A Robust Static Headspace GC-FID Method to Detect and Quantify Formaldehyde Impurity in Pharmaceutical Excipients

Formaldehyde is a highly reactive impurity that can be found in many pharmaceutical excipients. Trace levels of this impurity may affect drug product stability, safety, efficacy, and performance. A static headspace gas chromatographic method was developed and validated to determine formaldehyde in pharmaceutical excipients after an effective derivatization procedure using acidified ethanol. Diethoxymethane, the derivative of formaldehyde, was then directly analyzed by GC-FID. Despite the simplicity of the developed method, however, it is characterized by its specificity, accuracy, and precision. The limits of detection and quantification of formaldehyde in the samples were of 2.44 and 8.12 µg/g, respectively. This method is characterized by using simple and economic GC-FID technique instead of MS detection, and it is successfully used to analyze formaldehyde in commonly used pharmaceutical excipients.


Introduction
Although considered pharmacologically nonactive, pharmaceutical excipients have critical e ects on drug product safety, e cacy, and quality. During early stages of drug formulation development, excipients are evaluated for appropriateness by studying their physical and chemical compatibility with drug substances under stressed and accelerated conditions. Depending on the results of the compatibility studies, formulator can exclude incompatible excipients from later stages of formulation development. Chemical incompatibility with drug substances may result not only from direct reaction with excipients but also from reaction with excipient impurities. ese impurities include formaldehyde which could be generated by autoxidation degradation of many pharmaceutical excipients such as polyvinylpyrrolidone (PVP) and polyethylene glycol (PEG) [1][2][3][4].
Formaldehyde has a low molecular weight and a high chemical reactivity. Furthermore, it has little UV activity and low detector sensitivity and speci city. Besides, it is soluble in both water and organic solvents. As a consequence, it is di cult to directly extract and determine formaldehyde in a speci c, accurate, and sensitive way. erefore, most of the analytical methods used for formaldehyde analysis employ chemical derivatization to improve its stability and modify the physicochemical properties to enhance its detectability [11,[13][14][15][16][17].
Many analytical methods have been developed to determine formaldehyde in pharmaceuticals and cosmetics. ese methods are based on colorimetric, spectrophotometric, uorescent, capillary electrophoresis, HPLC, GC, and GC-MS techniques [1,[17][18][19][20]. Lots of the reported methods have relied on derivatizing formaldehyde using acetylacetone or chromotropic acid reagents prior to colorimetric and/or spectrophotometric determination. However, the main disadvantages of these methods are the lack of appropriate speci city and/or sensitivity [1,17,20,21]. 2,4-Dinitrophenylhydrazine, the most popular derivatization reagent, has been used before determination of formaldehyde by HPLC [11,15,22]. However, some of these methods have limited selectivity or have su ered from di culty handling and injecting excipient samples which form viscous solution (such as PVP). In addition, other HPLC methods require tedious and multisteps extraction procedures and/or long analysis time [1,16,20,23]. Gas chromatography is the best choice for determination of volatile components, especially when associated with the headspace sampling technique. is technique provides a simple way to directly inject the extracted volatiles into GC. In contrast with the direct injection of the sample solution into GC inlet, the headspace injection allows the volatiles to be analyzed without interference by the nonvolatile matrix. By using the headspace technique, analysts can also simply derivatize target components by utilizing the headspace vial as a reaction vessel [24]. Few GC methods have been developed to determine formaldehyde in pharmaceutical excipients; however, the derivatization reaction, the extraction, and/or the sample preparation were generally complex and/or required long time [13,16,19,20]. del Barrio et al. reported a headspace GC-MS method for determination of formaldehyde in some excipients by converting it to diethoxymethane using acidi ed ethanol. In this method, the sample preparation was very simple and rapid, and the chemical derivatization can be carried out under mild conditions. However, the authors did not evaluate the capability of using FID for quanti cation [16]. Although FID is less sensitive than MSD, it is the most widely used detector for routine quanti cation in typical analytical laboratories. In addition, it is easier to operate and maintain [20,25].
As most of the spectrophotometric methods have some disadvantages, while other chromatographic methods are complex to be performed, the aim of this research is to develop and validate a static headspace (SHS) GC-FID method to determine formaldehyde, and then to use it as a screening method and a quality control tool to analyze this impurity in commonly used pharmaceutical excipients.

Instrumentation.
Experiments were conducted by Agilent Model 7890A gas chromatograph equipped with mass selective detector (MSD model 5975C) and associated with GC sampler 80 enhanced with Agilent PAL headspace option. Agilent MSD Productivity ChemStation E.02.01.1177 software was used to control, acquire, and process the chromatographic data. A 30 m × 0.25 mm i.d. ZB-WAX column with 0.25 µm lm thickness (Phenomenex, USA) was utilized for gas chromatographic separation.

GC Instrumental Conditions.
e injector was maintained at 170°C with a split ratio of 1 : 25. e column oven temperature program involved an initial temperature of 35°C for 5 min and increased at 40°C/min to 220°C and held for 1 min. e carrier gas was helium (99.999%) at a constant ow rate of 0.9 mL/min. FID was set at 280°C for quantication. MS detection was carried out at 230°C with full scan (31-250 amu) for identi cation.

Sample Preparation.
Samples were prepared by directly weighing 250 mg of the tested excipient into a 20 ml amber headspace vial (Supelco). After that, 5 mL of the solution of 1% (w/w) p-toluenesulfonic acid in ethanol was added to the content of each sample vial which was then immediately sealed with a magnetic screw cap lined with a butyl/polytetra uoroethylene septum (Supelco) and shaken for 2 minutes (until the content becomes completely dissolved). Finally, the prepared vials were sequenced and automatically moved to the incubator for completing the chemical derivatization of formaldehyde.

Standards Preparation.
e concentration of formaldehyde solution was determined by applying the iodometric method described in British Pharmacopoeia 2013 [5]. e concentration was 35.10% (w/w).
Standard solutions of formaldehyde were prepared in the acidi ed ethanol (1% p-toluenesulfonic acid). A stock standard solution of formaldehyde at 1251.063 µg/ml was prepared and used to prepare a series of standard solutions at lower concentrations by serial dilutions. en, 5 mL of each standard solution was transferred to a headspace vial and treated as described above.

Method Validation.
e method was validated in terms of specify, linearity, accuracy, repeatability, intermediate precision, limit of detection, and limit of quanti cation according to British Pharmacopoeia 2013 [5].

Statistical Analysis.
Results were statistically processed by t-test or one-way analysis of variance (ANOVA) in order to evaluate signi cant di erences (p < 0.05). Statistical Package for the Social Sciences (SPSS, 20) software was used for this purpose.

Development and Optimization of SHS-GC-FID Method
e identi cation of the derivative, the product of the chemical reaction of formaldehyde with ethanol, was con rmed by using the corresponding standard (diethoxymethane). As shown in Figure 1(B), the retention time of formaldehyde derivative peak (t R � 3.837 min) matched with that belonging to the corresponding standard peak eluted according to Figure 1(C). is indicates that the derivative was diethoxymethane.
To further con rm the identity of formaldehyde derivative, EI mass spectrum of the corresponding peak (shown in Figure 2) in full scan was acquired and identi ed using the NIST MS library as diethoxymethane.

Optimization of Sample Preparation and Headspace Sampling
(1) Preliminary Investigation. At the beginning, the sample preparation and the headspace sampling parameters described by del Barrio et al. [16] (the sample concentration: 100 mg/mL; the headspace parameter: incubation at 60°C for 15 min) were applied for analyzing and evaluating the recovery of formaldehyde derivative from the spiked samples of PEG 400 and PVP K-30. e results (n � 2) demonstrated that the mean recovery of formaldehyde derivative from PEG 400 was within the acceptable range (80-120%); however, it was not so for PVP K-30 (56%).
In order to explore the factors that leaded to such low recovery of formaldehyde derivative from PVP K-30, the sample matrix e ect and the previous incubation parameters were checked. First, the experiments carried out on the ethanolic solution of diethoxymethane standard showed that the incubation time of 15 min at the temperature of 60°C was su cient for diethoxymethane to reach the static headspace equilibrium. Second, the mean recovery of diethoxymethane standard was 110% (n � 2) when preparing PVP K-30 samples by adding 5 mL of the ethanolic standard solution.
us, there is no negative e ect of PVP K-30 sample matrix on formaldehyde derivative extraction. ird, the incubation time required to complete the derivatization reaction of formaldehyde in PVP K-30 was evaluated at the temperature of 60°C. As shown in Figure 3, the incubation for 15 min was not su cient for completing the derivatization reaction of formaldehyde. Rather, the incubation time should be adjusted at 45 min at least to ensure completion of the reaction. is di erence in the incubation times required for  Journal of Analytical Methods in Chemistry 3 PVP K-30 sample may be due to existence of a relationship between the physical nature of the sample and the kinetic of the derivatization reaction of formaldehyde.
(2) E ect of Sample Dilution on Incubation Time. e relatively high viscosity of ethanolic solution of PVP K-30, when prepared at concentration of 100 mg/mL, may be the reason behind the long time required to reach equilibrium for formaldehyde derivative (Figure 3). Although sample dilution can decrease sensitivity of the analytical method, however, it can reduce viscosity of the ethanolic solution of PVP K-30, which can in turn decrease the required time for both completing the derivatization reaction and reaching equilibrium for formaldehyde derivative between the sample and the headspace gas phases [24]. erefore, the sample concentration of 50 mg/mL was used to evaluate the required incubation time for the spiked PVP K-30 samples while xing the incubation temperature at 60°C. Figure 3 shows the diagram which correlates between the peak area of the derivative and the increased incubation time. e results demonstrated that there was no signi cant increase in the peak area of formaldehyde derivative after 30 min incubation. As a consequence, sample dilution from 100 mg/mL to 50 mg/mL gave the advantage of lowering the incubation time to 30 min (instead of 45 min at least). erefore, the sample concentration of 50 mg/mL was selected to complete the other experiments.
(3) E ect of Increasing the Incubation Temperature. Increasing the incubation temperature generally leads to an increase in the rate of the derivatization reaction and the headspace sensitivity. On the other hand, it results in decreasing the required time to reach the static headspace equilibrium [24]. erefore, the incubation temperature was raised to 70°C (higher temperatures were not experimented to avoid any potential degradation of the samples). e results (shown in Table 1) indicate that there was no signi cant increase (t-test, p > 0.05) in the detector response when the reaction mixture was incubated at 70°C (instead of 60°C) for 30 min. Hence, there was no increase in the sensitivity. In contrast, the incubation time required to complete the derivatization reaction of formaldehyde in excipient samples decreased. Figure 4 shows the plateau of the peak area of formaldehyde derivative with increase in incubation time at 70°C for both PEG 400 and PVP K-30. e data show no signi cant increase in the peak area of formaldehyde derivative after 15 min for PEG 400 and 25 min for PVP K-30.
As a consequence, the increasing of incubation temperature from 60°C to 70°C leaded to a decrease in the incubation time from more than 15 min to 15 min for PEG 400 and from 30 min to 25 min for PVP K-30. erefore, the incubation temperature of 70°C was selected.
(4) Headspace Sampling Parameters. Headspace sampling parameters other than incubation time (injection speed and injection volume) were optimized and set in a way that can achieve a balance between the sensitivity and the peak shape parameters. e nal headspace sampling parameters are given in Section 2.3.

E ect of Water Presence on Formaldehyde
Determination. Diethoxymethane formation reaction is susceptible to water presence which can reverse the reaction [16,26]. erefore, the tolerance of the derivatization reaction of formaldehyde to water was evaluated by adding it to several vials containing 5 mL of the reaction mixture (formaldehyde in acidi ed ethanol at a concentration of 1.5 µg/mL) at the following amounts: 0, 12.5, 25, and 50 mg (0, 5, 10, and 20% w/w relative to the sample weight 250 mg). After analysis and statistical processing, the results (shown in Table 2) manifested that there was no signi cant di erence (ANOVA test, p > 0.05) between the control (0% water) and the other samples.
As a consequence, excipient samples containing water ≤ 20% (w/w) can be analyzed without any signi cant e ects on formaldehyde determination. Higher amounts of water were not experimented because the acceptable limits of water content stated in the monographs of PEG and PVP are 1-2% and 5%, respectively [5,6].

Speci city.
e speci city of the method was determined by comparing the chromatograms (Supplementary Material Figure S1) of the blank (acidi ed ethanol), the stressed excipients dissolved in ethanol and the standard solution of formaldehyde in the blank. e chromatograms showed that there were no interferences at the retention time of formaldehyde derivative peak. In addition, all volatile component peaks, which appeared after analyzing spiked excipient samples, dissolved in the blank (Supplementary Materials Figures S2 and S3) were completely separated from the analyte peak (the resolution between the analyte peak and any other one was more than 2).

Linearity.
e linearity of the method was evaluated from injection of nine concentrations of formaldehyde at the range of 0.25-50 µg/ml. e analyte showed excellent linear behavior over the speci ed range with coe cient of correlation (R 2 ) value of 1.

Accuracy.
e accuracy of the optimized method was determined by spiking the excipient samples with known amounts of formaldehyde at four concentration levels (n � 3). e results (shown in Table 1) were expressed as mean recoveries. e recovery at all concentration levels was within the acceptable range (80-120%). is indicates that the method is accurate.

Precision.
e precision of the method was evaluated in terms of the repeatability (expressed as intraday precision) and the intermediate precision (expressed as interday precision). e repeatability of the method was established from triplicate (n � 3) injections of each spiked excipient sample at each concentration level. Intermediate precision (interday precision) was carried out by analyzing the spiked excipient samples, prepared in the same way and at the same concentration levels, on two di erent days by two di erent analysts. e results of the method precision (shown in Table 3) were expressed as a relative standard deviation (RSD%). All RSDs were less than 4%.
is indicates that the method is precise.

Limits of Detection and Quanti cation.
e limits of detection and quanti cation (LOD and LOQ, resp.) were evaluated based on signal-to-noise ratios of 3 and 10, respectively, and the respective values were found to be 121.80 and 406.00 µg/L. Because the sample concentration was 50 mg/mL, LOD in excipient samples was 2.44 µg/g, while LOQ was 8.12 µg/g.  3.2.6. Robustness. Changing the ow rate within ±0.1 mL/min had no e ect on the area of formaldehyde derivative peak, and all RSDs were less than 5%. is indicates the robustness of the method.

Performance of the Proposed Method.
Little analytical methods have been developed to determine formaldehyde in pharmaceutical excipients [13]. Table 4 shows a comparison between these methods and the proposed one.
Compared to other methods, the time required to complete formaldehyde determination using the developed method was generally faster. Recovery obtained with the presented procedure was comparable with those reported by others. Precision of the method was also close to those mentioned by others; however, it was better than that for SPME-based method. Although the proposed method had lower sensitivity than most of the other methods, however, it gives the opportunity to detect and determine formaldehyde impurity in pharmaceutical excipients below the limit stated in British pharmacopoeia (15 ppm for free formaldehyde in PEG) [5].

Analysis of Commercial Pharmaceutical Excipient
Samples. PVP and PEG are widely used in many pharmaceutical applications and dosage forms. ey are utilized in drug products over a wide range of concentrations. ese concentrations can reach up to 67% for PEG 400 and 25% for various grades of PVP [3,27]. e optimized SHS-GC-FID method was applied to analyze the presence of formaldehyde in several commercial PEG and PVP samples. e results of the analysis are summarized in Table 5.
Formaldehyde content in the excipient samples was extensively varied from 8.04 to 190.58 µg/g. ese variations depended on the nature, the storage conditions, and the source of the excipient samples.     Journal of Analytical Methods in Chemistry Formaldehyde was not detected in PVP K-30 samples from the companies A and B. However, PVP samples (from the companies C and D) contained trace amount of formaldehyde. Actually, there is no speci c pharmacopoeial limit for formaldehyde in PVP, but there is a general limit for aldehydes (500 ppm) which expressed as acetaldehyde [5,6].
According to British Pharmacopoeia, the limit of formaldehyde in PEG is 15 ppm [5]. Both PEG 400 from the company E and PEG 300 from the company G had formaldehyde levels exceeding the identi ed limit.
Although PVP K-25 from the companies C and D and PEG 400 from the company F contained trace levels of formaldehyde, however, such levels may cause critical drug products instability and degradation, especially when excipient/active pharmaceutical ingredient ratio is high, according to many literatures [1,2,4].

Conclusion
A SHS-GC-FID method was developed and validated for determination of formaldehyde in pharmaceutical excipients. Samples were simply prepared in headspace vials by adding acidi ed ethanol as a diluent and derivatization reagent. After that, the vials were automatically moved to the incubator where formaldehyde derivative was simultaneously formed and extracted. e developed method was linear, speci c, accurate, and precise over the speci ed range. e limits of detection and quanti cation were 2.44 and 8.12 µg/g, respectively. e method was also simple and rapid, and it does not require sophisticated instrumentations or large amounts of solvents. is method was used to screen PEG and PVP samples for formaldehyde. e tested samples contained varying levels of it. So, the method could be valuable in selecting appropriate excipients and/or excipient batches for pharmaceutical formulation. It is also important for selecting approved vendors of pharmaceutical excipients. In addition, it can be applied as a tool for testing the quality of pharmaceutical excipients routinely.

Conflicts of Interest
e authors declare that they have no con icts of interest.