Robert John Tempelman

Meta-analyses have become increasingly common, providing meaningful summaries of cumulative knowledge in the dairy science literature. Some of the corresponding meta-analytic techniques have been developed by knowledgeable dairy scientists, some of which pre-date tractable likelihood-based random or mixed effects model meta-analytic techniques and associated software developed by statisticians. This review compares various meta-analytic techniques on aggregate data (i.e., study-specific treatment or slope estimates and their standard errors) generated from simulated data involving regression, completely randomized designs (CRD), and Latin square design scenarios. In all cases, meta-estimates generated from the analysis of individual performance data (IPD), using the same statistical model as that used to simulate the data, were considered to be gold standard references for meta-estimates derived from various meta-analysis strategies on aggregate data. In all cases, likelihood-based techniques outperformed techniques developed by dairy scientists for meta-estimate proximity to corresponding IPD estimates. An extensive simulation study comparing meta-analytic techniques within a CRD framework indicated that these advantages widen with increasing study heterogeneity in effect sizes, smaller number of experimental replicates (i.e., cows) per treatment per study, and lower within study variability; nevertheless, the impact of meta-analytic methods on estimated standard errors of these meta-estimates were rather trivial. To best utilize aggregate data from Latin square studies in meta-analyses, a concerted effort is required to recover standard errors of mean differences rather than the standard errors of the means themselves. Perhaps the most compelling reason for choosing likelihood-based methods for meta-analysis is their ability to provide reliable prediction intervals on effect sizes noting that these intervals are currently under-reported in the dairy science literature. Compared with the reporting of meta-estimates and their standard errors, prediction intervals provide a far more appropriate indication of uncertainty on treatment effects in future studies and should be greater emphasized in extension or outreach efforts. Although R software packages like metafor are readily available for likelihood-based methods, both SAS and R code for linear mixed models can be readily modified to facilitate these analyses as demonstrated extensively in the supplementary materials of this review.

Inorganic sources of Mg are commonly used in dairy cow diets, but their availability varies significantly. This study assessed the relative availability of 4 commonly used inorganic Mg sources and a novel alkalinizing proprietary mineral blend [PMB; Multesium (GLC Minerals, LLC, Green Bay, WI, USA)]. The study was a duplicated 6 × 6 Latin square, with 12 nonlactating, non-pregnant Holstein dairy cows assigned to a square based on BW and parity. Cows were fed 90% of their voluntary DMI (diet contained 0.21% Mg). Each experimental period lasted 7 d. On d 2 of each period, urinary catheters were fitted. Total urine collection started on d 3 for 48 h with samples collected and measured every 12 h. On d 4, 30 g of Mg were administered as boluses with gelatin capsules: negative control (one empty capsule), magnesium oxide (MgO), magnesium sulfate (MgSO₄), calcium magnesium hydroxide [CaMg(OH)₄], calcium magnesium carbonate [CaMg(CO₃)₂], and PMB [a blend of Ca and Mg sources that includes CaMg(CO₃)₂, CaMg(OH)₄, and MgO]. Blood samples were collected at 0, 1, 2, 3, 12, and 24 h after treatment administration on d 4 of each treatment period. Urine and blood samples were analyzed for Mg and Ca concentration. Statistical analyses were conducted with PROC GLIMMIX including treatment, time, period, square, treatment × time, treatment × period, and time × period as fixed effects, and cow nested within square as a random effect in the model. Urinary Mg excretion for 4 of the Mg sources studied [PMB, MgO, CaMg(OH)₄, and MgSO₄] increased significantly, representing an increase of at least 40.8% relative to control. The supplementation of CaMg(CO₃)₂ did not significantly increase relative to control. There were no significant changes in blood Mg concentration with treatment; but, a significant treatment × time effect was observed. Calcium-rich sources [PMB, CaMg(OH)₄, CaMg(CO₃)₂] had lower blood Mg concentrations at 12 or 24 h after treatment than control and CaMg(CO₃)₂. Based on urinary Mg excretion 24 h after treatment, 4 of the Mg sources evaluated (including PMB) showed a similar availability, however, the availability of the commercial CaMg(CO₃)₂ source included in our study was similar to the negative control (no-supplemented cows).

Feed efficiency is important for economic profitability of dairy farms; however, recording daily DMI is expensive. Our objective was to investigate the potential use of milk mid-infrared (MIR) spectral data to predict proxy phenotypes for DMI based on different cross-validation schemes. We were specifically interested in comparisons between a model that included only MIR data (model M1); a model that incorporated different energy sink predictors, such as body weight, body weight change, and milk energy (model M2); and an extended model that incorporated both energy sinks and MIR data (model M3). Models M2 and M3 also included various cow-level variables (stage of lactation, age at calving, parity) such that any improvement in model performance from M2 to M3, whether through a smaller root mean squared error (RMSE) or a greater squared predictive correlation (R²), could indicate a potential benefit of MIR to predict residual feed intake. The data used in our study originated from a multi-institutional project on the genetics of feed efficiency in US Holsteins. Analyses were conducted on 2 different trait definitions based on different period lengths: averaged across weeks versus averaged across 28 d. Specifically, there were 19,942 weekly records on 1,812 cows across 46 experiments or cohorts and 3,724 28-d records on 1,700 cows across 43 different cohorts. The cross-validation analyses involved 3 different k-fold schemes. First, a 10-fold cow-independent cross-validation was conducted whereby all records from any one cow were kept together in either training or test sets. Similarly, a 10-fold experiment-independent cross-validation kept entire experiments together, whereas a 4-fold herd-independent cross-validation kept entire herds together in either training or test sets. Based on cow-independent cross-validation for both weekly and 28-d DMI, adding MIR predictors to energy sinks (model M3 vs. M2) significantly (P < 10^-10) reduced average RMSE to 1.59 kg and increased average R² to 0.89. However, adding MIR to energy sinks (M3) to predict DMI either within an experiment-independent or herd-independent cross-validation scheme seemed to demonstrate no merit (P > 0.05) compared with an energy sink model (M2) for either R² or RMSE (respectively, 0.68 and 2.55 kg for M2 in herd-independent scheme). We further noted that with broader cross-validation schemes (i.e., from cow-independent to experiment-independent to herd-independent schemes), the mean and slope bias increased. Given that proxy DMI phenotypes for cows would need to be almost entirely generated in herds having no DMI or training data of their own, herd-independent cross-validation assessments of predictive performance should be emphasized. Hence, more research on predictive algorithms suitable for broader cross-validation schemes and a more earnest effort on calibration of spectrophotometers against each other should be considered.

Large datasets allow estimation of feed required for individual milk components or body maintenance. Phenotypic regressions are useful for nutrition management, but genetic regressions are more useful in breeding programs. Dry matter intake records from 8,513 lactations of 6,621 Holstein cows were predicted from phenotypes or genomic evaluations for milk components and body size traits. The mixed models also included DIM, age-parity subclass, trial date, management group, and BW change during 28- and 42-d feeding trials in mid lactation. Phenotypic regressions of DMI on milk (0.014 ± 0.006), fat (3.06 ± 0.01), and protein (4.79 ± 0.25) were much less than corresponding genomic regressions (0.08 ± 0.03, 11.30 ± 0.47, and 9.35 ± 0.87, respectively) or sire genomic regressions multiplied by 2 (0.048 ± 0.04, 6.73 ± 0.94, and 4.98 ± 1.75). Thus, marginal feed costs as fractions of marginal milk revenue were higher from genetic than phenotypic regressions. According to the ECM formula, fat production requires 69% more DMI than protein production. In the phenotypic regression, it was estimated that protein production requires 56% more DMI than fat. However, the genomic regression for the animal showed a difference of only 21% more DMI for protein compared with fat, whereas the sire genomic regressions indicated approximately 35% more DMI for fat than protein. Estimates of annual maintenance in kilograms DMI/kilograms BW per lactation were similar from phenotypic regression (5.9 ± 0.14), genomic regression (5.8 ± 0.31), and sire genomic regression multiplied by 2 (5.3 ± 0.55) and are larger than those estimated by the National Academies for Science, Engineering, and Medicine based on NE_L equations. Multiple regressions on genomic evaluations for the 5 type traits in body weight composite (BWC) showed that strength was the type trait most associated with BW and DMI, agreeing with the current BWC formula, whereas other traits were less useful predictors, especially for DMI. The Net Merit formula used to weight different genetic traits to achieve an economically optimal overall selection response was revised in 2021 to better account for these estimated regressions. To improve profitability, breeding programs should select smaller cows with negative residual feed intake that produce more milk, fat, and protein.

The Resilient Dairy Genome Project (RDGP) is an international large-scale applied research project that aims to generate genomic tools to breed more resilient dairy cows. In this context, improving feed efficiency and reducing greenhouse gases from dairy is a high priority. The inclusion of traits related to feed efficiency (e.g., dry matter intake [DMI]) or greenhouse gases (e.g., methane emissions [CH₄]) relies on available genotypes as well as high quality phenotypes. Currently, 7 countries (i.e., Australia, Canada, Denmark, Germany, Spain, Switzerland, and United States) contribute with genotypes and phenotypes including DMI and CH₄. However, combining data are challenging due to differences in recording protocols, measurement technology, genotyping, and animal management across sources. In this study, we provide an overview of how the RDGP partners address these issues to advance international collaboration to generate genomic tools for resilient dairy. Specifically, we describe the current state of the RDGP database, data collection protocols in each country, and the strategies used for managing the shared data. As of February 2022, the database contains 1,289,593 DMI records from 12,687 cows and 17,403 CH₄ records from 3,093 cows and continues to grow as countries upload new data over the coming years. No strong genomic differentiation between the populations was identified in this study, which may be beneficial for eventual across-country genomic predictions. Moreover, our results reinforce the need to account for the heterogeneity in the DMI and CH₄ phenotypes in genomic analysis.

Feed efficiency has become an increasingly important research topic in recent years. As feed costs rise and the environmental impacts of agriculture become more apparent, improving the efficiency with which dairy cows convert feed to milk is increasingly important. However, feed intake is expensive to measure accurately on large populations, making the inclusion of this trait in breeding programs difficult. Understanding how the genetic parameters of feed efficiency and traits related to feed efficiency vary throughout the lactation period is valuable to gain understanding into the genetic nature of feed efficiency. This study used 121,226 dry matter intake (DMI) records, 120,500 energy-corrected milk (ECM) records, and 98,975 metabolic body weight (MBW) records, collected on 7,440 first-lactation Holstein cows from 6 countries (Canada, Denmark, Germany, Spain, Switzerland, and the United States), from January 2003 to February 2022. Genetic parameters were estimated using a multiple-trait random regression model with a fourth-order Legendre polynomial for all traits. Weekly phenotypes for DMI were re-parameterized using linear regressions of DMI on ECM and MBW, creating a measure of feed efficiency that was genetically corrected for ECM and MBW, referred to as genomic residual feed intake (gRFI). Heritability (SE) estimates varied from 0.15 (0.03) to 0.29 (0.02) for DMI, 0.24 (0.01) to 0.29 (0.03) for ECM, 0.55 (0.03) to 0.83 (0.05) for MBW, and 0.12 (0.03) to 0.22 (0.06) for gRFI. In general, heritability estimates were lower in the first stage of lactation compared with the later stages of lactation. Additive genetic correlations between weeks of lactation varied, with stronger correlations between weeks of lactation that were close together. The results of this study contribute to a better understanding of the change in genetic parameters across the first lactation, providing insight into potential selection strategies to include feed efficiency in breeding programs.

Resilience can be defined as the capacity to maintain performance or bounce back to normal functioning after a perturbation, and studying fluctuations in daily feed intake may be an effective way to identify resilient dairy cows. Our goal was to develop new phenotypes based on daily dry matter intake (DMI) consistency in Holstein cows, estimate genetic parameters and genetic correlations with feed efficiency and milk yield consistency, and evaluate their relationships with production, longevity, health, and reproduction traits. Data consisted of 397,334 daily DMI records of 6,238 lactating Holstein cows collected from 2007 to 2022 at 6 research stations across the United States. Consistency phenotypes were calculated based on the deviations from expected daily DMI for individual cows during their respective feeding trials, which ranged from 27 to 151 d in duration. Expected values were derived from different models, including simple average, quadratic and cubic quantile regression with a 0.5 quantile, and locally estimated scatterplot smoothing (LOESS) regression with span parameters 0.5 and 0.7. We then calculated the log of variance (log-Var-DMI) of daily deviations for each model as the consistency phenotype. Consistency of milk yield was also calculated, as a reference, using the same methods (log-Var-Milk). Genetic parameters were estimated using an animal model, including lactation, days in milk and cohort as fixed effects, and animal as random effect. Relationships between log-Var-DMI and traits currently considered in the US national genetic evaluation were evaluated using Spearman's rank correlations between sires' breeding values. Heritability estimates for log-Var-DMI ranged from 0.11 + 0.02 to 0.14 + 0.02 across models. Different methods (simple average, quantile regressions, and LOESS regressions) used to calculate log-Var-DMI yielded very similar results, with genetic correlations ranging from 0.94 to 0.99. Estimated genetic correlations between log-Var-DMI and log-Var-Milk ranged from 0.51 to 0.62. Estimated genetic correlations between log-Var-DMI and feed efficiency ranged from 0.55 to 0.60 with secreted milk energy, from 0.59 to 0.63 with metabolic body weight, and from 0.26 to 0.31 with residual feed intake (RFI). Relationships between log-Var-DMI and the traits in the national genetic evaluation were moderate and positive correlations with milk yield (0.20 to 0.21), moderate and negative correlations with female fertility (-0.07 to -0.20), no significant correlations with health and longevity, and favorable correlations with feed efficiency (-0.23 to -0.25 with feed saved and 0.21 to 0.26 with RFI). We concluded that DMI consistency is heritable and may be an indicator of resilience. Cows with lower variation in the difference between actual and expected daily DMI (more consistency) may be more effective in maintaining performance in the face of challenges or perturbations, whereas cows with greater variation in observed versus expected daily DMI (less consistency) are less feed efficient and may be less resilient.

Residual feed intake is viewed as an important trait in breeding programs that could be used to enhance genetic progress in feed efficiency. In particular, improving feed efficiency could improve both economic and environmental sustainability in the dairy cattle industry. However, data remain sparse, limiting the development of reliable genomic evaluations across lactation and parity for residual feed intake. Here, we estimated novel genetic parameters for genetic residual feed intake (gRFI) across the first, second, and third parity, using a random regression model. Research data on the measured feed intake, milk production, and body weight of 7,379 cows (271,080 records) from 6 countries in 2 continents were shared through the Horizon 2020 project GenTORE and Resilient Dairy Genome Project. The countries included Canada (1,053 cows with 47,130 weekly records), Denmark (1,045 cows with 72,760 weekly records), France (329 cows with 16,888 weekly records), Germany (938 cows with 32,614 weekly records), the Netherlands (2,051 cows with 57,830 weekly records), and United States (1,963 cows with 43,858 weekly records). Each trait had variance components estimated from first to third parity, using a random regression model across countries. Genetic residual feed intake was found to be heritable in all 3 parities, with first parity being predominant (range: 22-34%). Genetic residual feed intake was highly correlated across parities for mid- to late lactation; however, genetic correlation across parities was lower during early lactation, especially when comparing first and third parity. We estimated a genetic correlation of 0.77 ± 0.37 between North America and Europe for dry matter intake at first parity. Published literature on genetic correlations between high input countries/continents for dry matter intake support a high genetic correlation for dry matter intake. In conclusion, our results demonstrate the feasibility of estimating variance components for gRFI across parities, and the value of sharing data on scarce phenotypes across countries. These results can potentially be implemented in genetic evaluations for gRFI in dairy cattle.

Summary: Residual feed intake (RFI) is typically obtained as the difference between dry matter intake (DMI) observations and predictions from regression on known energy sinks, and effects of parity, days in milk, and cohort. Understanding the impact of parity on RFI estimation is important because variation across parities might affect the estimation of RFI phenotypes and, hence, the prediction of RFI breeding values. To investigate this, we calculated RFI considering the energy sinks nested or not nested (constant) within parity using weekly records of 5,813 lactating Holstein cows collected from 2007 to 2022 in the United States. In addition, genetic correlations between weekly RFI for parities 1, 2, and 3 were estimated using bivariate repeatability animal models as the genetic breeding values for RFI in each parity. Our results showed that nesting energy sinks within parity when computing RFI improves model goodness of fit, but the impact on the RFI phenotypes and the estimated breeding values appears to be minimal.