The plant material consisted of eight bean varieties supplied by the National Centre for Agricultural Research of Burundi and distributed and recommended in the Kirimiro region. Among them, three are twining (Kinure ndende,Kinure ngufi,and Muhoro), and five are semi-volubles (Mugwiza(RWR2154),Musengo(MLB122-94B),Rufutamadeni,Magorori, and Mukungugu). Overall, three varieties are conventional (Rufutamadeni, Kinure ndende, and Kinure ngufi) and five are bio-fortified (Mugwiza, Musengo, Muhoro, Magorori, and Mukungugu).

The trial was conducted using a randomized complete block design (RCBD), covering eight treatments, each corresponding to a bean variety (Phaseolus vulgaris L.). Due to the smaller quantity of seeds provided by the research center (ISABU), each treatment was repeated three times, resulting in a total of 24 elementary plots. The total area used for the experiment was 34.71 ares, corresponding to an average area of 1.44 ares per elementary plot.

The plots were set up with a spacing of 0.75 m between blocks and 0.50 m between plots to limit edge effects and interactions between treatments. The orientation and arrangement of the blocks were established to reduce soil fertility and moisture gradients, in accordance with the principles of agricultural experimentation.

Randomization of treatments within each block was performed randomly to ensure statistical independence and minimize bias. This design enables a rigorous assessment of intervarietal variability while controlling for the effects of environmental heterogeneities.

This experiment was conducted under natural climate conditions and during the rainy season (season B). Sowing was carried out in rows with two seeds per pocket for both climbing and bush beans. The spacing was 40 cm between rows, 20 cm between pockets for bush beans, and 50 cm between rows and 20 cm between pockets for climbing beans [27].

Manure was applied as a base dressing at a rate of 10 tons per hectare, while organo-mineral fertilizer FOMI-IMBURA (NPKCaMg: 9-22-4-13-2) was used at a rate of 150 kg per hectare for voluble and semi-voluble beans, and 10 tons per hectare of manure and 100 kg of organo-mineral fertilizer FOMI-IMBURA for dwarf beans during sowing as indicated by ISABU [24][27]. Weeding was carried out 3 weeks after sowing, and regular weeding was carried out thereafter to maintain the trials’ permanent cleanliness, and no pest or plant disease was declared. The stakes were installed before the bean showed its ability to climb. Harvesting was carried out at the physiological maturity of the seeds, which occurred at an average of 92 days.

For the present study, we collected agronomic and consumer acceptability data. For agronomic data, they were collected throughout the cropping cycle using standard measurement protocols. Both qualitative and quantitative variables were taken into account. Qualitative variables concern the description, management, and site characteristics, including fertilization type (mineral or organo-mineral), soil type (loamy or sandy), and previous land occupation (cereal or fallow). Quantitative variables included the emergence rate of the plant, plant height (cm), number of leaves per plant, number of flowers per plant, number of pods per plant, pod length (cm), and 100-seed weight (g). Measurements were taken on representative plants per plot, and mean values were computed for each variety.

To compare the yield of conventional and biofortified beans, we first need to describe statistical variables, and after, we used the Shapiro-Wilk test of normality because the sample size was small [28]. Because the simple size doesn’t obey the Central limit theorem (n < 30) [29], nonparametric tests (Kruskal-Wallis, Mann-Whitney) are valid for comparing the yield across bean varieties or categories [30].

For assessing the influence of morphological traits and yield components on yield, we used a stepwise Ordinary Least Squares to include variables that are as orthogonal as possible to each other [31]. The global model used is the following:

where:

Y i is the yield of the bean; X i the parameters taken into account that influence the yield.

Multiple tests were conducted to verify the model’s robustness. The regression specification error test of [32] was performed to detect general functional form misspecification. The normality and constant variance (homoscedasticity) of the residuals were also verified with the Shapiro-Wilk test [33] and the Breusch-Pagan test [34]. The significance of the coefficient was verified with the Wald test [35].

To assess consumers’ acceptability of nutrition perceptions, Fisher’s exact test was used.

Following the bean harvest in the 24 randomized blocks, a sample of 30 participants was randomly selected to take part in a sensory evaluation designed to analyze consumer perceptions of eight dishes prepared from the harvests of the eight bean varieties. The evaluation focused on four predefined criteria: taste, growing cycle length, cooking time, and texture. A rating grid was used, with a score scale from 0 to 10 for each criterion and for each dish, based on consumer appreciation. Each participant evaluated all eight dishes. The individual scores were then aggregated, and the averages calculated to obtain an average score for each criterion, which served as the basis for the analysis of consumer perception. From the average scores obtained, a Likert scale was appropriately constructed for all criteria.

During the experimentation, we grew eight varieties of beans distributed in biofortified and conventional beans. Among the eight, we had five (5) biofortified bean varieties, namely: Magorori, Mugwiza, Muhoro, Mukungugu, and Musengo. For the conventional ones, we had three varieties, namely: Kinure ndende, Kinure ngufi, and Rufutamadeni. The following table shows the descriptive statistics and mean comparison of the bean yield.

Results of Table 1 indicate that, except Mukungugu, no other variety reached its potential yield. The observed mean yields (639 - 1585 kg/ha) remain below the potential levels (1000 - 2000 kg/ha) [27]. This can be explained by the drought, which occurred during the first period of sewing. The student’s test of mean comparison between potential and observed yields across bean varieties indicated significant yield gaps for Kinure ndende, Kinure ngufi, Magorori, Musengo, and Rufutamadeni, demonstrating that these varieties performed well below their genetic potential (p < 0.05). In contrast, Mugwiza, Muhoro, and Mukungugu showed no significant differences, suggesting better environmental adaptation or stable performance under prevailing conditions. For Mukungugu, even if it goes above its potential yield, the difference is not significant. To compare the mean yield among varieties, a Kruskal-Wallis test was performed. The p-value of the Chi2 test shows that there’s no significant difference between the bean yield of different grown varieties. Notwithstanding, Mukungugu recorded the highest rank sum while Rufutamadeni recorded the lowest rank sum.

Table 1. Description of the yield of different bean varieties.

For comparing the bean yield between categories, the Mann-Whitney test was used. The results show us that the p-value (0.31) is greater than 0.05. This indicates that the yields of the two categories of beans are statistically equivalent. These results indicate that biofortification doesn’t imply an increase in yield. Though the difference is not marked enough to be confirmed by this test, the results indicate a trend in favour of biofortified beans.

To compare the acceptability of biofortified and conventional bean varieties according to the sensory, growing time, and practical attributes such as taste, texture, cooking time, growing period, and ease of preparation, we used Fisher’s exact test [36]. This analysis aimed to identify whether consumers perceived significant differences between the two categories of beans in terms of organoleptic quality and preparation characteristics.

The results of the table below (Table 2) show a strong and significant association for the taste (p = 0.009) and for the cooking time (p = 0.000). These results indicate a clear difference between these two criteria for the two categories of beans. The biofortified beans have a better taste than the conventional ones. For the cooking time, the conventional beans are distinguished by their very long cooking time, longer than the biofortified beans, which indicates that they need more energy for cooking. On the other hand, no significant association was found for the texture, growing period, and ease of preparation.

Table 2. Consumers’ perception analysis.

Among the variables, we have qualitative and quantitative ones. They have to be described separately.

Table 3. Description of quantitative variables.

Based on Table 3, only a few variations were found across the majority of criteria when the morphological characteristics of conventional and biofortified bean cultivars were compared. The two groups’ emergence rates (68.0% vs. 68.3%) were almost the same, indicating similar seed vigor and germination ability. The average pod length (12.93 cm vs. 10.67 cm) and number of seeds per pod (5.27 vs. 5.00) of biofortified beans were marginally greater, suggesting the possibility of improved grain filling and reproductive efficiency. On the other hand, conventional beans showed slightly higher plant height (119.6 cm vs. 110.7 cm) and 100-seed weight (42.67 g vs. 41.20 g), indicating stronger vegetative development. The Collar diameter was similar between the two groups (2.13 vs. 2.33 cm), and they had slightly more leaves per plant (34 vs. 26) but fewer flowers per plant (14.9 vs. 18.2), indicating a balance between vegetative and reproductive allocation. And finally, biofortified beans had slightly more pods per plant (8.07 vs. 7.33). Biofortified beans showed agronomic equivalency to conventional beans while maintaining competitive morphological traits and potential yield advantages.

For qualitative variables, based on Table 4, most of the tested plots were maintained using organo-mineral fertilizer (66.7%), while only 33.3% used just mineral manure, based on the descriptive analysis of production parameters. This indicates that farmers prefer integrated soil fertility management practices. The higher use of nutritionally enhanced genotypes promoted by agroecological programs is evident, as most plots (62.5%) were planted with biofortified bean varieties. Loamy soils constituted the majority of the samples (62.5%), with sandy soils making up 37.5%. These results suggest that the soil conditions were generally suitable for bean cultivation. Most fields (79.2%) had previously been planted with cereal crops, while only 20.8% were left fallow, implying short periods for soil regeneration. The production environment seems characterized largely by intensive and integrated management systems, which may influence the variability in bean yields and soil fertility responses.

Table 4. Description of qualitative variables.

The variables included in the model were selected automatically using a stepwise method, which adds explanatory variables that are statistically significant at 10 percent, ensuring a more parsimonious and robust final model. The findings reveal that, instead of reproductive features (pod number, flower number), vegetative morphological traits (plant height, Collar diameter, leaf number) have a greater effect on bean output. Using biofortified cultivars and applying organo-mineral fertilizers are examples of cultural practices that have been shown to increase yield. Overall, the model is statistically valid and well specified (R² = 0.857; satisfactory RESET test of omitted variable: p = 0.668, test of multicollinearity VIF = 2.24, Shapiro-Wilk test of normality of the residuals: p = 0.844, Breusch-Pagan tests of heteroscedasticity: p = 0.389, and Wald test for testing that at least one of the coefficients is different from zero: p = 0.000), confirming that the combination of strong vegetative vigor and sustainable cultural practices is the main driver for improving bean yield in the studied context. The following table illustrates the results of the linear regression model.

Table 5. Determinants of bean yield.

The results of Table 5 demonstrate that plant morphological factors have a significant impact on bean output. Specifically, there are positive and substantial impacts (p < 0.05) on plant height, crown diameter, and leaf number, suggesting that plant vigor is essential for productivity, particularly through improved photosynthetic capacity and efficient transfer of assimilates to reproductive organs. For the plant height, a one-centimeter increase in the plant’s height contributes a yield increase of 5.367 kg per ha, which is a strongly significant contribution. An increase in the Collar diameter also contributes strongly (p = 0.000) to the bean yield. An additional centimeter to the Collar diameter of the beans influences an increase of the yield of 747.017 kg per ha. The number of leaves per plant positively influences the bean yield. The results show that for an increase of one leaf in the average number of leaves per plant of a field, a significant increase of 5 kg in the yield can be reached.

On the other hand, yield is negatively impacted by the average number of flowers (p < 0.01), indicating that excessive blooming creates internal competition that restricts pod formation and filling. The average pod length has a little favorable impact (p = 0.056), which is in line with the trait’s direct contribution to seed production.

Although they are only slightly significant (p < 0.10), agronomic techniques (organo-mineral manure use) and the selection of biofortified cultivars also have a favorable impact on yield. These patterns demonstrate how agroecological advancements have enhanced soil fertility, allowing better cultivars to reveal their full genetic potential.