Estimating fractional rate of NDF degradation from in vivo digestibility

Fractional rate of degradation (kd) of potential degradable NDF (dNDF) was estimated based on in situ degradation profi les. In situ values were compared to values for kd which were calculated based on in vivo NDF digestibility and dNDF concentration for 61 forage samples including both young and very mature grasses and legumes. The calculation only failed for three samples, and in situ estimates correlated reasonably well to calculated values. Calculated kd for dNDF has a large potential for use in practice as a low resource method.


INTRODUCTION
Values for fractional rate of degradation (k d ) of potential degradable NDF (dNDF, % of dry matter (DM)) have so far mainly been based on NDF degradation profi les obtained by nylon bag or in vitro methods.However, for practical feed evaluation these methods are far too resource demanding (time, labour, costs) and dependent on access to fi stulated cows.Therefore alternative methods are necessary for practical use.For forage NDF, it is essential that NDF degradation parameters can be analysed or estimated on the actual sample.This means that a simple method is needed for estimation of k d based on biological/chemical laboratory methods.Further, only little information is available on in vivo validation of in situ values.The aim of the present paper is to compare k d for dNDF estimated from in situ incubations with k d which were calculated based on NDF digestibility measured in sheep fed at maintenance.

Material and analysis
Sixty one forage samples including grasses, legumes, whole crop cereals, whole crop maize and straw were analysed for in vivo NDF digestibility in sheep fed at maintenance (four sheep per forage sample) and for rumen degradability in situ (nylon bags).NDF degradation was performed using the in situ method according to Weisbjerg and Hvelplund (2005).Samples were freeze-dried, milled through 1.5 mm screen and incubated in nylon bags with pore size 37 µm (for 288 h 12 µm) for 0, 2, 4, 8, 24, 48 and 96 h in three dry cows fed a standard ration.As estimates for potential NDF degradability (pdNDF, % of NDF), degradabilities obtained after either 288 or 504 h rumen incubation were used.The reason for two methods was a change in the standard used in our laboratory.
After incubation residues were transferred quantitatively for NDF analysis (ash free) including amylase treatment.
Degradability was calculated relatively to 0 h residue to account for eventual particle losses, except for 288 h incubations where particle loss was negligible due to the small pore size in the nylon bag.Degradation profi le parameters were estimated using incubation times up to 96 h by PROC NLIN in SAS, using a simple exponential model including degradability asymptote and rate of degradation.Further, a model including a lag time was also run.

Calculation of k d
In vivo dNDF digestibility (D) can be found by combining NDF digestibility at maintenance and pdNDF.If D is known and assumed equal to rumen digestibility, and the rate of passage of dNDF is either measured or assumed, then the rate of dNDF degradation can be calculated.The simple one pool models for rumen passage behaviour, which do not account for the selective retention of newly ingested particles, underestimate effective degradation, and is therefore not suitable for this 'backwards' calculation.The two compartment model (Figure 1) by Allen and Mertens (1988) takes the selective retention of feed particles in the rumen into account: (Equation 1) This equation can be solved according to k d (Huhtanen et al., 2006): where k r and k p are fractional rates of passage from the non-escapable and escapable pool, respectively.Estimation of k d using this principle thus requires information on the rate of passage (k r and k p ), pdNDF and in vivo NDF digestibility in sheep fed at maintenance.Assuming a total mean retention time (MRT) in the rumen of the sheep fed at maintenance level of 50 h with 40% of the retention time in the non-escapable pool (Huhtanen et al., 2006) gives the values for k r and k p shown in Figure 1.However, this further relies on the assumption that rumen mean retention time and distribution of retention time between compartments are the same for all feed types, which can be questioned, but the assumption will probably only introduce a minor error.
Using the values from Figure 1 with total MRT of 50 h for obtaining k r and k p , Equation 2 result in the Equation 3. To illustrate the effect of MRT, calculations were also performed using a total MRT of 60 h with the same compartmental distribution, resulting in Equation 4. Equations 3 and 4 have been used for k d calculations in the present paper.k d = -0.041667+ [0.006944 + 0.0066667D/(1 -D)] 0.5 /2 (Equation 3) k d = -0.034722+ [0.004823 + 0.0046296D/(1 -D)] 0.5 /2 (Equation 4)

RESULTS
From the total of 61 samples, calculated k d could not be estimated for three samples (two red clover and one lucerne) due to higher measured in vivo NDF digestibility than measured in situ pdNDF.For the 58 samples with calculated k d , pdNDF varied from 0.410 to 0.936.Calculated k d varied from 0.019 to 0.211 per h, in situ degradation parameters for the model without lag time varied for k d from 0.011 to 0.186 per h and for the asymptote from 0.394 to 0.919.For the model with lag time in situ k d varied from 0.013 to 0.377 per h, asymptote from 0.394 to 0.910 and lag time from 0 to 6.8 h.
Plots of in situ k d against calculated k d are shown in Figure 2.For three very young grass/grass-clover crops very high in situ k d (0.16-0.19 per h, model without lag) were obtained.For calculated k d six samples of young crops obtained very high k d above 0.15 per h, as measured pdNDF was only slightly higher than measured in vivo digestibility of NDF for these samples.One sample of early cut grass-clover resulted in a very high in situ k d of 0.38 when estimated with a model including lag.The in situ k d from a model without lag (Figure 2a) were generally lower than calculated k d .Further, the variation around the regression line increased with increasing k d .In situ k d values from a model including lag time were more of the same magnitude as the calculated k d , with a regression coeffi cient of 0.89 and an intercept of -0.0006 very close to zero.Increasing MRT from 50 to 60 h generally increased the regression coeffi cient for both model types without changing R 2 .

DISCUSSION
The results show that it is possible to obtain reasonable degradation data from in vivo digestibility of NDF combined with dNDF concentration.Only for very young crops, where in vivo digestibility is close to pdNDF, random and method errors can induce situations where measured in vivo digestibility become higher than measured pdNDF and k d therefore cannot be calculated.Generally, k d estimated in situ using a model including lag time is more in accordance with k d calculated from in vivo data in a model without digestion lag time, indicating that the lag time found in nylon bags might be a shortcoming of the in situ method.Further, 60 h compared to 50 h for total rumen MRT used for calculating k d are in closer accordance with in situ data.
The calculation method for k d is limited by its very indirect approach where analytical and prediction errors can add up.Further, and maybe most critical, the in vivo digestibilities are often very close to the potential degradabilities, and therefore only small 'errors' on either dNDF or in vivo NDF digestibility have large infl uence on the estimated fractional rate of degradation.This is especially true for feed samples with high fractional rate of dNDF degradation, where only small analytical errors can result in dNDF digestibilities above 100%, and then no solution for k d can be found.
Sensitivity tests have shown that feed samples with high k d are very sensitive to an 'erroneous' overestimation of digestibility of dNDF, but less sensitive to underestimation.Sensitivity to 'errors' in ash and NDF concentration was found to be rather low.Samples with less degradable NDF are generally much less sensitive to 'errors' in analyses.
As mentioned above, the nylon bag method has been widely used to obtain NDF degradation parameters for modelling.There are, however, several indications showing that k d for dNDF found using the nylon bag method underestimate k d found using in vitro or in vivo methods (Huhtanen et al., 2006).This could be due to a less favourable micro-environment for NDF degradation in the nylon bags than in the surrounding rumen environment.
At the onset of NDF degradation there is often a delay (lag time), both with the nylon bag method and the in vitro method.Whether these in situ lag times are real or shortcomings of the in situ method or a mixture of both are not known.However, it is very important for the interpretation of degradation profi les to establish whether the lag times are real or not, as it highly affects the magnitude of the estimated k d .
Further, the fractional rate for dNDF degradation is not constant, as the name 'degradation rate constant' indicates.Besides the lag time, it seems, after the lag time has ended, that k d decreases as incubation time increases.However, this decrease is probably less problematic than the lag time for the interpretation of degradation profi les.But it is necessary with increased knowledge in this area to correctly interpret degradation profi les, and this implies that the k d found from in situ data cannot be taken as the absolute truth.

PRACTICAL IMPLICATIONS
For the use in feed evaluation on samples from practice, sheep digestibility experiments are far too resource-demanding.Using the Lucas principle, a true NDS (Neutral Detergent Solubles) digestibility of 101% and an endogenous loss of 90 g NDS per kg ingested dry matter (DM) were found (Weisbjerg et al., 2004a).When OM digestibility is known, this allows a prediction of NDF digestibility, as digested OM is equal to the sum of digested NDS and digested NDF.An evaluation of 2,337 observations showed an acceptable prediction of digested NDF in % of feed DM with no bias (Weisbjerg et al., 2004b).This principle enables an estimation of NDF digestibility from OM digestibility and NDF concentration, which both are routine analysis made in practical feed evaluation using in vitro methods and wet chemistry or near infra-red refl ectance spectroscopy (NIRs).When NDF digestibility and dNDF concentration are known, dNDF digestibility can be calculated.The advantage of this approach for estimation of fractional rate of degradation of dNDF is that it relies only on laboratory methods already used in practice.The method is at present used in practice in Denmark for estimation of k d for forages, where OM digestibility, and NDF and dNDF concentrations, are estimated using NIRS.

Figure 1 .
Figure 1.Model of two compartment passage of fi bre through the rumen with values for compartmental mean retention times (CMRT) and resulting passage rates used to calculate k d

Figure 2 .
Figure 2. Plot and regression of in situ k d from model without (a, c) or with (b, d) lag time against calculated k d .Total rumen MRT 50 h (a, b) and 60 h (c and d).Dotted line indicates y=x