The CCQM-K90 comparison is designed to evaluate the level of comparability of national metrology institutes (NMI) or designated institutes (DI) measurement capabilities for formaldehyde in nitrogen at a nominal mole fraction of 2 μmol mol−1. The comparison was organised by the BIPM using a suite of gas mixtures prepared by a producer of specialty calibration gases. The BIPM assigned the formaldehyde mole fraction in the mixtures by comparison with primary mixtures generated dynamically by permeation coupled with continuous weighing in a magnetic suspension balance. The BIPM developed two dynamic sources of formaldehyde in nitrogen that provide two independent values of the formaldehyde mole fraction: the first one based on diffusion of trioxane followed by thermal conversion to formaldehyde, the second one based on permeation of formaldehyde from paraformaldehyde contained in a permeation tube. Two independent analytical methods, based on cavity ring down spectroscopy (CRDS) and Fourier transform infrared spectroscopy (FTIR) were used for the assignment procedure. Each participating institute was provided with one transfer standard and value assigned the formaldehyde mole fraction in the standard based on its own measurement capabilities. The stability of the formaldehyde mole fraction in transfer standards was deduced from repeated measurements performed at the BIPM before and after measurements performed at participating institutes. In addition, 5 control standards were kept at the BIPM for regular measurements during the course of the comparison. Temporal trends that approximately describe the linear decrease of the amount-of-substance fraction of formaldehyde in nitrogen in the transfer standards over time were estimated by two different mathematical treatments, the outcomes of which were proposed to participants. The two treatments also differed in the way measurement uncertainties arising from measurements performed at the BIPM were propagated to the uncertainty of the trend parameters, as well as how the dispersion of the dates when measurements were made by the participants was taken into account. Upon decision of the participants, the Key Comparison Reference Values were assigned by the BIPM using the largest uncertainty for measurements performed at the BIPM, linear regression without weight to calculate the trend parameters, and not taking into account the dispersion of dates for measurements made by the participant. Each transfer standard was assigned its own reference value and associated expanded uncertainty. An expression for the degree of equivalence between each participating institute and the KCRV was calculated from the comparison results and measurement uncertainties submitted by participating laboratories. Results of the alternative mathematical treatment are presented in annex of this report.