2.1. Chemicals Solvents were of analytical grade (SIGMA–Aldrich, Buchs SG, Switzerland). 2,4 dinitrophenylhydrazine (2,4 DNPH) was purchased from Merck (Darmstadt, Germany). Formaldehyde in water-certified reference material (CRM) (4815 mg/L) was from SIGMA– Aldrich (Buchs SG, Switzerland). 2.2. Formaldehyde solution The certified value for formaldehyde in water CRM was 47.5 mg/L ± 8.91 (mean ± st. dev.) with an expanded uncertainty of 1.82, (k = 2.23). A stock solution of formaldehyde in water (500 mg/L) was prepared using deionized water. A matrix-free calibration curve was prepared at six concentrations: 1, 2, 5, 25, 50 and 100 mg/L. For matrix-matched calibrations, matrix samples (mango, fish and milk) were spiked before extraction at 1, 2, 5, 25, 50 and 100 mg/L. To calculate the bias of the method, a stock solution of formaldehyde CRM at 47.5 mg/L concentration was prepared following the instructions provided. 2.3. 2,4 dinitrophenylhydrazine working solution 2,4 DNPH was recrystallized prior to use. Recrystallization was performed by dissolving 10 g of 2,4 DNPH in 100 mL in hot analytical grade acetonitrile to form a saturated solution. After complete dissolution, the solution was cooled to room temperature, capped in a brown bottle and stored overnight at 4 C for crystallization. The crystals were collected by vacuum filtration. 150 mg of 2,4 DNPH crystals were accurately weighed, dissolved in 49.5 mL of acetonitrile and mixed with 0.5 mL of phosphoric acid (85%). 2.4. Derivatization kinetics and sample preparation Derivatization kinetics followed the procedure described by Claeys et al. (2009) but was slightly modified. Edible parts of the food; fruit flesh and fish fillets were used for the analysis. For derivatization kinetics, mango samples were ground, homogenized and spiked with 10 mg/L of formaldehyde standard. To sample aliquots of 5 g, 5 mL of acetonitrile were added, and the sample vortexed and then sonicated for 30 min. The samples were centrifuged at 5000 rpm for 5 min and the supernatant was passed through a 90 mm diameter Whatman 541 (Hardened Ashless) filter paper (SIGMA–Aldrich, Buchs SG, Switzerland). Two and a half milliliter of 2,4 DNPH was added to the extract and mixed well. Samples were incubated at 40 C for 30, 60, 90 and 120 min in a shaking water bath (model BS-11, Oxon, UK). Formaldehyde was quantitatively converted to its Schiff base in 60 min. In all experiments, derivatization time was set to 60 min. After incubation, the acetonitrile layer was collected, membrane filtered (0.45 lm) and injected into the HPLC. 2.5. High-performance liquid chromatography conditions Analyses were performed on a C18 Luna column (25 cm 4.6 mm id., 5 lm particle size), (Phenomenex, Utrecht, The Netherlands) using an HPLC (model SPD-M20A) coupled to a photodiode array detector (both manufactured by Shimadzu, Kyoto, Japan). The wavelength was set to 355 nm and the oven temperature at 30 C. Separation was achieved using isocratic elution with a mixture of water/methanol (35:65, v/v). The flow rate was 1.0 mL/min and the injection volume 20 lL. The total run time was 12 min. 2.6. Method validation The method was validated in terms of specificity, linearity, range, limit of detection (LOD), limit of quantification (LOQ), repeatability, intermediate precision and robustness. The specificity of the method was tested by injecting reagent blank (2,4 DNPH and phosphoric acid), sample blank and formaldehyde solution individually. For linearity, the determination coefficient (R2 ) was calculated from the responses of 0.1, 1, 2, 5, 25, 50 and 100 mg/L standards. The limit of detection was calculated by the expression 3.3 sy/x/slope, based on the assumption that, the standard deviation of the signal of a solution with a concentration near to the blank is roughly the standard deviation of y-residuals (sy/x). Generally, there is a normal distribution of 5% of occurring error type a or b and the curve intercepts at zero. The quantification limit was estimated by the expression 10 sy/x/slope (Miller & Miller, 1993). For repeatability and recovery studies, 5 samples of each of the matrices were spiked at nominal concentrations at the LOQ, 2xLOQ and 5xLOQ levels and extracted by the method described in Section 2.3. Recoveries were expressed in % and repeatability as the standard deviation (sr) and relative standard deviation (%) (RSDr). The repeatability limit (r) was calculated for a coverage factor of 99.9% (using the expression, r = 2.8 sr). Intermediate precision was the standard deviation (sip) and relative standard deviation (%) (RSDip) obtained from measuring six independent sample replicates spiked at 3 different levels on 3 different days (IUPAC, 2002). The trueness of a method is the expression of how close the mean of a set of results is to the true value. The quantitative expression of trueness is biased and it was calculated.