Why Doesn't Your Method Work When I Try It?
Abstract: Published analytical methods can fail for a variety of reasons, including poor instructions and changes in instrumental or chemical factors, in addition to human error.
The title of this contribution is a question that most experienced analysts are asked or ask of others. The most common reply is - "Have you read the instructions?" This is not merely facetious, as sometimes small but vital details are listed in the experimental description that are overlooked or ignored. Often the results or discussion section of a paper contains useful additional information on a method that is missed in the hurry to get it working.
Of course, authors do make mistakes and omit important parts of a method when it is written up for publication. A modification may have been made to a method late in the day and not properly documented. Sometimes, editors of journals insist that papers are verbose and must be substantially abbreviated before they can be published. When this happens, some of the description of methodology may be jettisoned. Most referees and editors are more concerned with the general value of the science than with the fine detail.
I have heard of authors who have deliberately omitted a vital part of a method to give themselves an advantage over the "opposition", and Paul Stumpf documented an example of this . Fortunately, I do not believe that this is a frequent occurrence.
On occasion, errors arise because of imprecision on the part of authors. For example, a chromatographic procedure carried out at "room temperature" in Dundee may be several degrees C different from that in Arizona. In summer, the latter may even be cooler because they have air-conditioning - a facility not often required in Scotland! Retention values in thin-layer chromatography especially are very susceptible to changes in relative humidity, a parameter that is rarely measured, never mind listed. When I worked in Ayr, it always seemed to be 100%, so at least it was consistent.
In HPLC methods, an important aspect often ignored in descriptions is the nature and volume of the solvent used to inject the sample. Ideally, this should be the starting solvent in a gradient elution procedure and the volume should be the minimum possible - certainly not more than 10 µL in analytical applications. The loop in the injection valve should not be bigger than this. Use of too large a volume or an inappropriate injection solvent will bring about appreciable peak broadening, for example, and can even cause single components to emerge as double peaks. In reversed-phase applications with lipids, the sample may not be sufficiently soluble in the mobile phase for this to serve as the injection solvent. The minimum volume of a solvent such as chloroform or dichloromethane must then be employed for injection purposes. Hexane should not be used as this is too similar to the octadecylsilyl stationary phase in its properties; it can carry some of the sample through the column without it being retained or it can interfere with the partition of the lipid analyte with the stationary phase, again causing the components to emerge as double peaks.
Analysts may do their best to duplicate conditions in a chromatographic separation, but it is rarely possible to duplicate instrumentation. In HPLC gradient applications, for example, different pumping systems may have very different characteristics. The size and design of the mixing chamber is of special importance in determining how close the nominal gradient is to reality; if it has too large a volume, the effect may be similar to having an isocratic step at the start of the run. The quality of a gradient can also depend on factors such as the state of wear in the pump valves and seals, and in copying an existing method you have no means of knowing whether the author obtained his separation because his system was deteriorating rather than the opposite.
Most but not all analysts are aware that there can be substantial differences between manufacturers in the chemistry used in preparing stationary phases. Problems are most often seen in the analysis of phospholipids by thin-layer chromatography, for example, when different brands or even batches of adsorbent can give very different results.
Problems are also often apparent with HPLC phases. The silica gel used as the foundation of the stationary phase can vary greatly in its activity. An octadecylsilyl phase obtained from manufacturer A can differ in many respects from that from manufacturer B, although both are equally reputable companies and their products have many satisfied users. No two bonded nitrile phases ever seem to be the same. Even when the analyst goes to the trouble of obtaining a column of exactly the same phase as that used in a published separation, there is no guarantee that it will function in the same way. It may be from a different production batch, so could differ in the coverage of bonded phase and the number of residual silanols, for example. We have documented a problem in laboratory preparation of columns for silver ion chromatography elsewhere on this site. One manufacturer designates columns with bonded sulfonic acid groups as their ‘SA’ range, while another uses this term for their amine bonded phases, so confusion inevitably results as I have seen to my cost. Also, production methods are not static and manufacturers are continually working to improve their materials; the new “improved” phase with the same name as the old may be better for most purposes but not for yours.
The history of use of an HPLC column can affect its performance. If it has been used with an acidic mobile phase, for example, any remaining acid may affect subsequent analyses. Similarly, a build up of impurities from residues of previous analyses can affect the properties of a column. You may know the history of your column, but not that of the author whose paper you are trying to follow.
Chemical and Other Problems
Solvents from different sources can differ appreciably in their properties. Chloroform is possibly the best known example. Pure chloroform is rather a nonpolar solvent. When it is stabilized with a small amount of ethanol, its polarity increases substantially. Nowadays, chloroform may be stabilized by amylene, and it seems likely that this will keep the polarity at the lower end of the scale. Batches of ethyl acetate can vary greatly in their polarity and chromatographic properties, depending on how much ethanol and acetic acid are present as contaminants. Ethers, including tetrahydrofuran, often contain appreciable amounts of peroxides that affect UV absorption and may interfere with solutes.
Derivatization procedures are another problem area. Methods that work well with 1 to 10 mg of material, often turn out to be unsatisfactory when the scale is reduced. Usually, the explanation is that traces of water have an increasing effect by inhibiting the reaction. If the sample scale is small enough, say the microgram level, the trace amounts of water physically adsorbed to the glass of a test tube can have a significant effect. Difficulties arising from contaminants being introduced also increase as the sample size decreases, and most analysts have a host of bad experiences to relate about impurities in reagents, solvents, etc.
Another cause for concern is overreliance on computers – “rubbish in produces rubbish out”. If a chromatography step is inefficient, for example, the computer may print out a result, but it will not mean much. I have seen one mass spectrometry paper, where computerized identification was used inappropriately so that nearly every fatty acid component in the sample was identified incorrectly.
In an early issue of Lipid Technology , John Craske showed more courage than I would in suggesting yet another reason for why things don't work in some laboratories - simple incompetence! I would not dream of arguing with John, and many of his comments regarding gas chromatography are also relevant to my topic. I am sure that there will be no argument if I say that all staff need proper training and supervision to work effectively.
In conclusion, there are a hundred and one ways for things to go wrong in repeating a chromatographic separation or chemical reaction. The more honest of us will confess that we may occasionally have difficulties in repeating some of our own published work. It is hardly surprising that others have this difficulty also. Most of us expect to have to make some minor modifications to published methods to get them to work in our laboratories. If I could make one plea, please do not publish that minor modification as a new method!
- Stumpf, P.K. A retrospective view of plant lipid research. Prog. Lipid Res., 33, 1-8 (1994) .
- Craske, J.D. Where have all the analysts gone? Lipid Technol., 5, 94-97 (1993).
This article has been updated appreciably from one by the author that first appeared in Lipid Technology, 6, 121-123 (1994) (by kind permission of P.J. Barnes & Associates (The Oily Press Ltd)), who retain the copyright to the original article.