Equivalent Column Selection in HPLC
HPLC methods are used today to control the quality of many chemical and pharmaceutical products. The methods are usually developed by optimizing the properties of the mobile phasewith a given column. If the original column is not produced any longer, with another column the separation will often change. Therefore, we were interested to find one or two equivalent columns which can replace the original column without changes in selectivity and robustness. This study will show a new way to compare columns and to select suitablereplacement columns. The presented procedure will allow to evaluate to use different columns with the same method, and to evaluate the robustness of the common method with different columns.
SHL Group details its approach to the challenges of emerging formulations
Some method validation reports suggest searching for an equivalent column in order to replace the original reference column with the ability to fulfil nearly the same quality of separation. This may be necessary in case the original column is no longer available on the market. Finding an equivalent column is difficult, because the same type of column from different manufactures, or even different batches of the same material from the same manufacturer, can result in different chromatographic selectivity’s .
There are several approaches to select equivalent columns. The most well-known is the Snyder-Dolan Hydrophobicity Subtraction Database (SDHSDB), further the Tanaka approach , and the USP database . These three approaches characterize stationary phases based on a fixed sample mixture measured under fixed experimental conditions. This is leading some-times to unexpected results, if two “similar” columns are used with a given sample under a given set of experimental conditions.
This study describes a novel approach to find an equivalent column for a reference column using modeling software, based on the approach described in . This approach is precise in its predictions and fulfills Quality by Design (QbD) criteria, which means, to test tolerance limits of the method. It uses the so called Critical Resolution Cube (CRC) to characterize peak movements with a large number of different columns for a given sample. This leads to a better understanding of the factors and their variabilities, influencing chromatographic separations.
This study describes a novel approach to find an equivalent column for a reference column using modeling software,
2. Experimental Part
Two different C18 columns were used in all experiments: ACE Excel 3 C18-PFP (75 x 4.6 mm, 3 μm), and HALO C18 (50 x 4.6 mm, 2.7 μm)(MacMod, USA).
Eluent A was water (H2O) buffered with 10mM formic acid to pH 2.0. Eluent B was varied between acetonitrile (B1) and methanol (B2). Flow rate (F) was 1.0 mL/min.
HPLC separations were performed on a Shimadzu LC-2010C with UV detection at 254 nm. Data collection, peak integration and evaluation were done using LabSolution 4.5 (Shimadzu -Europe, Duisburg, Germany). Peak tracking and method modeling were done with DryLab4 using the new Robustness Module and the Knowledge Management Module (version 4.3, Molnár-Institute, Berlin, Germany).
The experimental design (DoE) for simultaneous optimization of gradient time (tG), temperature (T) and ternary composition (tC) required twelve experiments as illustrated in Fig 1. The injection volume was 2 μL.
Fig. 1. Design of Experiments (DoE). First the runs at T1 were carried out in the order of 1->2, 5->6 and 9->10. Than the oven was heated up to T2: 60°C and the runs were 11->12,
7->8 and finally 3->4. tC1 was acetonitrile(AN) and tC2 was methanol (MeOH).
3.Results and Discussion
3.1 Establishing a common working point
After the twelve experiments on each column according to the DoE (s. Fig 1) were carried out, two 3-dimensional resolution models (“Cubes”) for each column could be calculated by DryLab4 (s. Fig.2). After that, a “Common Cube” was established automatically from the two Cubes by taking at the same point (same combination of tG, T and tC) the smallest critical resolution value (Rs,crit) from the two Cubes.
The two Cubes and their Common Cube at the tG-T-plane for tC = 74 % (B2 in B1) = (AN:MeOH)(26:74)(V:V), are illustrated in Fig. 2. The common Cube was calculated using a proprietary algorithm. Within this Cube, an optimum “Common Working Point” (CWP) for both columns could be localized. The CWP is at that point the one with the highest Rs,crit within the Common Cube. It was found at CWP = (tG = 28 min, T = 27 °C, tC = 74 %B2 in B1) with Rs,crit = 2.78 (See Fig. 2).
Fig. 2. 3D-tG-T-tC models showing the tG-T-plane for tC = 74 %B2 in B1, at tG = 28 min and T = 27 °C. A and B are the 2 equivalent columns and C. is the Common Cube where both columns are considered.
Each point within a Cube corresponds to a precisely modeled chromatogram, and each Cube represents more than a million virtual experiments. Red areas within the Cubes indicate, that the Analytical Target Profile (ATP), i.e., the critical resolution, was above or equal 2.5 (Rs,crit ≥ 2.5), could be achieved. Blue areas indicate co-elution of critical band pairs (Rs,crit = 0). The CWP is marked with a node and an arrow.
3.2 Evaluating the robustness of the common working point on both columns
The robustness of a method is defined by the ICH as the methods capability to remain unaffected by small variations in the selected conditions.
Six method parameters tG, T, tC, F, Start %B, and End %B were assessed for the calculations.
For this study we used three distinct values for the level of each parameter at, above, and below the selected point (tolerance limits: 0.0, +1.0, -1.0 for tG [min], T[°C], Start %B, and End%B [%], and 0, +0.1, -0.1 [mL/min] for F ). The required critical resolution (as the ATP) was set to Rs,crit = 2.0. The Robustness Module calculates a chromatogram for every possible combination of the variables and their tolerance limits. Since six parameters were used for the robustness testing, the total number of modeled experiments was equal to 36 = 729.
Fig. 3 shows the results of the robustness testing for the two columns at the CWP. All the 729 modeled experiments on each column were above the required resolution, i.e., both columns had a 100% success rate. Furthermore, Tab. 1 and Fig. 4 show the comparison between the predicted and experimental Rs,crit at CWP.
Fig. 3. Frequency distribution of the critical resolution of the 36 = 729 modeled experiments at the working point of tC = 74 % (B2 in B1), at tG = 28 min and T = 27 °C.
Fig. 4. Predicted and experimental chromatograms at CWP (tG = 28 min, T = 27 °C, tC = 74 % (B2 in B1)). As we can see, there is a high correlation between predicted (blue) and measured (red) chromatograms and their retention times.
With the latest version of DryLab4 it is now possible to develop an HPLC method, which could be used on different equivalent columns with the selected ATP according to Quality by Design (QbD) principles. Additionally, the robustness was tested for each column with good success at the “Common Working Point” for robust routine Quality Control of drug products. The use of different column chemistries will increase for the purpose to achieve the maximum on robustness in routine Quality Control according to QbD.
The authors thank the Shimadzu Europe Corporation for their support with the HPLC instruments and MAC-MOD Analytical Inc. for providing columns for this work.
 L. R. Snyder, J. W. Dolan and P. W. Carr, "The hydrophobic-subtraction model of reversed-phase column selectivity," Journal of Chromatography A, 1060, 77-116, (2004).
 M. R. Euerby, M. James and P. Petersson, "Practical implications of the “Tanaka” stationary phase characterization methodology using ultra high performance liquid chromatographic conditions," Journal of Chromatography A, 1228, 165-174 (2012).
 USP-Database. [Online]. Available: http://apps.usp.org/app/USPNF/columnsDB.html (accessed April 2016).
 R. Kormány, I. Molnár and H.-J. Rieger, "Exploring better column selectivity choices in ultra-high performance liquid chromatography using Quality by Design principles," Journal of Pharmaceutical and Biomedical Analysis, 80, 79-88 (2013).
Share this article