Verification of a compendial method: Water Determination by Karl Fischer Titration
In lyophilizates of pharmaceutical drugs, the residual moisture content must be determined. It must be as low as possible in order to delay possible degradation reactions during storage and thus ensure the declared shelf life. Such a water determination can be performed applying the Karl Fischer titration. This titration method is a widely used procedure for quantitative water determination in a variety of samples and was developed by Karl Fischer in 1935. In the European Pharmacopoeia (Ph. Eur.), chapter 2.5.12 covers the volumetric (semi-micro) determination and chapter 2.5.32 covers the coulometric titration. The choice of method is primarily based on the amount of water present in the sample.
Analytical principle of water determination according to Karl Fischer
The procedure of this compendial method is based on the fact that iodine can be reduced by sulfur dioxide in the presence of water (and absence of alcohols) (1). However, if a suitable alcohol (such as methanol, CH3OH) comes into play, it grabs the sulfur dioxide and forms an acidic ester that is neutralized by a suitable base such as imidazole (abbreviated as RN) (2) and the iodine is, at first glance, out of luck. However, the methyl sulfite anion (CH3SO3) is converted to methyl sulfate (CH3SO4) by the present water, resulting in reduction of the yellow-brown iodine (I2) to colorless iodide (3).
2 H2O + SO2 + I2 → SO42− + 2 I− + 4 H+ (1)
CH3OH + SO2 + RN → (RNH)·(CH3SO3) (2)
(RNH)·(CH3SO3) + I2 + H2O + 2 RN → (RNH)·(CH3SO4) + 2 (RNH)·I (3)
Since this reaction is dependent on the water present, it can only proceed until the water present in the sample is completely consumed. Accordingly, once all water has been used, further iodine added from the titration solution can no longer be reduced and the brown coloration resulting from excess iodine indicates the end point.
In contrast to the volumetric determination described above, in the coulometric method the iodine required for the reaction is generated electrochemically by oxidation of iodide at an anode in the reaction cell of the instrument. This consumes electricity. The iodine produced reacts as described above or remains in solution after reaching the end point. The current expended to generate the iodine is thus directly proportional to the amount of quantitatively reacted water from the sample.
Practical performance of a coulometric Karl Fischer titration for water determination
A coulometer from a well-known manufacturer and an appropriate anolyte solution are used as well as dry formamide as solvent. The coulometer consists of the reaction cell, the electrodes (iodine-generating anode and iodine consumption-measuring cathode) and a magnetic stirrer.
A blank determination is performed using formamide. After removing the label and cleaning the outer wall of the vial, the sample (in our case a lyophilizate) is first weighed and then mixed with a defined volume of formamide and resuspended. A defined volume is removed and introduced into the reaction cell filled with electrolyte solution. This can be done using a syringe, as in our example. Alternatively, it would also be possible to introduce the sample by evaporation. For this, the sample is heated in a special oven and the vapor is transported to the reaction cell by means of inert gas. For the water determination, the sample is then stirred for 30 seconds and then titrated until a stable end point is obtained.
The remaining resuspended sample vial is completely dried after thorough cleaning and then cooled in a desiccator. Afterwards, the weight is determined once again. To calculate the water content of the sample, the blank value is subtracted from the measured sample value and divided by the product of the volume used and a further factor. This further factor considers the weight at the beginning and after drying and the volume of formamide added for resuspension.
Verification of a coulometric Karl Fischer titration
Both chapters of the pharmacopoeia provide information on the parameters to be considered for the verification of the compendial method. In addition to evaluating trueness (via linearity tests), it makes sense to investigate precision and to have a look at the "baseline drift" to establish appropriate system suitability test (SST) criteria, since the trueness and precision of the measurement depend on how well the ingress of atmospheric moisture into the system can be avoided. These validation parameters are also consistent with the ICH Q2(R1) requirements for a content determination except for specificity, which is also required. However, if we reconsider the analytical principle on which this method is based, it is clear that this method is absolutely specific and thus specificity does not require any additional verification.
For our example, we are in a regulated pharmaceutical environment, i.e., a laboratory with controlled room conditions, trained staff, and qualified and regularly maintained equipment.
To be able to determine the trueness via linearity experiments, we must first perform the linearity experiments. For a good experimental set up, we still need 2 pieces of information: a) the volume in which our lyophilizate will be resuspended for therapeutic use (let's say 15 mL) and b) the maximum allowed residual moisture according to our specification (such as e.g. 3%). We first prepare a spiked solution by weighing exactly 15 g of water (the resuspension volume) into e.g. a 30 mL volumetric flask and then filling it until volume with formamide solution of known (i.e., previously determined) water content. Then we perform a water determination of our resuspended sample to be examined and note the result. Afterwards, we perform 5 more determinations, each time adding a different increasing volume of spiking solution. Since our sample is only allowed to contain a maximum residual moisture of 3%, 1, 2, 3, 4 and 5% can be suitable for this purpose, even though a range of 80-120% (i.e. from 2.4 to 3.6%) is only necessary to be evaluated for content determinations. For evaluation, we plot the theoretical cumulative water volumes against the measured cumulative water volumes and apply linear regression. To evaluate linearity, we use the coefficient of determination (R2) and define that it should be, for example, at least 0.99.
To assess the trueness, we determine the recovery for each spike level used by dividing the measured water content by the theoretical water content and multiplying by 100. As an acceptance criterion, we might have set an interval of 90-110%, for example.
To get an impression of the precision, 2 analysts could perform 6 measurements each on two different days, which would cover repeatability and intermediate precision, but of course a matrix approach would also be conceivable. For our example we define a relative standard deviation (RSD) of 15%.
Now that we have already examined linearity, trueness, and precision, it is of course also possible to derive the working range from these data. It is defined as the range in which the validation parameters linearity, trueness and precision are successfully met.
Additionally, we also told at the beginning that we would like to investigate the drift to establish the SST criteria. For example, we discover getting drift values between 8 und 12 µg/min in our measurements, which is absolutely fine because it is below the maximum of 20 µg/min specified by the manufacturer of our coulometer as standard value for the starting drift. A further SST criterion could be the spread (in terms of RSD) of e.g. 3 blank determinations.
If all previously defined acceptance criteria will be compliant, the verification of this compendial method would now be fully performed and successfully completed.
However, during research for this article, I came across another interesting report from the University of Tartu and the Romanian National Institute of Metrology [1]. In this report, they describe the results of their verification of this compendial method, obtained using two different coulometers (each with oven systems for sample introduction) from two different manufacturers at their different sites. In addition to linearity determinations (each with an R2 of > 0.999 in the range studied) and limit of quantification studies (with results ranging from 1.4 bis 11 µg depending on the instrument, sample introduction with or without oven, and evaluation technique), various robustness experiments were performed. Thus, with direct injection (i.e. without oven), the polarization current between the indicator electrodes, the titration speed and the endpoint determination via relative or absolute drift or via indicator electrode potential were investigated. In addition, the influence of the flow velocity of the carrier gas in the oven was evaluated.
With regard to the polarization current, it was found that, depending on the instrument and the current applied, a lower current results in longer titration times and, due to background drift, to higher standard deviations and also shows the greatest relative difference from the expected result. The influence of the titration speed preset in the instruments was small for both instruments. Although the preset "slow" resulted in the longest titration times and highest standard deviations for both instruments, in general, relative differences to the expected value of less than 1% were found for all three settings of both instruments. Regarding endpoint determination, it was observed that the use of an absolute drift criterion was better than a relative drift criterion for both instruments but varying the indicator electrode potential did not seem to exert a great effect. When evaluating the influence of the flow velocity of the carrier gas in the oven, the largest deviations from the expected value were obtained with the lowest flow velocities for all samples studied in both devices.
Furthermore, the measurement uncertainty was evaluated in detail. Although very interesting, it would go far beyond the scope of this blog article to get into detail here, so we recommend that the curious reader take a look at the original publication ????.
Finally, it should be mentioned that the two Ph. Eur. monographs 2.5.12 and 2.5.32 are not interchangeable and that the application of the other, unregistered method is considered an alternative method and thus requires appropriate cross-validation.
References
[1] Aro R., Jaluske L., Leito I., Nicolescu I., Ionescu G. (2016). Validation report and uncertainty budget of Coulometric Karl Fischer titrator with an oven system