input parameter on the performance of the unit operation. Use of risk assessment tools such as "Failure Mode and Effects Analysis" is used commonly to prioritize these variables into a reasonable number that can be examined via small-scale studies (13,26,31). High-throughput process development tools also are useful in achieving this objective in a time- and resource-efficient manner. 

Figure 3: (a) Illustration showing the exponential increase in the number of experiments required with the number of factors that are being investigated in a full-factorial design of experiments (DOE) study. (b) Number of steps, raw materials, input parameters, and output parameters for a typical process chromatography step. The figure illustrates the complexity of unit operations utilized in manufacturing of protein products. 

Relationship between the product and the clinic: 

Biotech products tend to be complex and subject to a lot of potential modifications, as listed in Table I (22,27). These include: desired process modifications such as pegylation; desired host cell modifications such as glycosylation, phosphorylation, sulfation, and acylation; undesirable host cell-related modifications such as truncation, acylation, glycation, and methylation; and undesirable process modifications such as aggregation, oxidation, and deamidation. Each of these modifications potentially can impact safety and efficacy of the product. In addition to the product-related modifications, there are numerous host cell-related impurities (host cell proteins, DNA, viruses, and so forth) and process-related impurities (antifoam, extractables, leachables, and so forth) that can be present in the final product. Glycosylation of proteins has been known to impact the drug's efficacy and stability, as well as the immunogenicity. Changes in degradation of the drug can produce novel antigenic epitopes, which in turn can result in changes to immunogenicity and metabolic half-life. The glycosylation pattern is a function of the host cell (microbial, insect, mammalian, and so forth) and the operating conditions of the bioreactor. Understanding how aggregate species interact with the immune system to enhance immune responses has been the focus of a recent review (23). The author suggests that the highly arrayed protein antigens, such as those present in nondenatured aggregate species, are capable of inducing antibody responses even in the absence of any help from the T-cells. Use of animal models and studies in understanding the effects of protein aggregates also was discussed. Host cell impurities, such as host cell proteins, DNA, and endotoxin can impact immunogenicity of the drug significantly. Endotoxins have been known to signal T-cells, resulting in activation of the B-cells and an immune response to the drug. 

Table I: Illustration of the various quality attributes (QA) that need consideration during production of a biotech product and the various analytical tools that are used for characterization of the QA. The list is not intended to be exhaustive and is rather being presented to highlight the diversity of attributes and analytical methods. Adapted from 22.  

Also shown in Table I is the gamut of high-resolution, orthogonal analytical tools that often are utilized to characterize some of these species.  For example, protein aggregates often are considered a critical quality attribute due to their potential impact on immunogenicity of the drug, and these are characterized by multiple analytical techniques including capillary electrophoresis (CE), high performance liquid chromatography (HPLC), analytical ultracentrifugation (AUC), and light scattering. This enables a thorough comparison of the biochemical characteristics of the follow-on protein product versus the reference product.  

Our lack of understanding of how the product quality attributes impact a biotech drug's safety and efficacy in the clinic is best illustrated by the case of Eprex. Use of this product for treatment of anaemia in patients with renal dysfunction has been associated with a higher incidence of pure red cell aplasia (PRCA), which is a severe form of anemia accompanied by near stoppage of production of the red blood cells. While it is not clear what triggered PRCA in this case, the two most common hypotheses include the change in formulation (from human serum albumin to polysorbate 80) and the presence of leachates from the uncoated rubber syringe stoppers into the drug (7). This example illustrates our limited understanding of biotech products with respect to which attributes drive their safety and efficacy.  

With all the challenges that come with manufacturing and commercialization of biosimilars, the very significant cost savings that are likely to emerge from their approval is likely to serve as the motivation for the regulatory authorities to create a path that allows for examination and approval of biosimilars (in the U.S.) or continue on the path that has been laid out already (in Europe).  The regulators have the daunting task of keeping the balance between the financial benefit from allowing approval of biosimilars and the safety of the patients. It is expected that while the topic of approval of biosimilars will stay controversial, decisions will need to be taken on a case-by-case basis and would depend upon a multitude of factors, as illustrated in Figure 4. The more we understand the starting materials and how they impact the performance of the process, the critical process parameters (CPP) of the process that impact the quality attributes of the product, and the critical quality attributes of the product that impact safety and efficacy in the clinic, the less we will rely on the clinical studies (assuming that comparability with respect to the relevant attributes has been demonstrated). It should be evident from the discussion here that generic approaches that have been taken for small molecules are not sufficient for protein products, even more so for complex biotech products, such as monoclonal antibodies. 

Figure 4: Illustration of some of the significant factors that come into play when comparing a follow-on protein product to the parent molecule. The two extremes are shown in color. Green represents the case of minor process changes (improvements) that may be undertaken by the sponsor of an already licensed product. If the scope of the changes is sufficiently small, no clinical studies may be required. Red represents the case of filing a new biologic product with no predecessor, and thus, requires a full clinical dataset to support the drug’s safety and efficacy. Follow-on protein product applications will be in the middle of these two extremes with the answers to the presented questions driving the expectations from the regulatory agencies.  

A recent paper has proposed some solutions for the aforementioned challenges (22). Postapproval pharmacovigilance will continue to play a critical role, as the clinical database is likely to be limited for biosimilars at the time of approval. In the preapproval period, only a few patients might receive the product in a relatively limited time period compared to the number of patients that will use the product during its complete lifecycle. The postapproval pharmacovigilance program needs to address the potential risks that often are associated with biotech drugs, especially immunogenicity. EMEA provides specific guidance on this subject (16), and often, these plans are part of the postapproval commitments that the biosimilar manufacturer makes to the regulatory agency. 

"Validation Viewpoint" Co-Editor Ira S. Krull is an Associate Professor of chemistry at Northeastern University, Boston, Massachusetts, and a member of LCGC's editorial advisory board.

is a biotech CMC consultant and an associate professor with the Department of Chemical Engineering,  Indian Institute of Delhi, India. He is also a member of 

(1) Annex Guideline on Similar Biological Medicinal Products Containing Biotechnology-Derived Proteins as Active Substance: Non Clinical and Clinical Issues – Guidance on Similar Medicinal Products Containing Somatropin. Committee for Medical Products for Human Use. EMEA/CHMP/94528/05 (CHMP adopted February 2006). 

(2) Annex Guideline on Similar Biological Medicinal Products Containing Biotechnology-Derived Proteins as Active Substance: Non Clinical and Clinical Issues – Guidance on Similar Medicinal Products Containing Recombinant Erythropoietins. Committee for Medical Products for Human Use. EMEA/CHMP/BMWP/94526/2005 (CHMP adopted March 2006). 

(3) Annex Guideline on Similar Biological Medicinal Products Containing Biotechnology-Derived Proteins as Active Substance: Non Clinical and Clinical Issues – Guidance on Similar Medicinal Products Containing Recombinant Granulocyte-Colony Stimulating Factor. Committee for Medical Products for Human Use. EMEA/CHMP/BMWP/31329/2005 (CHMP adopted February 2006). 

(4) Annex Guideline on Similar Biological Medicinal Products Containing Biotechnology-Derived Proteins as Active Substance: Non Clinical and Clinical Issues – Guidance on Similar Medicinal Products Containing Recombinant Human Soluble Insulin. Committee for Medical Products for Human Use. EMEA/CHMP/BMWP/32775/2005 (CHMP adopted February 2006). 

http://www.drreddys.com/products/popups/reditux.htm

(9) Guideline on Similar Biological Medicinal Products. Committee for Medical Products for Human Use. CHMP/437/04 (CHMP adopted September, 2005). 

(10) Guideline on Similar Biological Medicinal Products Containing Biotechnology-Derived Proteins as Active Substance: Quality Issues. Committee for Medical Products for Human Use. EMEA/CHMP/BWP/49348/05 (CHMP adopted February, 2006). 

(11) Guideline on Similar Biological Medicinal Products Containing Biotechnology-Derived Proteins as Active Substance: Non-Clinical and Clinical Issues. Committee for Medical Products for Human Use. EMEA/CHMP/42832/05 (CHMP adopted February, 2006). 

(12) Guideline on Immunogenicity Assessment of Biotechnology-Derived Therapeutic Proteins. Committee for Medical Products for Human Use. EMEA/CHMP/BMWP/14327/2006 (CHMP adopted December, 2007). 

Biotechnology Today

The authors provide an introduction to the topic of biosimilars, review the available regulatory guidance, and discuss the various scientific challenges that are present.

Biosimilars: Introduction, Concerns, and Opportunities

I wrote my first chromatography paper in 1978, shortly after transferring from the University of Georgia to the University of Minnesota. This work (1) concerned the optimization of gradient elution by use of the Simplex algorithm popularized by Deming and Morgan 

2). Before the move to Minnesota, my research had focused on biochemical applications of electroanalytical and thermoanalytical methods, especially the use of immobilized enzymes. There are many commonalities between separation science and immobilized enzymes, and it was in this context that I first became acquainted with Csaba Horvath's work. In fact, I had been unaware of his seminal role in the development of high performance liquid chromatography (HPLC).  

I owe a great deal to my mentors in electroanalytical and thermoanalytical methods, including my Ph.D. advisor, Joe Jordan at Penn State University, and Steve Feldberg at Brookhaven National Laboratories, but especially Lou Meites at Brooklyn Poly (now the Polytechnic Institute of New York). 

The title of this article is borrowed, with thanks, from remarks made by Peter Uden of the University of Massachusetts in his address at the Eastern Analytical Symposium (EAS) several years ago when he received the EAS Award in Chromatography. He looked over the audience and welcomed a group of "recovering electrochemists." I just know he was looking at me in particular. Actually, I took this as a compliment. I have always said that electrochemistry is great training for separations scientists, as it involves many concepts related to mass transport and phase equilibria, both being subjects in which I had been well instructed and that facilitated my transition to chemical separations. 

I do recall my earliest encounters with chromatography. In my sophomore organic laboratory course, we were taught how to assemble a classical chromatography column by slurry packing a glass column about ½ in. i.d. by 12 in. long and separating some colored product from "gunk." It was great to see a colored — albeit very irregular — band move down the column. But I can't say I remotely understood what was going on. The lab focused on technique, not theory. Certainly I had never heard of terms like "HETP" and "isotherm," and knew nothing about the van Deemter equation. (I would like to think this is no longer true for many "users" of chromatography, but sadly I feel this is still too much the case. Chromatography is far too often used as a "black box.") 

During my junior year, I had a course called "Advanced Analytical Chemistry," which would now be referred to as "Instrumental Analysis." We used Meites and Thomas' wonderful textbook, 70% of which was electroanalytical chemistry, 25% spectroscopy (IR, UV but no nuclear magnetic resonance [NMR], atomic absorption [AA], or mass spectrometry [MS]), with one barely intelligible chapter on chromatography in which the de Vault equation was derived. Little did I realize that this equation was utterly fundamental to chromatography.  

I got a good bit deeper into the theory of column dynamics when, some years later, Steve Weber of the University of Pittsburgh asked me to help with some fine points of the de Vault equation and its consequences for nonlinear chromatography in his wonderful chapter on the dynamics of liquid chromatography in Brown and Hartwick's book on HPLC (3). However, in my undergraduate course on instrumental analysis there were no lectures on chromatography and certainly no laboratory experiments.  

Instruction in this field has surely come a long way. The last time I taught a course on Instrumental Analysis for our senior undergrad chemistry majors, we spent half of the 14 weeks on gas chroamtography (GC), HPLC, and capillary zone electrophoresis (CZE). (When I hear that our undergraduate curriculum in analytical has not changed, I want to ask, "Since when—last year?") The rest of the course had a fair amount of electronics and computers, atomic spectroscopies, FT-IR, and only five or six lectures (two weeks) on electrochemistry.   

My next contact with chromatography was very casual. Brooklyn Poly was housed principally in a converted razor blade factory in downtown Brooklyn. One day between classes I saw a group of graduate students in the stairwell performing a strange rite involving a narrow copper tube that ran from the eighth floor to the basement. One of them had been my instructor, so I asked him what the "bleep" they were doing. He explained that they had a difficult separation to do and so were packing a 100-foot GC column (this was clearly well before the era of fused-silica capillaries). Packing required a team of nearly a dozen. One guy stood on each floor to tap the side of the tube to keep the particles flowing and settle the bed, and two were at the top to add the diatomaceous earth and support the tube. Assuming a 100-ft column and some reasonable estimates of the height equivalent to a theoretical plate (HETP), they might have generated 10,000–12,000 plates, a trivial accomplishment by today's standards. 

Other than a seminar by Cal Giddings, I had little additional exposure to chromatography in graduate school at Penn State, although one of my predecessors in Joe Jordan's laboratory had studied thermal conductivity detection in GC (principally because Jordan was interested in thermistors as temperature sensors). I also recall a grad student colleague who ran into trouble reproducing a GC separation he had done using a well-known laundry powder as the stationary phase. It turned out the detergent's manufacturer had changed the formulation, so he was never able to do the separation again.  

After graduate school and a brief postdoc at Stanford Medical School, in 1969 I joined a marvelous group at the University of Georgia that included Don Leyden, Dave Hercules, and Leon Klatt, with Rudi Seitz joining a bit later. All have proven to be lifelong friends, wonderful colleagues, and mentors. I learned a great deal about many things from all of them. But not one of us knew a thing about chromatography. We tried and failed to hire Milos Novotny in the very early 1970s. Shortly after this great disappointment, "Buck" Rogers happened to be an analytical seminar speaker at Georgia. In private conversations with Buck, we all asked more or less the same question: "Who should we try and hire?" A few days later we met for our weekly lunch, looking quizzically at one another and asking "Lloyd who?" revealing a mutual, stupendous ignorance of Lloyd Snyder's already significant contributions to chromatography. At that time, Lloyd was working for Technicon, and within two years I was recommending a great student, Larry Bowers, to postdoc with Lloyd. Larry's interests in clinical chemistry pulled him toward a more academic position. In the years since, Lloyd and I have published jointly many times and frequently act as sounding boards for one another's ideas and papers. Lloyd has been a most valued collaborator and has contributed immeasurably to my work.  

My real introduction to chromatography came when my Georgia colleagues and I decided that we simply had to offer a chromatography course for our graduate students — and it fell to me to develop one. After surveying the literature (and nascent state) of LC in 1972, I found Littlewoods' marvelous book on GC. I taught a basic course out of that monograph, augmenting it liberally from Dean's book on chemical separations, the chapter on separations from Laitinen's book 

 and a few chapters from Sidney Siggia's 

I had relied heavily upon Siggia's book on quantitative organic analysis as a graduate student, and he was a personal hero of mine. I still recall my first and only meeting with Siggia at a great talk he gave at Georgia. In the middle of this talk, he illustrated dramatically his emphasis on "problem solving" in analytical chemistry by telling of an industrial colleague who came to beg for his help one day. "Why," this overwrought chemist wondered, "is my catalyst dead?" What struck Siggia was that he, an analytical chemist, was being viewed as a team member and problem solver rather than as a technician who merely responds to the demand "How much XYZ is in this sample?" To some extent, he eschewed academic analytical research that is often purely "theoretical." Instead, Siggia's approach was very much in the tradition of Lundell's classic paper "The Chemical Analysis of Things as They Are" (4). As a brief perusal of any analytical chemistry journal will show, even though the technological basis of this paper is entirely outdated, his philosophical points are as valid today as they were 60 years ago when the paper was written. 

The first Nobel laureate I had contact with was Archer Martin. I think I was in my second year at Georgia, and Joe Jordan, who had organized a symposium on thermochemical analysis for an American Chemical Society (ACS) meeting, invited me to give a presentation. I vividly recall an absolutely insightful and fundamentally important question from an elderly gentleman in the audience. I responded, and he followed up with a question that revealed (I thought) a profound ignorance of the subject. I responded again, perhaps a bit cavalierly. I then sat down next to Joe, who whispered in my ear, "You don't know who that was, do you?" Of course I said, "No." Joe said, "That's Archer Martin!" I think he consciously (and conscientiously) deleted the comment "you jerk."  

A few years later, I had several offers to leave Georgia. But I had a distinct leaning toward Minnesota. It was the home of I.M. (Piet) Kolthoff, who was, by any measure, the founder of scientifically based analytical chemistry in the U.S. Also, I was very impressed with Paul Gassman, then chair of the chemistry department and subsequently president of the ACS. But as I think my students would hasten to point out, Minnesota also had the best sports fishing of any state where I had an offer (see Figure 1). 

Figure 1: Fall fishing at St. Cloud Minnesota with daughter Kelly in the background. 

Shortly after starting at Minnesota in 1977, I decided that it would be very helpful if there were an organization in the area similar to the Chromatography Forum of the Delaware Valley. Based upon discussions with several local sales representatives, Larry Bowers (my former graduate student at Georgia who was already on the faculty of Minnesota's Medical School) and I identified a group of local chromatography "gurus." These included Prof. Mark Brenner of the Department of Horticulture (afterwards dean of the graduate school at Minnesota); Don Hagen of 3M, a real leader in the area and subsequently a great friend; Larry Bell, an independent tech representative; and a number of others. We met at a bar near campus called the "Improper Fraction" on St. Patrick's Day 1978 (see Figure 2). The Minnesota Chromatography Forum (MCF) was the result of that lunch meeting. And while the Improper Fraction is no longer with us, the MCF just celebrated its 30th anniversary (Figure 3). 

Figure 2: First Minnesota Chromatography Forum Spring Symposium with Les Ettre, Walt Jennings, and founding fathers of MCF. 

One of the MCF's earliest invited speakers was Les Ettre of Perkin-Elmer, a pioneer of capillary GC (Figure 2). He reminded us that a University of Minnesota faculty member, Leroy Sheldon Palmer, was the first American to use chromatography in his research work. So Les suggested that we institute an award in Palmer's name (Figure 3). 

Figure 3: Reunion of Palmer Award Winners 2010 (30th Anniversary). 

During a subsequent visit, Les wanted to document various aspects of Palmer's life. So Don Hagen and I took him to Palmer's grave site so he could photograph it. Unfortunately, the in-ground marker was covered with dirt and overgrown with grass. We cleaned it off and shined it up by pouring some of Don's coffee on it. It was surreal.  

Over the years, MCF meetings gave me the chance to get to know some of the really great contributors to chromatography, including Csaba Horvath, Georges Guiochon, Lloyd Snyder, Al Zlatkis, Jack Kirkland, Klaus Unger, and many others who would be very well known to the readers of 

One of the great experiences of my life was my involvement with HPLC 1994 which was held in the Twin Cities. Due to a heart attack in 1993, Larry Bowers, the co-organizer, did all of the heavy lifting and of organizing and running the  conference which despite a slow economy was very successful. He did a magnificent job. We both gained the experience of working with the best chromatographers and separation scientists in the world (Figure 4). 

Figure 4: Visitors during HPLC 1994 (Jack Kirkland and Csaba Horvath). 

As I said, my real introduction to chromatography was through my teaching, but when I shifted my research into chromatography, I spent a lot of time with Cal Giddings' 

(5), Lloyd Snyder's monograph on adsorption chromatography (6), and Karger, Horvath, and Snyder's famous 

. In terms of LC, Snyder and Kirkland's classic 

was essential, along with the five wonderful volumes of Horvath's 

Advances and Perspectives in High Performance Liquid Chromatography

. During my early days at Minnesota, my group had a series of all-day Saturday meetings in which we systematically reviewed Horvath's encyclopedic work. These weekly sessions were immediately christened "Csabathons" by an irrepressibly irreverent graduate student who shall remain nameless to protect her identity.  

 Over the years, our various LC instruments and computers have acquired relevant nicknames such as "Lloyd," "Csaba," "Barry," and "Georges" (though one year we had a rebellious student who named his instrument "Homer" after his favorite Simpsons character). 

I had always been a believer in reading broadly in the chemical literature, and this inevitably led me to Mort Kamlet and Bob Taft's work on the physical organic chemistry of solvation. It is self-evident that chromatographers should be interested in solvation processes. For me, the result was a major series of papers with Mort, Bob, and Mike Abraham combining chromatography with physical organic chemistry (7) and the chemical engineering of fluid phase equilibria (especially through the work on solubility parameters by Chuck Eckert of Georgia Tech).   

Without doubt, Lloyd Snyder and Csaba Horvath have had an immense impact on my career and work in chromatography. There are too many stories to tell about Lloyd and Csaba individually, so I will tell one involving both. 

Figure 5: Attendees at the Lloyd Snyder Celebration in 2001 prior to Maastrich meeting of HPLC 2001 organized by John Dorsey and Elsevier. 

chief U.S. editor for liquid chromatography. So when he retired shortly before the Maastricht meeting of the International Conference on Column Liquid Chromatography (2001), Elsevier and John Dorsey, the incoming editor, sponsored a symposium just before the ICLC to honor Lloyd (Figure 5). It was a marvelous event, and many of Lloyd's friends participated, including Csaba, whom I had the privilege of introducing. Now, only a month earlier I had introduced Csaba as the keynote speaker at the annual MCF spring symposium. So I already had a very elaborate introduction with slides detailing Csaba's many seminal contributions. But because everyone at Lloyd's symposium knew Csaba personally, I saw no point in a formal lengthy introduction. Instead, after lunch I stood up and said, "I have always wanted to do this." Then, mustering my best Ed McMahon interpretation, I shouted, "Heeeere's Csaba!" The Americans in the audience, including Csaba, got quite a kick out of my salute. However, at the break after Csaba's talk, Hans Poppe, not understanding my homage by comparing Csaba to Johnny Carson, started to reprove me for disrespecting Csaba. Fortunately, Peter Schoenmakers, who had worked for several years at Shell in the U.S. and was a close friend and associate of Hans', intervened and politely explained that I was indeed saluting Csaba and treating him as a first-name-only celebrity, not demeaning him. Subsequently Hans and his wife gave me a lift to Amsterdam. Lloyd and others warned me that Hans liked to drive fast. It was a hair-raising trip. 

Figure 6: 2D LC traces of various complex materials. 

In many ways, my current research in fast LC×LC is the outgrowth of an interest, born in the late 1980s, in stable stationary phases and a long-standing interest in chemical selectivity and phase transfer thermodynamics (see Figures 6 and 7). If, as someone once said, all chemists can be divided into two classes, thermodynamicists and kineticists, I am wholeheartedly and unabashedly a thermodynamicist. 

Figure 7: Pete Carr and Dwight Stoll examining a 2D LC system. 

Chromatography — the research and the interactions with the hundreds of colleagues, graduate students, and collaborators whom I have met because of it — has been immensely stimulating and, yes, fun. I hope my students and I have contributed something to both its theory and practice.  

 received his B.S. in Chemistry (1965) from the Polytechnic Institute of Brooklyn, where he worked with Professor Louis Meites, and a Ph.D. in Analytical Chemistry at Pennsylvania State University (1969) under the guidance of Professor Joseph Jordan. He was a research assistant and associate at Brookhaven National Laboratory during the summers of 1965 and 1966 and a postdoctoral associate with David Glick at Stanford University Medical School (1968). From 1969 until 1977, he was with the University of Georgia (Athens).

In 1977, he joined the University of Minnesota, where he became Professor of Chemistry in 1981. He has won numerous awards including the Eastern Analytical Award in Separation Science, the Dal Nogare Award of the Chromatography Forum of the Delaware Valley, and the ACS (Supelco) Award in Chromatography. Most recently he was awarded the Horvath Medal and the Martin Medal. 

(3) S.G. Weber and P.W. Carr, "The Theory of the Dynamics of Liquid Chromatography," in 

Chemical Analysis: A Series of Monographs on Analytical Chemistry and Its Applications

, Vol. 98, Ch. 1 (John Wiley and Sons, Inc., Hoboken, New Jersey, 1989), pp. 1–115. 

 (Analytical Edition), 5, 221–2251 (1933). 

 (Marcel Dekker, Inc., New York, New York, 1965). 

, (Marcel Dekker, Inc., New York, New York, 1965). 

(7) M.J. Kamlet, P.W. Carr, R.W. Taft, and M.H. Abraham, 

History of Chromatography

Well-known industry veteran and LCGC EAB member Peter Carr looks back at the early days of the chromatography field and reflects on his own experiences.

Early Adventures in Chromatography: The Story of a "Recovering" Electrochemist

Sample preparation is a key step in chemical analysis because it is, in some cases, the slowest and most complicated step overall. Sample preparation is necessary to extract, isolate, and concentrate the analytes of interest from complex matrices into a suitable solvent because most analytical instruments cannot handle the sample matrix directly. Therefore, sample preparation has a direct impact on accuracy, precision, and quantitation limits and is often the rate-determining step of the analytical process, especially when trace determination has been the purpose.  

For the isolation and concentration of analytes from aqueous matrices, the technique of liquid–liquid extraction (LLE) is still the most common extraction method even though it is time consuming and labor intensive. The technique generally is considered to be expensive since ultrapure solvents are necessary for trace analysis; it also is potentially hazardous, generates waste, and is nonenvironmentally benign and nonsustainable. Nevertheless, with improvements in analytical techniques, especially in detector sensitivity and specificity, even classical LLE has undergone a renaissance in that smaller samples can be used, which implies that extraction solvent volume can be reduced, sometimes to only a few milliliters. Now small-volume LLE can be fairly routine and can be automated.  

As sample sizes continue to get smaller, techniques that provide efficient extraction are needed. Thus, a series of reduced solvent and solventless miniaturized sample preparation techniques has been developed. Many of these techniques are based on a great reduction in the volume ratio of acceptor to donor phase. In this installment of "Sample Prep Perspectives," we shall look at some of these microextraction techniques and demonstrate how the application of hollow fibers can improve the extraction efficiency, experimental robustness, and allow automation. We will focus on direct immersion hollow-fiber techniques, in which the extracting solvent is in direct or indirect contact with an aqueous sample phase (donor).  

Brief Review of Microextraction Techniques

 The first successful modern microextraction technique was solid-phase microextraction (SPME) developed by the group of Pawliszyn at University of Waterloo, Ontario, Canada (1,2). This extraction technique has become the standard that other microextraction techniques have tried to emulate and outdo. The SPME technique uses a small polymer-coated fiber that is placed in solution or in the headspace of a sample for a period of time. Analytes diffuse by convection to the surface of the polymer coating until equilibrium is achieved. Once withdrawn from the sample source, in the case of gas chromatography (GC), the fiber is placed into the hot injection port and the sorbed analytes are thermally desorbed into the GC column to be separated and detected. This SPME approach was soon commercialized for both manual and automated operation. For high performance liquid chromatography (HPLC), the SPME technique never proved as successful. In this case, the analytes had to be solvent extracted (or stripped) from the polymer coating, which proved to be a slower process than thermal desorption. To overcome some of these problems, an in-tube SPME technique was developed in which analytes were sorbed by repeatedly aspirating and dispensing the sample liquid through a short piece of GC capillary whose interior walls were coated with a stationary phase (3). Then, the sorbed analytes were displaced by the HPLC solvent to the column for analysis. 

A variation of the SPME technique was introduced by Baltussen and coworkers (4,5) in 1999. The stir-bar sorptive extraction technique (SBSE) involves a glass stir bar (having a magnet within) coated with polymeric sorbent (for example, polydimethylsiloxane). The available surface area of the coated stir bar is substantially higher than the surface area of an SPME coated fiber (50–250 times greater). Hence, larger amounts of analyte can be extracted and overall sensitivity is improved. A special thermal desorption unit is used to transfer sorbed analytes into the GC. The stir bar is placed in an aqueous (or other) environment, stirred for an appropriate time (usually tens of minutes), removed, dried to remove any water droplets, and placed in the thermal desorption unit for further analysis. Depending upon the octanol–water partition coefficients of the analytes, sample volume, stir-bar size, and analytical system being used (usually GC–mass spectrometry [MS]), detection limits of 1 ng/L have been obtained. The technique has been commercialized by Gerstel (Baltimore, Maryland) as the Twister extraction system. 

 The first attempt to emulate the SPME experiment for LLE was performed in the late 1990s (6,7). Here, a single drop (sometimes called the receiving or acceptor phase) of an immiscible extraction organic solvent (1–10 μL) was suspended from the end of a microsyringe needle in an aqueous solution (donor phase) with stirring (see Figure 1) or in a continuous flow system. This technique became known as single-drop microextraction (SDME), a subset of the general technique of liquid-phase microextraction (LPME). Common with static microextraction techniques, the process is rather slow requiring tens of minutes. 

Figure 1: Schematic of the SDME apparatus. 

The LPME techniques use a very small amount of organic solvent, usually only a few microliters. Due to the high volume ratio of the donor aqueous phase to the acceptor phase, analytes are concentrated during the extraction. In the case of a hanging drop (for example, SDME), once equilibrium was achieved, the drop could be retracted into the syringe barrel and injected directly into a GC, HPLC, or capillary electrophoresis (CE) instrument. The extraction of analytes from the matrix relies upon the diffusion and partition of target compounds between the solvent and aqueous phase. Therefore, the whole procedure is a process dependent upon equilibrium rather than exhaustive extraction. Parameters that accelerate the diffusion can increase the extraction efficiency, such as stirring the sample or increasing the extraction system temperature. Other parameters affecting the extraction efficiency include the volume of the solvent drop, the solvent selection, pH adjustment, salting-out effect, and so on.  

Although SDME is inexpensive, simple, and efficient, the stability of the hanging drop has always been a problem. If the stirring rate was too fast, the drop could be dislodged from the syringe needle during high-speed stirring. Although selection of a syringe with a beveled needle tip, suitable solvent, and small volume of solvent (1–2 μL) can obviate this difficulty, they cannot solve the problems completely, thus limiting the development and application of the technique. Any slight solubility of the acceptor liquid in the aqueous solvent could cause the drop to change size with time, thereby affecting extraction efficiency. In addition, biological samples, such as plasma, can emulsify substantial amounts of organic solvents that also can affect the stability of the hanging drop during the extraction process. 

Furthermore, drop-based LPME works best with a clean matrix because particles or bubbles in the sample matrix affect the extraction by making the drop unstable and are potentially detrimental to the analytical instrument. To protect the solvent drop, compromises must be made during the method optimization, such as shortening the equilibrium time, using a lower stirring speed, and so on, which negatively impact the extraction efficiency. In other words, the SDME method is a very good sample preconcentration technique but is generally unsuitable as a sample clean-up procedure. 

Hollow-Fiber Liquid-Phase Microextraction: 

To circumvent the problem of drop stability, Pedersen-Bjergaard and Rasmussen (8) came up with the idea to contain the drop (or small volume of organic solvent) within the lumen of a porous hydrophobic hollow fiber, typically made of polypropylene. Thus, the drop was "protected" against mechanical disturbance within the confines of the fiber. The technique has been referred to as hollow-fiber liquid-phase microextraction (HF-LPME). With the presence of the hollow fiber, extractions could be performed in either a two-phase mode or a three-phase mode. The two-phase mode consists of the organic solvent (a few microliters) in the lumen as well as being immobilized in the wall  pores of the hollow fiber. The extracting solvent (acceptor) is not in direct contact with the sample solution because the porous membrane of the hollow fiber serves as a barrier to the aqueous sample. The porous hollow fiber also could serve as a three-phase extraction system in which the hollow fiber is first treated with organic solvent that provides a supported liquid membrane (SLM) with the immiscible solvent filling the pores of the membrane. An acceptor solution, usually aqueous, is placed inside the lumen of the hollow fiber. Thus, the three-phase system could be used to perform liquid–liquid–liquid microextraction (LLLME). The SLM technique was pioneered by Audunsson (9) and popularized by Jonsson and Mathiasson (10). Later, we will get into more details of the workings of HF-LPME. 

Dispersive liquid–liquid microextraction: 

A technique called dispersive liquid–liquid microextraction (DLLME) is based upon a ternary component solvent system (11,12). In this method, also useful for extractions of analytes from aqueous media, a small volume of extracting solvent (higher density than water and about 1–3% of the total volume of extracting mixture) and a disperser solvent (miscible with extracting solvent and aqueous sample solvent) are mixed together and are rapidly injected into the aqueous sample. A cloudy mixture is formed due to the tiny droplets of immiscible extracting solvent in the aqueous environment. Due to the high surface area of these extracting solvent droplets, analyte extraction is quite efficient and extremely rapid. The extracting solvent containing analyte is then centrifuged and the sedimented phase collected for direct analysis or further workup. 

Solidified floating drop microextraction: 

Solidified floating drop microextraction (SFODME) is a technique similar to DLLME in that a small volume of immiscible organic solvent with a melting point near room temperature is floated on the surface of an aqueous sample containing analyte (13). The phases are stirred for a period of time and the mixture transferred to an ice bath. The organic solvent is solidified, removed, and placed in a vial where it melts, thereby releasing the analyte for further analysis. This approach of DLLME allows the use of an extracting solvent that is less dense than water. 

Hollow fibers are used in a wide variety of applications, the major one being their use for wastewater treatment but in the laboratory, they have been used for microfiltration and reverse osmosis. They are made of organic polymers such as polypropylene, polyethersulfone, polyester, and other materials and also can be constructed from inorganic materials such as titania and zirconia. The hollow fibers are relatively inexpensive and can be considered for single use only; in addition, the materials are recyclable. To remove potential contaminants, prior to use in any analytical application, hollow fibers should be rinsed in acetone or other compatible solvent several times, then dried at room temperature. 

The apparatus for static HF-LPME is illustrated in Figure 2. Rather than the spherical configuration of a drop, the HF-LPME configuration provides a rod-like shape of extracting solvent. The rod-like configuration increases the solvent surface area because for the same volume of liquid, the surface area of a sphere is lower. The contact area between sample solution and extracting solution is thus much more substantial than if the solvent were spherically shaped. The direct benefit of the change in solvent configuration is better extraction efficiency. Another significant benefit is that higher stirring speeds can be applied during the extraction procedure, because the solvent is protected by the hollow fiber and its stability is enhanced. In addition, the disposable nature of the hollow fiber eliminates the possibility of sample carryover and ensures high reproducibility; and the pores in the walls of the hollow fiber cause it to display some selectivity by preventing the extraction of macromolecules, such as proteins, and particles from the sample matrix. Therefore, HF-LPME is no longer just a preconcentration technique. It also can provide simultaneous sample clean up, and it is suitable for use with complex sample matrices. 

Figure 2: Schematic of the HF-LPME apparatus. 

The selection of organic solvent used to fill in the pores of hollow fiber is a critical parameter in the success of the method. There are several requirements for the organic solvent. Firstly, it should be immobilized easily in the hollow-fiber pores, and it should be nonvolatile. Typically, the hollow fibers used in most HF-LPME experiments are made of hydrophobic polypropylene, and the solvent selected should have good affinity for it to prevent solvent loss during the extraction, and also to achieve stable solvent impregnation. Secondly, the solvent should be immiscible with water as it serves as a barrier to the aqueous sample. Lastly, the solubility of the analyte of interest in the solvent should be higher than that in the aqueous sample matrix. Usually a relatively nonpolar, high viscosity solvent will meet those requirements, and typical solvents used include 1-octanol, 

-hexyl ether, undecane, toluene, and so forth. However, these relatively nonpolar solvents usually have low solubility for the polar compounds, thus negatively impacting the extraction efficiency for polar compounds. Recently, ionic liquids have been used as solvents for HF-LPME applications (14–17). Ionic liquids are polar and nonvolatile, but have been demonstrated to generate stable solvent impregnation. Figure 3 shows scanning election microscopic images of the inner surface of the hollow fiber before and after ionic liquid impregnation (7). By comparing the two photographs, it is obvious that the ionic liquid was immobilized in the hollow-fiber pores effectively and an ionic liquid membrane for extraction formed. The advantages of using an ionic liquid as the extraction solvent include high affinity for polar compounds and compound-dependent selectivity. These specific features of ionic liquids have broadened the applications of HF-LPME to polar compound analysis. 

Figure 3: Scanning election microscopic images of the inner surface of the hollow fiber (a) before and (b) after ionic liquid impregnation. 

The HF-LPME procedure is simple and includes just a few steps (see Figure 2). Before extraction, the hollow fiber (typically, 1.5–10 cm in length, 200-μm thick, 600-μm i.d. with a 0.2-μm wall pore size), usually sealed at one end and attached to the syringe needle on the other end, is soaked into the immiscible solvent that results in the immobilization of the solvent into its wall pores. The solvent forms a thin layer within the wall of the hollow fiber. A microsyringe usually is used to support the hollow fiber and also to introduce the acceptor phase into its lumen. The hollow fiber with added acceptor solution is then placed into a sample vial filled with the aqueous sample, usually with a volume of several milliliters. To speed up the extraction, the sample is stirred extensively. The analytes are extracted from the aqueous sample through the solvent layer in the pores of the hollow fiber, then into the acceptor phase inside its lumen. After a certain time of extraction to achieve equilibrium (often in the range of 10–30 min), the acceptor phase is withdrawn into the microsyringe and then injected directly into the instrument for analysis. 

Depending upon the mode, the lumen of the hollow fiber can be filled with several microliters of the same solvent (two-phase mode) or an aqueous acceptor phase (three-phase mode). In two-phase HF-LPME, organic solvent impregnated in the hollow-fiber segment is exposed as a solvent rod in a stirred aqueous sample. Nonpolar analytes in the aqueous sample then can be extracted largely through the hollow fiber into the acceptor solvent by diffusion. Because the solvent used is usually GC-amenable, the extracted sample can be then injected directly into the GC system for analysis. This two-phase HF-LPME is more suitable for the extraction of relatively hydrophobic analytes (17–19).  

In this installment, the subject of LPME is reviewed, with emphasis on the use of hollow fibers.

LCGC North America-08-01-2010

System Contamination

Why Does It Take So Long To Validate a Chromatography Data System?

Biosimilars: Introduction, Concerns, and Opportunities

Early Adventures in Chromatography: The Story of a "Recovering" Electrochemist

The Use of Hollow Fibers in Liquid-Phase Microextraction

The "Chrommunity" Welcomes Two Distinguished New Members

Vol 28 No 8 LCGC North America August 2010 Regular Issue PDF