Comparing three filter-bag types for accuracy and precision of in sacco undigestible neutral detergent fibre determination of various dicotyledon forages

Introduction

Undigestible neutral detergent fibre (uNDF) has become an analytical standard for several feed-evaluation systems to quantify the potential digestible NDF fraction of a feed (Sniffen et al. 1992; Volden 2011). Unlike, neutral detergent fibre (NDF), uNDF is a feed fraction with a digestibility of zero. Subtracting uNDF from NDF improves the predictability of the dry-matter digestibility (DMD) of a feed (Mertens 2015). Furthermore, performing total faecal collection for determination of DMD is not practical in most animal studies, and therefore dietary internal markers have been used to estimate apparent degradation after passing through the digestive tract, i.e. from faecal spot samples (Tamminga et al. 1989; Huhtanen et al. 1994). Undigestible NDF has been found to be a suitable internal maker for this purpose in several studies (Morris et al. 2018; Beck et al. 2023).

However, uNDF analysis is not commercially available in New Zealand and the in sacco nylon-bag incubation methods or in vitro methods are labour-intensive/expensive (Lund et al. 2004; Foster et al. 2023). The ANKOM F57 bags (ANKOM Technology, Macedon, NY, USA) are cheaper than nylon bags and can be used sequentially for both NDF analysis and rumen incubations and therefore reduce labour requirements and costs compared with using nylon bags or in vitro methods. Furthermore, the F57 bags can be used for boiling samples in neutral detergent solution in a cooking pot, which allows performing NDF analysis on a large number of samples at once (McRoberts and Cherney 2014), reducing the labour requirements for NDF analysis and, consequently, for uNDF analysis. However, the initial analysis with ANKOM F57 bags in our laboratory showed some challenges for uNDF determination, especially with forage rape that had a low fibre content (Della Rosa et al. 2022a), which might result in an overestimation of the DMD. (To note, acid detergent lignin was also below the detection limit when the same forage sample was analysed using the ANKOM F57 method by a commercial laboratory). Dicotyledon forages such as forage rape (and other forage brassicas), lucerne and white clover have different and less tensile fibre than do monocotyledons and typically are pulverised during grinding and turn into a dusty bulky fine powder, independent of grinding screen size. The relatively large pore size of 25 μm of the F57 bag might have allowed undigested small particles to escape from the bag (Damiran et al. 2008) and therefore reduce the uNDF value. ANKOM has released a new bag (F58) with a pore size of 6–9 μm for fibre analysis of samples that have had significant pulverization during grinding. Unlike the F57 bag, the F58 bag is not specifically designed for in vitro or in sacco rumen incubations, but the pore size of F58 bag is similar to that of the 12 μm of the Saatifil bag (Saati S.p.A., Veniano, Italy) that is recommended for uNDF analysis for the NorFor feed-evaluation system (Volden 2011). The objective of the current study was to compare the uNDF for various forages after 12 days of in sacco rumen incubation by using ANKOM F57 and F58 filter bags and Saatifil bags. Second, the effect of sequentially determining NDF and uNDF and bag sealing direction were tested for the two ANKOM bags. The hypothesis was that uNDF determined with the F58 would be similar to uNDF determined with the Saatifil bag, whereas uNDF determined with the F57 would be underestimated.

Materials and methods

Two in sacco rumen incubation experiments with a fistulated non-lactating non-pregnant Jersey dairy cow was conducted. The rumen incubations adhered to the guidelines of the 1999 New Zealand Animal Welfare Act and the AgResearch code of ethical conduct, and animal manipulations were approved by the AgResearch Animal Ethics Committee (#1933). The cow was grazing ryegrass-based pasture during the in sacco rumen incubations.

In Experiment 1, samples from different forage species (Table 1; some of which have known high (e.g. plantain) or low (e.g. forage rape) uNDF contents were compared with ryegrass (Della Rosa et al. 2022a; 2022b)) were weighed in sextuplicate into three filter-bag types, namely, Saatifil (12 μm pore size, 10 × 10 cm, fabric code PES 12/6), and ANKOM F57 and F58 bags (25 μm 6–9 μm pore size, respectively, dimensions of the trapezoid shaped bags were approximately 5 cm at the top, 3 cm wide at the bottom and 5 cm high). The bags were incubated for 12 days in the rumen of a fistulated non-lactating dairy cow grazing ryegrass-based pasture. Two other procedures also were tested in Experiment 1. A second set of F57 and F58 bags containing substrates was analysed for NDF followed by the determination of uNDF, and a third set of F57 and F58 bags containing substrates was heat-sealed with the sides of the bags pushed to each other (instead of the flat side of the bag) to maintain an gas pocket inside the bag (Fig. 1).

Fig. 1.

The Saatifil bag (left) and ANKOM F57 and F58 bags sealed on the flat sides (top middle and right) and side-by-side (bottom middle and right) to create a gas pocket.

In Experiment 2, 10 samples from dicotyledon forages (Table 1), which are known to result in fine powdery material after grinding, were used, including samples from six white clover cultivars to enable the determination of within-species variance. The forage samples were each weighed into sextuplicate bags for each bag type (Saatifil, F57 and F58) and incubated for 12 days in the rumen of one fistulated cow grazing ryegrass based pasture.

Results

Combined results of Experiments 1 and 2

The average uNDF (±pooled within sample CV%) for all forage samples across the two experiments was 5.4% (9.8), 6.7% (15.2) and 6.3% (16.0) of DM for Saatifil, F57 and F58, respectively. The mean bias for uNDF (±95% CI) relative to Saatifil was 1.3 (0.3–2.3) for F57 and 1.2 (−0.1–2.5) for F58 (Fig. 2). The mean squared prediction error (MSPE, % of DM) relative to Saatifil was 2.1% of DM for F57 and 2.7% of DM for F58. The 95% CI indicated that the intercept and slope for the orthogonal regression of Saatifil vs F57 and F58 were not different from zero and one, respectively, suggesting that the uNDF values of F57 and F58 were similar to Saatifil. The correlation of Saatifil with F57 was stronger than for F58 (r = 0.65 and 0.40, respectively).

Fig. 2.

Orthogonal regression (top graphs) and Bland–Altman plots (bottom grahps) of undigestible neutral detergent fibre (uNDF, % of DM) determined using the F57 or F58 bags compared with the Saatifil bag (left-hand and right-hand graphs, respectively) of various forages incubated in the rumen for 12 days across Experiments 1 and 2 (see Table 1). The Bland–Altman plot shows the mean bias (solid red line), 95% confidence interval for the mean bias (red dotted lines) and limits of agreement (solid black lines). In the orthogonal regressions, the 95% confidence interval for F57 versus Saatifil was −2.423–1.628 for the intercept and 0.585–1.139 for the slope and F58 versus Saatifil was −11.097–2.773 for the intercept and 0.448–2.581 for the slope.

Discussion

The hypothesis that uNDF determined with the F58 would be similar to the uNDF determined with the Saatifil bag and that uNDF determined with the F57 would be underestimated was not supported by findings of the current study. Instead, the F58 bag did not appear to result in improved uNDF determination compared with using the F57 bag, and both resulted in similar values as did the Saatifil bag; however, numerically, uNDF values were greater and with a larger within-forage CV for the two ANKOM bag types. Furthermore, sealing the ANKOM bags from side-to-side to create a air space in the bag or sequentially analysing uNDF after NDF analysis did not result in great changes in the uNDF determination with the F57 and F58 bags, compared with sealing the bags on the flat sides and without pre-NDF determination.

Although the F57 has a larger pore size (25 μm) than recommended (~12 μm), on the basis of findings with nylon bags (Volden 2011; Krizsan et al. 2015), results of the current study and by Norris et al. (2019) did not indicate that the F57 loses more undigested particles than does the F58 bag, which has a smaller pore size of 6–9 μm. This was similar to the findings of Norris et al. (2019), but different from the findings of Adams et al. (2020) who observed that uNDF was approximately 20% greater in F58 than in F57 bags. Casali et al. (2009) tested and observed no undigested particle losses from F57 bags with different types of feed, but they found particle losses from nylon bags with 50 μm pore size. Using electron microscopy, they observed that the fabric of the nylon bag had a geometric arrangement, whereas the fabric of the F57 bag had an irregular mesh arrangement, which might affect the optimum pore size for uNDF analysis. It was observed that the external layer of most of the F58 bags was peeled and rolled, retaining some particles after washing (M. M. Della Rosa, pers. obs.). This could have contributed to the greater within-sample CV observed for the F58 bags. Similar observations for F58 bags were described by Adams et al. (2020).

The uNDF values of clover and ryegrass with the three bag types fell largely within the range of 2.5–11.1% of DM across grass/clover samples from Denmark analysed according to Volden (2011) and reported in Krämer et al. (2012) and the NorFor feed tables (https://feedstuffs.norfor.info/), suggesting that values observed in this study were within the normal range. However, numerically the uNDF values were, on average, greater with the F57 bag during both experiments and F58 bag in Experiment 2 than with the Saatifil bag, suggesting that the true uNDF might not yet have been reached.

A factor found to affect uNDF was the sample size in the bag, with uNDF being ~10% less when the sample size was reduced from 40 to 20 mg/cm² (Norris et al. 2019). The sample size in the bag in the current study was approximately 12.5 mg/cm² for the two ANKOM bags and slightly less at 9.65 mg/cm₂ for the Saatifil bag. Fluffy, bulky samples can make the ANKOM bag appear very full and become compacted when the bag is sealed. Therefore, we tested whether sealing the ANKOM bags from side-to-side, which created an air space in the bag, would improve the uNDF determination. Especially in the F57 bag, it appeared to make the uNDF determination worse, both in terms of a numerically larger mean bias than in the Saatifil bag and a greater within-sample CV. Coblentz et al. (2019) compared incubating samples in F57 bags at 12.5 mg/cm² or 6.25 mg/cm² in DaisyII in vitro system and found that results with the smaller sample size were in closer agreement with results using the in vitro method, based on Goering and Van Soest (1970), which was based on incubating each sample in buffered rumen fluid in Erlenmeyer flasks, followed by NDS washing after incubation and filtering through crucibles. Therefore, reducing the sample size in the bag might further improve the accuracy of the uNDF determination.

Sequentially determining NDF and uNDF versus direct determination of uNDF had little effect on the accuracy and precision of the uNDF determination. In contrast to these findings, Battelli et al. (2020) found that DMD estimation in pigs was more accurate when using uNDF determination after NDF analysis than using uNDF directly. These suggest that sequential determination of NDF and uNDF can be employed.

Feed uNDF was found to be 10–29% less when incubating in the rumen for 576 h than when incubating 288 h (in F57 and F58 bags) (Norris et al. 2019; Adams et al. 2020). The faecal recovery of NDF was over 100% for both 288 and 576 h of incubation, as well as for the F57 and F58 bas in the study of Adams et al. (2020), which indicates that the true uNDF was still not reached, even after 576 h of incubation. This was mainly due to the uNDF of the feed, which reduced to a greater extent between 288 and 576 h of incubation than did the uNDF of faeces (Norris et al. 2019; Adams et al. 2020). In contrast, Beck et al. (2024) observed that the recovery of uNDF was near 100% when incubating F57 bags in the rumen for 288 h and this was shortened to 120 h if the feed was predigested with pepsin and HCl before rumen incubations. These studies did not compare the ANKOM bags with nylon-type filter bags with a small pore size. It is therefore not clear whether uNDF in these bags would further decrease when incubated for more than 288 h in the rumen of fistulated cows.

The within-sample CV (precision) was similar for the three bag types tested in the first experiment, but the pooled CV was less for the Saatifil bag in the second experiment than for the F57 and F58 bags. The similar pooled CV for the F57 and F58 bags in the current study was different from the findings of Adams et al. (2020) who observed a greater CV for F58 than for F57 bags. The within-sample CV of Experiment 1 was similar to that observed in Norris et al. (2019), but this was larger than that observed in Experiment 2 and that observed by Adams et al. (2020). This might be partly due to the feeds incubated and, in the current study, the bags were hand-washed after Experiment 1 while a washing machine was used after Experiment 2. However, Cherney et al. (1990) observed similar variance for filter bags washed by hand or with a washing machine, whereas residual DM was greater in hand-washed than machine-washed bags.

The clover samples in the F57 bags in Experiment 2 had a smaller within-sample CV than did the chicory and kale samples in the F57 bags in Experiment 2, suggesting that there might be a difference in the precision of the uNDF determination for different forages, as was also observed by Norris et al. (2019). The samples of chicory, forage rape and kale contained less NDF and less uNDF than did the clover samples and this results in a lower signal-to-noise ratio, which might increase the error (imprecision) of the uNDF determination. The signal-to-noise ratio can be improved by incubating a greater amount of sample, but the F57 bag comes only in one size, and, as indicated above, incubating a greater sample in the F57 is not advisable (Norris et al. 2019).

The within-sample CV for the NDF analysis of the feed samples was 6.0%, which also contributed to the error of the uNDF analysis. Norris et al. (2019) modelled that the number of bags incubated per sample needs to be increased by three- to four-fold to improve the precision to enable detection of differences between feeds or diets. Although there are many recommendations in the literature to improve the accuracy of the uNDF determination, there are very limited recommendations for improving the precision of the uNDF determination (Foster et al. 2023).

Implications

The F58 bag did not appear to result in improved uNDF determination compared with using the F57 bag, and both resulted in similar values as with the Saatifil bag; however, numerically uNDF values were greater and with a large within-sample CV for the two ANKOM bag types. The results of the current study suggest that some further modifications to the protocol need to be made to improve the accuracy and precision of the uNDF determination using the F57 bag. The F57 bag can be used sequentially (i.e. weigh samples only once) to determine the different fibre fractions (NDF and uNDF) and using the simple kitchen pot method for NDF determination, making it a more practical alternative than uNDF determination with the Saatifil bag.