We performed spectrophotometric analysis using a GENESYS 10S UV-vis brand single-beam UV-visible spectrometer (Series: 2M1N299002, with a xenon lamp), an Agilent 8453 brand double-beam UV-visible spectrophotometer (2140-0605, with deuterium lamp), and a 1 cm-wide quartz cell. The analytical balance we used for weighing was Gram FV-220C (IPESAGE S.A.S, France). A HANNA brand HI 9126 pH meter was used.

The following chemicals were used as reagents to perform this method: concentrated hydrochloric acid (37%) for analysis of brand univ AR^® (SAAR3063054LCA), concentrated sulfuric acid (98%) for analysis from Merck (CAS 7664-93-9/112080), oxytetracycline RS (96%) from VWR Chemicals (CAS 2058-46-0/6153-64-6), and distilled water from the Medicines and Foodstuffs Quality Control Laboratory (LACOMEDA) of the Faculty of Pharmaceutical Sciences of the University of Kinshasa.

The following conditions were determined for the analysis of oxytetracycline using a UV spectrophotometer: 0.01 N hydrochloric acid as the dissolution solvent; 268 nm as the working wavelength; 5 μg/ml as the target concentration (after three repeated measurements with precision and specificity to the analyte) and a blank which constituted 0.01 N hydrochloric acid.

In this research, we defined the calibration range from 5 to 25 μg/ml. Over this interval, we evaluated the different validation criteria of the analytical methods. These criteria focused on accuracy, precision in terms of repeatability and inter-series precision, measurement uncertainty, accuracy profile, and linearity (using a range of concentrations with five points). The validation was carried out at three concentration levels, with three repetitions for 3 days.

For the first studied criterion of the validation, we assessed the specificity/selectivity of the developed method by scanning both the standard calibration and the standard of validation solutions prepared at the target concentration level with different wavelengths using the DAD. This was to determine, within the interval of between 200nm and 400 nm, what the maximal absorption wavelength would be when only the analyte could absorb UV. The matrix effect was then suppressed from the whole solution to avoid its impact during measurements. The specific wavelength was established at 268 nm for selective determination without the detection of impurities, matrix, or degradation products.

Based on ICH guidelines Q2, we assessed the interference influence using the matrix components. The validation standard was scanned within the wavelength 200–400 nm and no excipient was absorbed at the targeted wavelength (268 nm). Some of the components showed a significant absorption over the target. Povidone was absorbed at 365 nm, but sodium hydroxy methane sulfinate, magnesium oxide, and methyl-1 pyrrolidone did not show significant absorption. After scanning, all excipients’ influence was suppressed to perform analysis at the analyte's maximal absorption wavelength.

Table 1 illustrates the results of calculated bias at the different concentration levels. Based on these results, we found that the absolute bias at all levels was between −0.0066 and +0.0313, and the relative bias (%) was between −0.0333 and +0.2888. As these results were within specified limits included in the [-2%, +2%] interval, we concluded that the estimated bias at each level of concentration was acceptable.

The acceptance limits for the relative bias are ±2% for a drug product assay (International Organization for Standardization, 2017).

Table 2 shows that the recovery rate (%) is 98.09–101.7 using the 95%–105% confidence interval.

Thus, by observing the results of Tables 1 and 2, we can determine that the developed oxytetracycline assay method is true and can give appropriate results at the proposed concentration levels.

Table 3 groups the method’s precision results in terms of repeatability and intermediate precision. This was assessed by determining the three replicates of the three sample solutions obtained separately at different levels of concentration. Their preparations were made to be close so that the repeatability was constant across the concentration range. The acceptance limits for the relative bias are ±2% for a drug product assay (International Organization for Standardization, 2017; USP XL, 2020). The intermediate precision was also studied to evaluate the effect of random events on the method precision. For this purpose, we considered additional parameters such as performing analysis on different days, the use of a secondary UV-spectrophotometer different from the first, and the method used by different analysts. Additionally, the intermediate precision afforded acceptable values, comprised within [-3%, +3%] as required for the drug product assay (USP XL, 2020).

The relative standard deviation for the repeatability at the different concentration levels was between 0.471 and 1.162. We found that, at all concentration levels, the repeatability RSD was less than 2% (International Organization for Standardization, 2017). This reflects the low intra-day dispersion of the results obtained.

Moreover, the RSD for intermediate precision at the different levels was between 1.762 and 2.192 at all concentration levels, and the intermediate precision RSD was less than 3% (International Organization for Standardization, 2017).

Observing these results, we can therefore determine that the method is acceptable in the dosage range of 5–25 μg/ml according to the precision criterion.

According to the general approach, uncertainty evaluation requires estimating the random and systematic errors of the method. This should provide estimates of the within- and between-lab components of variance combined with an estimate of uncertainty coupled with the method trueness (Bratinova et al., 2009). Therefore, in this research, we determined random error separately using within-laboratory reproducibility and systematic error, and then calculated the uncertainty following the formula given in the literature.

Table 4 presents the results of the measurement uncertainty calculation. The uncertainty of a measurement, is broadly defined, as the area within which the true value is likely to lie. The relative expanded uncertainties (%) observed at the different concentration levels are between +3.041 and +4.961. We found that, at all concentration levels, the relative expanded uncertainty was less than 5% (International Organization for Standardization, 2017).

As presented in the Figure 1 and Table 5, the method accuracy profile was etablished on three concentration levels. This criterion allowed us to assess the difference between each individual result and the accepted reference value from both the random and systematic errors. The calculated results used for calculating beta tolerance were first multiplied by an appropriate factor and then utilized for accuracy determination, as given in the following table.

The method accuracy was assessed by determining the recovery using well-known analyte concentrations in an appropriate matrix. As we developed a UV-vis method, the assay results obtained were compared with those from an established analytical method for more certainty. For drug product assay, the acceptance criteria are fixed between 95.0% and 105.0% (USP XL, 2020).

The β tolerance intervals were located within the acceptance limits, referring to the linear regression model, the risks of having measurements outside the acceptance limits observed at the different concentration levels were between +0.3110 and +3.157. The literature indicates that the risk that the β tolerance intervals are outside the acceptance limits should not exceed 5% (International Organization for Standardization, 2017).

We found that the risk is less than 5% at all levels of concentration; thus, the developed method is accurate and fits analytical determination at all levels of concentration.

Figure 2 shows the linearity curve of the oxytetracycline assay method by UV-Visible.

In Figure 2, we observe that the determination coefficient (R ²) of the developed method is equal to 0.9997. This value is close to the values determined by Prasad (2010), Sebayim and hasson (2019), and Naumoska et al. (20122)—0.9994, 0.9999, and 1, respectively.

Linearity is proven when R ² is greater than 0.99 (International Organization for Standardization, 2017); thus the linearity of our method is proven.

For a better appreciation of this linearity, we used the linearity profile based on the tolerance interval and acceptance limits shown in Figure 3.

As can be seen in Figure 3, which presents the linearity profile of the method, the tolerated interval represented by the broken blue lines is inside the acceptance limits defined by the black dotted lines; thus, the linearity of this method is proven.

The validated method was then successfully applied to 47 samples, divided into three doses (50 mg/ml, 100 mg/ml, and 200 mg/ml), 18 batches which were manufactured by eight pharmaceutical companies in five countries.

All the results were calculated following the Beer-Lambert law to express the analyte concentration thus: A = l × ε × c where A is the analyte absorbance, ε is the specific absorption coefficient, and l is the optical trajectory distance.

The assay was performed with UV-visible spectrophotometry. However, this method requires precise dilutions and a good cleaning of the cuvette, as well as avoiding any pipetting error. For this study, the acceptable content of oxytetracycline should be between 90% and 110% (Naumoska et al., 2012). Based on these limits, out of the 47 samples analyzed (Table 6), 28 samples were compliant (i.e. 59.57%) against 19 samples non-compliant (i.e. 40.43%).

The cases of non-compliance observed after the determination of the active ingredient using the developed and validated method were under-dosing (18 samples, i.e. 38.3%) and over-dosing (01 sample, i.e. 2.13%).

Studies carried out on the quality control of injectable oxytetracycline for veterinary use in Mali revealed a non-compliance rate of 41.86% (15.38% of these non-compliant samples presented a low content of active ingredient) for N'DONG et al. and 10.83% for DICKO Sako A.A., while that carried out in Niamey in Niger on the quality control of antibiotics by Amadou et al. showed a non-compliance rate (under dosage) of 14.28% (Dicko, 2010; USP, 2019).