J. Dairy Sci. 86:2852–2863 © American Dairy Science Association, 2003. Predicting Cholesterol, Progesterone, and Days to Ovulation Using Postpartum Metabolic and Endocrine Measures1 C. C. Francisco,* L. J. Spicer,* and M. E. Payton† *Department of Animal Science †Department of Statistics Oklahoma State University Stillwater 74078-0425 ABSTRACT The objective of this study was to examine relationships among metabolic and endocrine factors that may influence ovarian activity during early lactation. Holstein cows (n = 19) were bled twice each week to determine plasma concentrations of insulin (INS), glucose, cholesterol, insulin-like growth factor-1 (IGF-I), and progesterone (P4). Feed intake and milk production were recorded daily while body weights and milk composition were recorded weekly. Relationships among plasma cholesterol and P4, and days to first and second postpartum ovulation were modeled with energy balance (EB), dry matter intake, milk yield and composition, plasma metabolites, and hormones using the backward elimination technique of multivariate regression analysis. Variables that contributed the most to predicting plasma cholesterol concentrations were dry matter intake × SNF using model 1 (production variables) and the square of glucose (i.e., glucose2) using model 3 (plasma hormones and metabolites). For plasma P4 concentrations, EB (model 2, production variables) and IGF-I (model 3, plasma hormones and metabolites) were the major predictors. The production variables EB and percentage of milk lactose were the greatest contributors to the models predicting days to first and second postpartum ovulations, respectively. Of the plasma hormones and metabolites evaluated, IGF-I2 was the most significant predictor of days to first postpartum ovulation, whereas glucose2 and INS were the significant predictors of days to second postpartum ovulation. Plasma IGF-I, glucose, and INS have been implicated in ovarian functions and their significant contributions to these models are consistent with possible important roles in postpartum return to ovarian competence. Received August 17, 2002. Accepted April 9, 2003. Corresponding author: L. J. Spicer; e-mail: igf1Leo@okstate.edu. 1 This work was approved for publication by the director, Oklahoma Agricultural Experiment Station and supported by Agtech, Inc., Waukesha, WI. (Key words: energy balance, metabolism, lactation, reproduction) Abbreviation key: EB = energy balance, FAT = milk fat percentage, P4 = progesterone. INTRODUCTION The return to ovarian function of early postpartum cows requires optimal conditions for several metabolic and endocrine factors. Factors related to early postpartum ovarian competence include proposed relationships between energy balance (EB) and postpartum interval to first ovulation (Butler and Smith, 1989; Spicer et al., 1990). Variables that are significantly correlated with each other include milk yield with days to first postpartum ovulation (Stevenson et al., 1983; Nebel and McGilliard, 1993), plasma cholesterol with plasma progesterone (P4), conception rate, and number of recoverable embryos (Kweon et al., 1986; Grummer and Carroll, 1988; Spicer et al., 1993b), plasma P4 concentration with days open and pregnancy rate (Folman et al., 1973; Villa-Godoy et al., 1988), plasma IGF-I with ovarian function (Lucy et al., 1992; Spicer et al., 1993b; Beam and Butler, 1997), and plasma glucose with pulsatility of LH (Butler and Smith, 1989; Canfield and Butler, 1990). Simple correlation coefficients (r) among some of these variables are significant but range from r = 0.06 to 0.64 (Villa-Godoy et al., 1988; Spicer et al., 1990; Lucy et al., 1992). Because ovarian activity is not solely influenced by any single condition or metabolite, it is necessary to integrate the relationships among various metabolic and endocrine factors. So far, few modeling studies have attempted to compare production variables with metabolic and endocrine variables of lactating dairy cows to predict postpartum ovarian competence. Stevenson and Britt (1980) utilized changes in concentrations of estradiol, LH, glucocorticoids, number of ovarian follicles, milk yield, energy intake, BW, and cross products of these variables between wk 1 and 2 postpartum to model first postpartum ovulation. The nine-variable model predicting days to first ovulation shows change 2852 MODEL COMPARISONS FOR LACTATING DAIRY COWS and cross product of estradiol concentrations, and number of follicles as the most important factors. The scope of Stevenson and Britt (1980) model is limited because variables were measured for only a period of 14 d postpartum, thus it may not predict cows with >14 d to first ovulation. Because most dairy breeders do not routinely breed cows at first ovulation, measurement of variables over a longer period to cover second and third postpartum ovulations would be preferable. Another model (Heuer et al., 2000) predicted average herd mean EB in early lactating cows using various dietary inputs, body condition score, milk components (fat, protein, lactose), and beta-hydroxy-butyrate but did not include any endocrine variables. In this later study, fat-to-protein ratio, milk fat and milk protein concentrations substantially explained the variability in EB. In the past few years, several metabolic and endocrine modulators have been identified that affect reproductive function including IGF-I, insulin, glucose, cholesterol and P4 (Spicer and Echternkamp, 1995; Butler, 2000). Thus, it is necessary to expand the previous models to include those reproductive modulators to delineate their contributions in predicting postpartum reproductive competence of early lactating dairy cows. Plasma cholesterol concentrations increase between calving and wk 6 postpartum in dairy cows (Carroll et al., 1990; Spicer et al., 1993b; Francisco et al., 2002), and are correlated with plasma P4 (Spicer et al., 1993b; Francisco et al., 2002), conception rates and number of recoverable embryos (Kweon et al., 1986; Grummer and Carroll, 1988). The association of increased plasma cholesterol concentration with increased luteal-phase P4 secretion in early lactating dairy cows (Spicer et al., 1993b) merits further investigation. Understanding what production and hormonal factors contribute to variation in plasma cholesterol may lead to insights that may help improve reproductive efficiency in dairy cattle. Because measurement of blood cholesterol has become economical and automated via quick tests (e.g., Johnson and Johnson Advanced Cholesterol kit, BioScience 2000, CholesTrack) for assessment of health and metabolic status of humans (Koda-Kimble and Young, 1988; Lloyd, 1991; Stehbens, 2001), perhaps this practice can be extended in dairy cows to assess metabolic and/or reproductive capacity postpartum. However, further research is needed to establish the relative importance of a given metabolite for such a use. Therefore, the goal of this study was to examine relationships among production variables, and metabolic and endocrine variables that may affect ovarian function, and ultimately identify key contributors (via modeling) that may regulate plasma cholesterol, P4, and days to first and second postpartum ovulation in early lactating dairy cows. 2853 MATERIALS AND METHODS Data collection was previously described in another study (Francisco et al., 2002). Briefly, data were collected from 19 pluriparous cows from 1 to 12 wk postpartum. Cows (n = 19) were individually fed a TMR consisting of sorghum silage (31.9%), alfalfa hay (21.2%), cottonseed (8.0%), and concentrate (38.9%). Nine cows were supplemented with propionibacteria (17 g/d). The TMR was sampled weekly and composited monthly for analyses. Feed intake and milk production were recorded daily, whereas BW were recorded weekly. Milk was collected twice daily (0300 and 1500 h) and analyzed weekly. Blood was collected twice each week via coccygeal venipuncture, plasma harvested and assayed for plasma concentrations of P4, glucose, insulin, cholesterol, and IGF-I. Weekly EB was calculated using net energy of intake, for maintenance, and that secreted in milk as described previously (Francisco et al., 2002). Multiple regression models were developed to predict plasma cholesterol and P4 concentrations and days to first and second postpartum ovulations as indicators of metabolic and reproductive status of the cows, respectively. The basic model used to obtain the relationships of the observed variables was: Log Yi = βo + β1X1 + β2X2 + ⴢⴢⴢ + βkXk + εi where Yi represents a dependent variable (either plasma cholesterol or plasma P4 concentrations), βo represents the intercept, β1 and β2 are the true regressions for observed X, k is the number of independent variables, and εi is the random error associated with the ith observation. All variables in the model were expressed as weekly means from wk 1 to 12 postpartum. Data from control (n = 10) and treated (n = 9; supplemented with Propionibacteria) cows were pooled because most of the variables were not found to be significantly different between treatments (see Francisco et al., 2002). Variables in the model included all the possible combinations of cross products and squares of all the variables except week postpartum. The first group of variables included the “production” variables EB, BW, DMI, FCM, and percentage of milk fat, milk protein, lactose, and SNF. The second set of variables incorporated the metabolic and hormonal variables, plasma INS, glucose, and IGFI concentrations. These two sets of variables were used separately in obtaining the “best” model for predicting plasma cholesterol (models 1 and 3) and P4 (models 2 and 4) concentrations during the first 12 wk postpartum. The weekly means ± SD (n = 228) of the various variables used in the models for predicting plasma cholesterol and P4 concentrations are summarized in Table Journal of Dairy Science Vol. 86, No. 9, 2003 2854 FRANCISCO ET AL. Table 1. Mean, SD of the variables used in predicting plasma cholesterol and progesterone (P4) in lactating dairy cows (models 1 to 4). Models 1 and 31 Model 52 Model 63 Variable Mean SD Mean SD Mean SD Energy balance, Mcal/d BW, kg DMI, kg/d FCM, kg/d Milk lactose (LACT),% Milk fat (FAT), % Milk protein (PROT), % SNF, % −0.6 641.8 23.7 34.8 5.0 3.1 3.2 8.9 6.2 56.9 4.0 6.7 0.2 0.8 0.6 0.6 −4.1 646.5 21.4 33.8 4.9 3.3 3.5 9.2 3.4 56 2.6 3.9 0.1 0.5 0.3 0.4 −3.4 649.6 22.2 34.9 5.0 3.3 3.4 9.2 3.5 54.5 1.7 5.0 0.1 0.8 0.3 0.4 Models 2 and 41 Cholesterol, mg/dl Progesterone (P4), ng/ml IGF-I, ng/ml Insulin (INS), ng/ml Glucose, mg/dl 177.6 1.1 106.0 0.4 60.1 53.5 1.3 60.0 0.2 5.0 Model 72 132.8 — 78.0 0.3 59.6 Model 83 22.6 — 39.0 0.1 6.0 148.0 — 98.0 0.4 60.8 22.1 — 41.0 0.1 1.8 1 Means based on n = 228. Means based on n = 15. 3 Means based on n = 11. 2 1. Plasma cholesterol and P4 values were log transformed because it increased the coefficient of determination (R2). To determine the impact of week postpartum in predicting cholesterol (models 1 and 3) and P4 (models 2 and 4) concentrations, it was included as one of the variables in one group of models and not included in another group of models. The same datasets were used to predict days to first (models 5 and 7) and second (models 6 and 8) postpartum ovulations as above without cross products and squares of the variables. In the models using metabolic and hormonal variables (models 7 and 8), cross products and plasma P4 were not included in the models because P4 concentrations were used to determine if and when the cows had ovulated postpartum. Based on P4 concentrations, there were 15 and 11 cows that showed first and second postpartum ovulations, respectively. Weekly means of the variables from 1 to 5 wk (n = 75) and 1 to 7 wk (n = 77) postpartum were averaged and used to generate the models for days to first and second (Table 1) postpartum ovulation, respectively, because the intervals to first and second postpartum ovulations averaged 5 and 9 wk, respectively. The backward elimination technique of the multiple regression procedure (SAS, 1988) was used to obtain the models used in the study. Initially, the backward regression technique includes all the variables in the first model, calculating an R2 for each variable. This is followed by removal of the least significant variable with the lowest R2 contribution to the model, thereby producing the second model, and so forth. Only the variables that remained significant at the P < 0.05 or 0.01 during each permutation were included in the final Journal of Dairy Science Vol. 86, No. 9, 2003 model unless stated otherwise. Models were selected based on high R2, small residual variations between predicted and observed values, and a low number of variables included in the model. RESULTS Model Selection Models for plasma cholesterol and P4. Models 1 and 2 were generated using production variables (Table 1). The model that “best” predicted plasma cholesterol concentrations (model 1) contained eight variables. All of the variables in model 1 were significant at P < 0.01 and had a total R2 = 0.685. The relative contribution of each variable, as indicated by partial R2 (Table 2), showed that the interaction of DMI × SNF is the most important component in the model contributing 63% (R2 = 0.433). This is followed by FCM × SNF (R2 = 0.192), which explained 28% of the composite model R2 (Table 2). There were other significant variables (i.e., FCM × LACT, BW × LACT, FCM2, FAT2, FCM × FAT, and BW) in the model but they each contributed less than 5% of the total R2 (data not shown). The “best” fit model for plasma P4 concentrations (model 2) included three production variables all significant at P < 0.05 (Table 2), and had a lower R2 = 0.291 compared with model 1 (plasma cholesterol). Two more variables (i.e., DMI × FAT and FCM × FAT) were significant but contributed less than 5% to the total R2 of model 2. The difference in R2 between models 1 and 2 indicates that the variables that predicted plasma cholesterol explained more of the variation than the model that predicted plasma P4 (Table 2). 2855 MODEL COMPARISONS FOR LACTATING DAIRY COWS Table 2. Variables and proportion of R2 in the model predicting plasma cholesterol and progesterone using weekly measurements (n = 228) of production variables, hormones and metabolites in lactating dairy cows. Model Variables1 Partial R2 % Contribution to the model DMI × SNF FCM × SNF 0.43** 0.19** 63.2 28.1 EB EB × FAT FCM × LACT 0.18** 0.05** 0.04** 61.1 17.1 14.2 GLU2 IGF-I IGF-I × GLU IGF-I × INS INS 0.20** 0.07** 0.04** 0.03** 0.02** 55.0 18.9 11.8 8.8 5.5 IGF-I GLU2 0.25** 0.07** 74.2 21.8 1. Plasma cholesterol model Total R2 0.68 2. Plasma progesterone model 0.29 3. Plasma CHOL model 0.37 4. Plasma progesterone model 0.33 1 Variables included are only those that contributed > 5% of the total R2. EB = energy balance, GLU = glucose, INS = insulin. **P < 0.01. Models 3 and 4 included only three variables (plasma insulin, glucose, and IGF-I) and their cross products. The models for plasma cholesterol (model 3) and plasma P4 (model 4) have similar R2 values (Table 2) when based on this combination of mean weekly measurements of hormones and metabolites (Table 1). Model 3 indicates that glucose2 contributes over half (55.0%) to the total R2 indicating that it is the most important variable in predicting plasma cholesterol concentrations (Table 2). This is followed by IGF-I (18.9%) and IGF-I × glucose (11.8%). The rest of the significant variables (i.e., IGF-I × INS and INS) each contributed less than 10% to total R2 (Table 2). The R2 of the model predicting plasma P4 (model 4) was 0.3319 of which plasma IGF-I and glucose2 contributed 74.2 and 21.8%, respectively (Table 2). The only other significant variable was IGF-I2 (4.0%). Week postpartum, when included in models 1 to 4, had the highest contribution to the total R2 of the models, ranging from 63 to 88% (Table 3). In model 1 Table 3. Variables, R2 and percent contribution of week postpartum when included in plasma cholesterol and progesterone models (models 1 to 4) using weekly measurements of hormonal, metabolic, and production variables in lactating dairy cows. Production variables1 Model 1, plasma cholesterol model Week EB × FAT FCM × LACT Model 2, plasma progesterone model Week BW × SNF EB × LACT Hormones and metabolic variables Model 3, plasma cholesterol model Week GLU2 Model 4, plasma progesterone model Week INS × GLU IGF-I Partial R2 % Contribution Total R2 0.52** 0.09 0.08 71.6 12.5 10.3 0.732 0.25** 0.02 0.03 67.9 8.1 8.0 0.369 0.53** 0.05 88.2 7.6 0.595 0.26** 0.12** 0.03** 62.9 28.0 8.1 0.410 1 Variables included are only those that contributed 5% of the total R2. EB = energy balance, GLU = glucose, INS = insulin. **P < 0.01. Journal of Dairy Science Vol. 86, No. 9, 2003 2856 FRANCISCO ET AL. Table 4. Variables, regression coefficients and proportion of R2 in the model predicting days to first (n = 15) and second (n = 11) postpartum ovulation using the average of weekly measurements of production variables (wk 1 to 5, model 5; wk 1 to 7, model 6) and hormones and metabolites (wk 1 to 5, model 7; wk 1 to 7, model 8) in lactating dairy cows. Variables1 Model 5, EB DMI FCM Model 6, LACT FCM DMI Model 7, IGF-I Model 8, GLU2 INS Regression coefficients Partial R2 SE Partial R2 (%) first postpartum ovulation model Total R2 0.350 −5.85 5.80 4.99 3.11 2.39 2.44 0.209* 0.093* 0.048* 59.78 26.57 13.67 −51.87 −1.17 3.42 13.02 0.46 1.21 0.434** 0.224* 0.101* 57.24 29.49 13.28 second postpartum ovulation model 0.759 first postpartum ovulation model 0.282 −0.14 0.063 68.64 −0.025 40.94 0.015 0.282** 100.0 second postpartum ovulation model 0.318 0.23* 0.09* 72.7 27.3 **P < 0.01 for Models 5 and 6; P < 0.05 for Models 7 and 8. *P < 0.05 for Models 5 and 6; P < 0.10 for Models 7 and 8. 1 EB = energy balance, GLU = glucose, INS = insulin. (plasma cholesterol), week postpartum contributed 71.6% to model R2, whereas EB × FAT and FCM × FAT contributed a total of 22.8% (Table 3). Percent contribution of variables that both appeared in model 1 without and with week postpartum vary such as DMI × SNF (63.2 vs. 2.1%), FCM × SNF (28.1 vs. 2.1%), and FCM × LACT (2.7 vs. 10.3%). The other significant variables that contributed less than 5% of the model R2 were IGF-I and IGF-I × glucose. In model 2 (plasma P4), week postpartum contributed 67.9% of total R2 (Table 3). The variable EB appeared in model 2 with and without week postpartum but at different contributions to the total model R2 (3.4 vs. 61.1%). Other variables (i.e., PROT2, EB2, EB × FCM, LACT, EB × DMI, PROT, and SNF) contributed significantly but less than 5% of the total model R2. Week postpartum, when included in model 3 (plasma cholesterol) contributed 88.2% of the total R2 (Table 3). The common variables of model 3 with (Table 3) and without (Table 2) week postpartum were plasma IGF-I (3.6 vs. 18.9%), glucose2 (7.6 vs. 55.0%) and IGF-I × glucose (0.2 vs. 11.8%). When included in model 4 (plasma P4), week postpartum contributed 62.9% of the total R2 (Table 3). Variables common in model 4 (plasma cholesterol) with (Table 3) and without (Table 2) week postpartum was plasma IGF-I concentration contributing 8.1 and 74.2%, respectively. Plasma INS concentration was a significant variable in model 4, but it contributed less than 5% of total R2. Models for days to first and second postpartum ovulation. Models 5 and 6 predicted days to first and second postpartum ovulation, respectively, using the average weekly means (wk 1 to 5, model 5 and wk 1 to 7, model 6) of production variables (Table 1). Model 5 Journal of Dairy Science Vol. 86, No. 9, 2003 showed that EB had the highest partial R2 (0.21) while for model 6, percentage of milk LACT generated a partial R2 of 0.43 (Table 4). Model 6 had a higher total R2 compared with model 5 (0.76 vs. 0.35) having the same number of variables (Table 4). Both models had FCM and DMI as common variables. models 7 and 8 were derived using the average of weekly means (wk 1 to 5, model 7 and wk 1 to 7, model 8) of plasma hormones and metabolites (Table 1) to predict days to first and second postpartum ovulation, respectively (Table 4). The square of plasma IGF-I concentrations was the only significant predictor for days to first postpartum ovulation (model 7), while days to second postpartum ovulation (model 8) showed that plasma INS and the square of plasma glucose concentrations contributed 75 and 25%, respectively, to the total R2 (Table 4). All other variables were not significant. If days to first and second postpartum ovulation were modeled on a weekly basis, rather than using means of wk 1 to 5 or wk 1 to 7, the production variables that influenced first postpartum ovulation were EB, FCM, and percentage of milk protein (Table 5). Plasma hormones and metabolites that influenced first postpartum ovulation were plasma cholesterol concentration that successively appeared from 2 to 5 wk postpartum and plasma IGF-I concentration that emerged 3 times (Table 6). The production variables that influenced second postpartum ovulation were FCM, DMI and percentage of milk lactose (Table 5). In the model of plasma hormones and metabolites, plasma glucose concentration emerged in 5 of 7 wk as a significant variable that influenced second postpartum ovulation (Table 6). In 2857 MODEL COMPARISONS FOR LACTATING DAIRY COWS Table 5. Variables and R2 of the models predicting days to first and second postpartum ovulation in lactating dairy cows using measurements of production variables on a weekly postpartum basis.1 Week First postpartum ovulation 1 2 3 4 Second postpartum ovulation 1 2 3 4 5 6 7 Production and milk component variables2 R2 EB (0.15)3 FAT (0.06) FCM (0.05) FCM (0.15) 0.161 FCM (0.03) LACT (0.05) LACT (0.01) EB (0.01) FCM (0.09) FCM (0.06) FCM (0.01) PROT (0.10) PROT (0.10) EB (0.12) 0.336 LACT (0.07) DMI (0.02) EB (0.02) SNF (0.12) 0.559 0.251 SNF (0.04) 0.712 0.391 0.571 PROT (0.02) DMI (0.04) LACT (0.03) DMI (0.01) SNF (0.01) 0.869 0.439 BW (0.12) PROT (0.01) FAT (0.13) 0.652 0.964 1 Multiple regression using backward elimination technique of variables that contributed to R2 at P < 0.15. EB = energy balance; FAT = percentage of milk fat; PROT = percentage of milk protein; FCM = fatcorrected milk; LACT = percentage of milk lactose. 3 Level of significance (P-value) for the variables that contribute at P ≤ 0.15 to the weekly model. 2 several cases (Table 6, wk 4 to 7), addition of the squared values (e.g., INS2) provided a significantly greater fit to the model than with only the linear term (e.g., INS), indicating a nonlinear relationship. Model Prediction Models for plasma cholesterol and P4. Observed plasma cholesterol concentrations and predicted values using models 1 (production variables) and 3 (plasma hormones and metabolites) are plotted by week postpartum in Figure 1 and reflects only those variables that contributed greater than 5% of the total R2. Predicted plasma cholesterol values derived using model 1 were more similar to observed values than were cholesterol values predicted using model 3. The weekly postpartum residuals between observed plasma cholesterol concentration and predicted values ranged from −1.0 to 0.89 (model 1, production variables) and −0.97 to 0.86 (model 3, plasma hormones and metabolites). With a similar range of residuals but higher R2 for model 1 (R2 = 0.68) vs. model 3 (R2 = 0.37), model 1 was better than model 3 in predicting plasma cholesterol concentrations. Generally, the residuals between observed and predicted values using the regression coefficients of models 1 (production variables) and 3 (plasma hormones and metabolites) for predicting plasma cholesterol concentrations, congregate within -0.5 to 0.5 except for some outliers (Figure 2). Observed plasma P4 concentrations and predicted values derived using models 2 (production variables) and 4 (plasma hormones and metabolites) are plotted in Figure 3. Observed values and predicted values of plasma P4 concentrations using models 2 and 4 were similar starting from 4 wk postpartum. Residuals between observed plasma P4 concentrations and predicted values ranged from −3.5 to 3.5 (model 2, production variables) and −3.4 to 3.8 (model 4, plasma hormones and metabolites). Even though the residuals are slightly less dispersed in model 2 than in model 4, models 2 and 4 are comparable in predicting plasma P4 concentration as indicated by similar R2 (Figure 3 and Table 1). Generally, the residuals for models 2 (production variables) and 4 (plasma hormones and metabolites) (Figure 4) used for predicting plasma P4 concentrations fall between −2 to 2, a much wider range than residuals obtained from models 1 and 3 used for predicting plasma cholesterol (Figure 2). The lower and upper diagonal borders of Figure 4 are likely due to the nature of the plasma P4 data (i.e., 89 values were < 0.10 ng/ml). Models for days to first and second ovulation. The residuals between observed and predicted days to first postpartum ovulation ranged from −21 to 32 d for Journal of Dairy Science Vol. 86, No. 9, 2003 2858 FRANCISCO ET AL. Table 6. Variables and R2 of the models predicting days to first and second postpartum ovulation in lactating dairy cows using measurements of plasma hormones and metabolites on a weekly postpartum basis.1 Week First postpartum ovulation 1 2 3 4 5 Second postpartum ovulation 1 2 3 4 5 6 7 Plasma hormones and metabolic variables2 INS (0.12)3 IGF-I (0.03) IGF-I (0.04) CHOL (0.001) CHOL (0.07) INS (0.03) IGF (0.11) GLU (0.02) CHOL (0.02) GLU (0.03) GLU (0.05) CHOL (0.03) R2 GLU2 (0.03) GLU2 (0.03) GLU (0.03) CHOL2 (0.01) CHOL (0.14) INS (0.001) IGF-I (0.02) 0.379 0.642 0.412 GLU (0.02) CHOL2 (0.07) INS2 (0.03) IGF2 (0.08) 2 CHOL (0.001) 2 INS (0.001) 2 GLU (0.02) 0.911 0.563 GLU2 (0.11) 0.517 0.149 0.471 INS (0.02) GLU2 (0.03) GLU2 (0.05) GLU (0.01) CHOL2 (0.02) INS2 (0.03) 0.711 0.486 0.391 CHOL (0.03) 2 2 INS (0.01) 2 GLU (0.01) 0.822 a Multiple regression using backward elimination technique of variables that contributed to R2 at P < 0.15. INS = plasma insulin; GLU = plasma glucose; IGF-I = plasma insulin-like growth factor-1; CHOL = plasma cholesterol. 3 Level of significance (P-value) for the variables that contributed at P ≤ 0.15 to the weekly model. 2 model 5 (production variables) and −22 to 36 d for model 7 (plasma hormones and metabolites, Figure 5). Because the dispersion of the residuals for the two models is equivalent and the R2 of model 5 is greater than that of model 7, model 5 (production variables) appears slightly better predictor of days to first postpartum ovulation than model 7 (plasma hormones and metabolites). The difference between predicted and observed values for days to second postpartum ovulation ranged from −6 to 11 d for model 6 (production variables) and −15 to 9 d for model 8 (plasma hormones and metabolites) (Figure 5). Because the dispersion of the residuals for model 6 is less than model 8 and the R2 of model 6 Figure 1. Predicted weekly means (n = 228) of log cholesterol with week postpartum using models 1 (production variables) and 3 (plasma hormones and metabolites) compared with actual observed values. SD = model 1 (0.26), model 3 (0.13), and Observed (0.18). Figure 2. Plot of residuals of observed-predicted log cholesterol vs. predicted log cholesterol values using regression coefficients of model 1 (production variables) and model 3 (plasma hormones and metabolites). Journal of Dairy Science Vol. 86, No. 9, 2003 MODEL COMPARISONS FOR LACTATING DAIRY COWS 2859 Figure 3. Predicted weekly means (n = 228) of log progesterone with week postpartum using models 2 (production variables) and 4 (plasma hormones and metabolites) compared with actual observed values. SD = Model 2 (0.79), model 4 (0.69), and Observed (1.12). is over threefold that of model 8, days to second postpartum ovulation is best predicted by model 6 (production variables). Figure 5. Plot of residuals of observed-predicted days to first postpartum ovulation versus predicted days to first (n = 15) and second (n = 11) postpartum ovulation values using regression coefficients of models 5 and 6 (production variables) and models 7 and 8 (plasma hormone and metabolites). DISCUSSION Prediction of plasma cholesterol levels was better using model 1, based on production and milk composition variables, than using model 3, based on endocrine and metabolic variables. This implies that DMI and milk production and its components are more important factors driving changes in plasma cholesterol than are hormones and metabolites. In model 1, the product of SNF with DMI and FCM accounted for a significant percentage of the variation in plasma cholesterol but those variables appeared to be less important when week postpartum was included in the model. This is not surprising because during early postpartum the increase in plasma cholesterol concentrations coincides with the increase in DMI and FCM with week postpar- Figure 4. Plot of residuals of observed-predicted log progesterone vs. predicted log progesterone values using model coefficients of model 2 (production variables) and model 4 (plasma hormones and metabolites). tum (Carroll et al., 1990; Spicer et al., 1993b; Francisco et al., 2002). In addition, the percentage of SNF in milk decreases with increasing milk production during the early postpartum period (McDonald et al., 1995; Francisco et al., 2002). These interactions in model 1 (i.e., DMI × SNF, FCM × SNF) indicate that the response of plasma cholesterol may be dependent on DMI and FCM but this dependence changes depending on the level of SNF. Previously, plasma cholesterol concentrations were consistently important in predicting nutritional status of lactating and nonlactating dairy cows (Kronfeld et al., 1982), and the fact that cross products that included DMI and EB were significant contributors to variation in plasma cholesterol levels in model 1 of the present study support the results of Kronfeld et al. (1982). In model 3, the variables that significantly contributed to variation in plasma cholesterol concentration were plasma glucose, INS, and IGF-I concentrations. This is consistent with reports that plasma INS and IGF-I concentrations increase concomitantly with plasma cholesterol during early lactation (Carroll et al., 1990; Spicer et al., 1993b; Francisco et al., 2002). Model 3 was even more precise in predicting plasma cholesterol when the week postpartum was included along with plasma hormone and metabolite variables. Previous studies have documented the importance of cholesterol as a precursor of ovarian steroidogenesis (Kronfeld et al., 1980; Grummer and Carroll, 1988; Spicer et al., 1993a) and thus understanding the variables that contribute to changes in plasma cholesterol concentrations Journal of Dairy Science Vol. 86, No. 9, 2003 2860 FRANCISCO ET AL. may help understand factors that may contribute to reproductive success in postpartum dairy cows. Models 2 and 4 appeared to be equivalent at predicting plasma P4 concentrations using production variables and plasma hormones and metabolites, respectively, because of their similar R2 and range of residuals. In model 2, weekly EB accounted for about 18% of the variation in plasma P4 and was the most important predictor of the model. Energy balance had been shown in several studies to modulate plasma P4 concentrations during early postpartum (Villa-Godoy et al., 1988; Spicer et al., 1990, 1993b). Concentrations of P4 have been associated positively with fertility (Folman et al., 1973) and pregnancy rates (Sklan et al., 1991) and negatively with EB (Villa-Godoy et al., 1988; Spicer et al., 1990, 1993b) and days open (Sklan et al., 1991). In addition, a negative EB reduces the weight of corpus luteum (Apgar et al., 1975) and decreases steroidogenic activity of luteal tissue (Villa-Godoy et al., 1990). Thus, the weekly energy status of the cows measured as the difference between net energy intake and energy expended for lactation plus maintenance in the present study, served as a good predictor of return to normal ovarian activity as measured by P4 concentrations. In model 4, IGF-I accounted for 25% of the variation in plasma P4 and served as the best predictor of plasma P4 levels in early postpartum cows. Plasma IGF-I concentrations increase with week postpartum as does plasma P4 concentrations (Abribat et al., 1990; Spicer et al., 1990) and are positively correlated with energy status and P4 production in early postpartum cows (Spicer et al., 1990, 1993b; Lucy et al., 1992). Also, IGFI stimulates bovine luteal cell P4 production (Sauerwein et al., 1992; Liebermann et al., 1996). Thus, it is not surprising that IGF-I is a good predictor of plasma P4 concentrations in early postpartum cows. Values predicted by both models 2 and 4 closely matched observed plasma P4 concentrations after wk 4 of lactation suggesting that both models were more precise from 4 to 12 wk postpartum period. Because the dispersion of the residuals for model 2 were slightly less than for model 4, but R2 of model 4 was slightly greater than that of model 2, both models appear similar in their ability to predict plasma P4 concentrations. Thus, plasma P4 concentration could be predicted as well by EB as by plasma IGF-I concentrations, and further imply that IGF-I may in part mediate the effect of EB on luteal function as previously suggested (Spicer et al., 1990). The cholesterol models were more accurate models (i.e., lower residuals and greater R2) than the P4 models, probably because plasma cholesterol concentrations steadily increase with week postpartum and then plateau after 8 wk postpartum compared with plasma P4 concentrations that start low initially and then increase Journal of Dairy Science Vol. 86, No. 9, 2003 and decrease in a cyclic pattern after the first postpartum ovulation. Adding week to the model marginally increased the R2 in the P4 models (models 2 and 4) compared with a large effect of week in predicting cholesterol concentrations (models 1 and 3) associated with the more consistent increase in plasma cholesterol as the postpartum period progressed. Week postpartum when included in the model was the best single predictor of plasma P4 and cholesterol in models 1 to 4. The main predictors of the models without week postpartum became secondary predictors when week postpartum was in the model. For instance, in model 3 (plasma cholesterol), the variable glucose2 largely predicted the model (55% of total R2) without week postpartum but accounted for only 7.6% of total R2 when week postpartum was included in the model. Also, most variables that correlate to the number of days or weeks postpartum were normally overshadowed by week postpartum. For example, IGF-I accounted for 74.2% of the variability of the P4 model (model 3) without week postpartum but only 8.1% when week postpartum was included in the model. However, running the models without week helped to determine which production, endocrine, and metabolic variables contributed more to each model independent of postpartum intervals. Predicting days to first postpartum ovulation using model 5 (production variables) generated a higher R2 (0.35) compared to model 7 (0.28) (plasma hormones and metabolites) although model 7 contained only 1 variable. The production variable in model 5 that “best” predicted days to first postpartum ovulation was EB, but DMI and FCM were also significant contributors of the model. Models predicting first postpartum ovulation on a weekly basis also gave EB as one of the most influential variables (significant in 2 of 4 wk modeled) but FCM was also significant in 2 of 4 wk modeled. The small sample size (i.e., n = 11) may have contributed to the variation in the weekly models. This model supports the findings of others that milk production and DMI and hence EB during the early postpartum period is a critical factor to the return to reproductive cyclicity (Butler and Smith, 1989; Spicer et al., 1993b; Staples et al., 1998). In model 7, the squared term of plasma IGF-I accounted for the most variation in days to first postpartum ovulation. Because plasma IGF-I concentrations steadily increase with week postpartum, plasma IGFI may reach a certain concentration that may serve as a signal that first postpartum ovulation may occur. Generally, plasma IGF-I peaks around 7 to 8 wk postpartum long after which the first postpartum ovulation normally takes place (Abribat et al., 1990; Spicer et al., 1990; Francisco et al., 2002). Furthermore, IGF-I is MODEL COMPARISONS FOR LACTATING DAIRY COWS positively correlated with P4 (as mentioned earlier), which is necessary to maintain pregnancy after ovulation. On a weekly basis, models for first postpartum ovulation appeared to be influenced most by plasma cholesterol (significant in 4 of 5 wk modeled) and IGFI (significant in 3 of 5 wk modeled) concentrations the latter of which coincides with the results of model 7. Plasma cholesterol is a known precursor of P4 (Savion et al., 1982; Grummer and Carroll, 1988), and necessary to maintain pregnancy after ovulation. Also, IGF-I is a potent stimulator of HDL and LDL metabolism and P4 biosynthesis by ovarian cells (Veldhuis et al., 1987; Veldhuis and Gwynne, 1989). Models 6 (production variables) and 8 (plasma hormones and metabolites) used to predict days to second postpartum ovulation generated R2 values of 0.76 and 0.32, respectively. Although the greater R2 in model 6 can be partially explained by the greater number of variables generated compared with model 8 (3 vs. 2 variables), model 6 had smaller dispersion of residuals than model 8. In model 6, percentage of milk lactose, FCM, and DMI influenced the interval of return to second postpartum ovulation. This finding is similar to the results generated when first postpartum ovulation was modeled on a weekly postpartum basis: percentage of milk lactose, FCM, and DMI were significant factors in 3 of 7 wk, 4 of 7 wk, and 3 of 7 wk modeled, respectively. This result is consistent with the fact that return to reproductive competence is influenced by the EB of the cow, and that EB is partially explained by the DMI, amount of milk produced and variability of milk components according to the models of Heuer et al. (2000). Model 8 (plasma hormones and metabolites), showed that concentrations of plasma glucose and INS had important associations with return to second postpartum ovulation. Similar results were obtained when second postpartum ovulation was examined weekly in the postpartum period: glucose or glucose2 were significant effects in 4 of 7 wk modeled, and INS or INS2 were significant effects in 3 of 7 wk modeled. Because glucose is the main source of energy for ovarian function (Rabiee et al., 1997), and influences bovine thecal cell steroidogenesis in vitro (Stewart et al., 1995) it may play a major role in achievement of postpartum ovulation. Rabiee and Lean (2000) found that there was a positive and highly significant cross correlation (r = 0.5) between the uptake of glucose and cholesterol and suggested that glucose may promote cholesterol uptake into ovarian cells or vice versa. Also, low plasma INS concentrations were associated with decreased LH pulsatility (Butler and Smith, 1989; Canfield and Butler, 1990). Moreover, INS administered in vivo increases estradiol concentrations in follicular fluid of superovulated cattle (Simpson et al., 1994). In cultured granulosa cells, INS 2861 augments FSH-induced P4 production (Langhout et al., 1991) and LDL metabolism (Veldhuis and Gwynne, 1989), increases proliferation (Langhout et al., 1991; Spicer et al., 1993a) and stimulates aromatase activity (Spicer and Francisco, 1997; Spicer and Chamberlain, 1998; Spicer et al., 2002). Thus, lower INS that is normally observed in early lactating cows (Koprowski and Tucker, 1973; Smith et al., 1976; Francisco et al., 2002) likely affects days to return to postpartum ovulation in part by the lack of sufficient direct action on the ovary. During early postpartum where lactation overrides other physiological processes (Bauman and Curie, 1980; Swanson, 1989), glucose, a precursor of milk lactose, may become limiting to other physiological processes such as postpartum ovulation (Staples and Thatcher, 1990). Interestingly, milk lactose was the most important production variable whereas plasma INS and glucose were the most important hormone/metabolite variables modeling days to second postpartum ovulation. The energetic efficiency of converting glucose to milk lactose is 0.98 (McDonald et al., 1995). If early postpartum DMI is low and if a cow produces more milk, then glucose supply to the ovary may not be sufficient for a cow to return to postpartum ovulation. Results of model 5 are consistent with this latter statement because EB, DMI, and FCM all accounted for significant variation in days to first postpartum ovulation. Contrary to this, Stevenson and Britt (1980) showed that energy intake and milk yield during the first 2 wk of lactation provided small contributions to the model predicting days to first ovulation. However, energy intakes and EB are low and milk production is increasing during the first 2 wk postpartum. In contrast, in the current study, models covered a longer postpartum period (5 to 12 wk) and a greater range of physiological conditions than in the model of Stevenson and Britt (1980). In the present study, some of the same variables (i.e., IGF-I, cholesterol, glucose, and insulin) emerged in the models predicting days to first and second postpartum ovulation when values for each week were modeled, and their significant contribution to these models are consistent with a possible important role in postpartum return to ovarian competence. Further research will be needed to ascertain whether some of those variables can be used in an applied way to monitor reproductive competence in early lactating dairy cows. CONCLUSIONS Relationships of plasma cholesterol and P4 with EB, DMI, milk yield and composition, metabolites and hormones, and week postpartum were modeled using multivariate regression analysis. Models using EB, DMI, Journal of Dairy Science Vol. 86, No. 9, 2003 2862 FRANCISCO ET AL. FCM, BW, and milk composition as independent variables consistently predicted both plasma cholesterol and P4 quite well as shown by the minimal residuals between predicted and observed values. Models using metabolites and hormones as independent variables predicted plasma cholesterol more accurately than they predicted plasma P4 based on the magnitude of residuals. The primary variables involved with predicting plasma cholesterol were DMI × SNF (model 1, production variables) and the square of glucose (model 3, plasma hormones and metabolites) contributing most to the total model R2 (63 and 55%, respectively). For P4, EB (model 2, production variables) and IGF-I (model 4, plasma hormones and metabolites) were the primary contributors to the total model, providing 61 and 74% of the total model R2 values, respectively. Week postpartum when included in the model was the major predictor in models 1 to 4, although cross products of EB were significant contributors to models 1 and 2 and IGF-I and (or) cross products of glucose were significant contributors of models 3 and 4. Days to first and second postpartum ovulation were “best” predicted by production variables EB and milk lactose percentage (models 5 and 6). Plasma metabolic and hormonal predictors for days to first and second postpartum ovulations were IGF-I2 (model 7, plasma hormones and metabolites) and INS (model 8, plasma hormones and metabolites), respectively, and support previous studies that have implicated both IGF-I and insulin as important regulators of ovarian function. Data analysis from other herds or studies should be used to further validate these results. REFRENCES Abribat, T., H. Lapierre, P. Dubreuil, G. Pelletier, P. Gaudreau, P. Brazeau, and D. Petitclerc. 1990. Insulin-like growth factor-I concentration in Holstein female cattle: variations with age, stage of lactation and growth hormone-releasing factor administration. Domest. Anim. Endocrinol. 7:93–102. Apgar, J., D. Aspros, J. E. Hixon, R. R. Saatman, and W. Hansel. 1975. Effect of restricted feed intake on the sensitivity of the bovine corpus luteum to LH in vitro. J. Anim. Sci. 41:1120–1123. Bauman, D. E., and W. B. Curie. 1980. Partitioning of nutrients during pregnancy and lactation: a review of mechanisms involving homeostasis and homeorhesis. J. Dairy Sci. 63:1514–1529. Beam, S. W., and W. R. Butler. 1997. Energy balance and ovarian follicle development prior to first ovulation postpartum in dairy cows receiving three levels of dietary fat. Biol. Reprod. 56:133– 142. Butler, W. R. 2000. Nutritional interactions with reproductive performance in dairy cattle. Anim. Reprod. Sci. 60–61:449–457. Butler, W. R., and R. D. Smith. 1989. Interrelationships between energy balance and postpartum reproductive function in cattle. J. Dairy Sci. 72:767–783. Canfield, R. W., and W. R. Butler. 1990. Energy balance and pulsatile LH secretion in early postpartum dairy cattle. Domest. Anim. Endocrinol. 7:323–330. Carroll, D. J., M. J. Jerred, R. R. Grummer, D. K. Combs, R. A. Pierson, and E. R. Hauser. 1990. Effect of fat supplementation and Journal of Dairy Science Vol. 86, No. 9, 2003 immature alfalfa to concentrate ratio on plasma progesterone, energy balance, and reproductive traits of cattle. J. Dairy Sci. 73:2855–2863. Folman, Y., M. Rosenberg, Z. Herz, and M. Davidson. 1973. The relationship between plasma progesterone concentration and conception in post-partum dairy cows maintained on two levels of nutrition. J. Reprod. Fertil. 34:267–278. Francisco, C. C., C. S. Chamberlain, D. N. Waldner, R. P. Wettemann, and L. J. Spicer. 2002. Propionibacteria fed to Dairy cows: Effects on energy balance, plasma metabolites and hormones, and reproduction. J. Dairy Sci. 85:1738–1751. Grummer, R. R., and D. J. Carroll. 1988. A review of lipoprotein cholesterol metabolism: importance to ovarian function. J. Anim. Sci. 66:3160–3173. Heuer, C., W. M. Van Straalen, Y. H. Schukken, A. Dirkzwager, and J. P. T. M. Noordhuizen. 2000. Prediction of energy balance in a high yielding dairy herd in early lactation: model development and precision. Livest. Prod. Sci. 65:91–105. Koda-Kimble, M. A., and L.Y. Young. 1988. Pages 29–31 in Applied Therapeutics: The Clinical Use of Drugs. 4th ed. Applied Therapeutics, Vancouver, WA. Koprowski, J. A., and H. A. Tucker. 1973. Bovine serum growth hormone, corticoids and insulin during lactation. Endocrinology 93:645–651. Kronfeld, D. S., S. Donoghue, J. M. Naylor, K. Johnson, and C. A. Bradley. 1980. Metabolic effects of feeding protected tallow to dairy cows. J. Dairy Sci. 63:545–552. Kronfeld D. S., S. Donoghue, R. L. Copp, F. M. Stearn, and R. H. Engle. 1982. Nutritional status of dairy cows indicated by analysis of blood. J. Dairy Sci. 65:1925–1933. Kweon, O. K., H. Ono, K. Osasa, M. Onda, K. Oboshi, H. H. Uchisugi, S. Kurosawa, H. Yamashina, and H. Kanagawa. 1986. Factors affecting serum cholesterol level of lactating Holstein cows. Jpn. J. Vet. Sci. 48:481–486. Langhout, D. J., L. J. Spicer, and R. D. Geisert. 1991. Development of a culture system for bovine granulosa cells: effects of growth hormone, estradiol, and gonadotropins on cell proliferation, steroidogenesis and protein synthesis. J. Anim. Sci. 69:3321–3324. Liebermann, J., D. Schams, and A. Miyamoto. 1996. Effects of local growth factors on the secretory function of bovine corpus luteum during the oestrous cycle and pregnancy in vitro. Reprod. Fertil. Dev. 8:1003–1011. Lloyd, J. K. 1991. Cholesterol: should we screen all children or change the diet of all children. Acta Paediatr. Scand. (Suppl.) 373:66–72. Lucy, M. C., J. Beck, C.R. Staples, H. H. Head, R. L. De La Sota, and W. W. Thatcher. 1992. Follicular dynamics, plasma metabolites, hormones and insulin-like growth factor I (IGF-I) in lactating cows with positive or negative energy balance during the preovulatory period. Reprod. Nutr. Dev. 32:331–341. McDonald, E, R. A. Edwards, J. F. D. Greenhalgh, and C. A. Morgan. 1995. Page 217 in Animal Nutrition. 5th ed. Longman Singapore Pub. (Pte) Ltd., Singapore. Nebel, R. L., and M. L. McGilliard. 1993. Interactions of high milk yield and reproductive performance in dairy cows. J. Dairy Sci. 76:3257–3268. Rabiee, A. R., and I. J. Lean. 2000. Uptake of glucose and cholesterol by the ovary of sheep and cattle and the influence of arterial LH concentrations. Anim. Reprod. Sci. 64:199–209. Rabiee, A. R., I. J. Lean, J. M. Gooden, B. G. Miller, and R. J. Scaramuzzi. 1997. An evaluation of trans-ovarian uptake of metabolites using arterio-venous difference methods in dairy cattle. Anim. Reprod. Sci. 48:9–25. SAS/STAT User’s Guide. 6th ed. 1988. SAS Inst. Inc., Cary, NC. Sauerwein, H., A. Miyamoto, J. Gunther, H. H. Meyer, and D. Schams. 1992. Binding and action of insulin-like growth factors and insulin in bovine luteal tissue during the oestrous cycle. J. Reprod. Fertil. 96:103–115. Savion, N., R. Laherty, D. Cohen, G. M. Lui, and D. Gospodarwicz. 1982. Role of lipoproteins and 3-hydroxy-3-methyl-glutaryl coenzyme A reductase in progesterone production by cultured bovine granulosa cells. Endocrinology 110:13–22. MODEL COMPARISONS FOR LACTATING DAIRY COWS Simpson, R. B., C. C. Chase Jr., L. J. Spicer, R. K. Vernon, A. C. Hammond, and D. O. Rae. 1994. Effect of exogenous insulin on plasma and follicular insulin-like growth factor I, insulin-like growth factor binding protein activity, follicular oestradiol and progesterone, and follicular growth in superovulated Angus and Brahman cows. J. Reprod. Fertil. 102:483–492. Sklan, D., U. Moallem, and Y. Folman. 1991. Effect of feeding calcium soaps of fatty acids on production and reproductive responses in high producing lactating cows. J. Dairy Sci. 74:510–517. Smith, R. D., W. Hansel, and C. E. Coppock. 1976. Plasma growth hormone and insulin during early lactation in cows fed silage based diets. J. Dairy Sci. 59:248–261. Spicer, L. J., E. Alpizar, and S. E. Echternkamp. 1993a. Effects of insulin, insulin-like growth factor-I, and gonadotropins on bovine granulosa cell proliferation, progesterone production, estradiol production, and (or) insulin-like growth factor-I production in vitro. J. Anim. Sci. 71:1232–1241. Spicer, L. J., and C. S. Chamberlain. 1998. Influence of cortisol on insulin- and insulin-like growth factor-I (IGF-I)-induced steroid production and on IGF-I receptors in cultured bovine granulosa cells and thecal cells. Endocrine 9:153–161. Spicer, L. J., C. S. Chamberlain, and S. M. Maciel. 2002. Influence of gonadotropins on insulin- and insulin-like growth factor-I (IGFI)-induced steroid production by bovine granulosa cells. Domest. Anim. Endocrinol. 22:237–254. Spicer, L. J., and S. E. Echternkamp. 1995. The ovarian insulin and insulin-like growth factor system with an emphasis on domestic animals. Domest. Anim. Endocrinol. 12:223–245. Spicer, L. J., and C. C. Francisco. 1997. The adipose obese gene product, leptin: evidence of a direct inhibitory role in ovarian function. Endocrinology 138:3374-3379. Spicer, L. J., W. B. Tucker, and G. D. Adams. 1990. Insulin-like growth factor-I in dairy cows: Relationships among energy balance, body condition score, ovarian activity, and estrous behavior. J. Dairy Sci. 73:929–937. Spicer, L. J., R. K. Vernon, W. B. Tucker, R. P. Wettemann, J. F. Hogue, and G. D. Adams. 1993b. Effects of inert fat on energy balance, plasma concentrations of hormones and reproduction in dairy cows. J. Dairy Sci. 76:2664–2673. 2863 Staples, C. R., J. M. Burke, and W. W. Thatcher. 1998. Influence of supplemental fats on reproductive tissues and performance of lactating cows. J. Dairy Sci. 81:856–871. Staples, C. R., and W. W. Thatcher. 1990. Relationship between ovarian activity and energy status during the early postpartum period of high producing dairy cows. J. Dairy Sci. 73:938–947. Stehbens, W. E. 2001. Coronary heart disease, hypercholesterolemia, and artherosclerosis. I. False premises. Exp. Mol. Pathol. 70:103–119. Stevenson, J. S., and J. H. Britt. 1980. Models for prediction of days to first ovulation based on changes in endocrine and non-endocrine traits during the first two weeks postpartum in Holstein cows. J. Anim. Sci. 50:103–112. Stevenson, J. S., M. K. Schmidt, and E. P. Call. 1983. Factors affecting reproductive performance of dairy cows first inseminated after five weeks postpartum. J. Dairy Sci. 66:1148–1154. Stewart, R. E., L. J. Spicer, T. D. Hamilton, and B. E. Keefer. 1995. Effects of insulin-like growth factor-I and insulin on proliferation, and basal and luteinizing hormone-induced steroidogenesis of bovine thecal cells: involvement of glucose and receptors for insulin-like growth factor I. J. Anim. Sci. 73:3719–3731. Swanson, L. V. 1989. Interactions of nutrition and reproduction. J. Dairy Sci. 72:805–814. Veldhuis, J. D., and J. T. Gwynne. 1989. Insulin-like growth factor type I (somatomedin-C) stimulates high-density lipoprotein (HDL)-supported progesterone biosynthesis by swine granulosa cells in vitro. Endocrinology 124:3069–3076. Veldhuis, J. D., J. E. Nestler, and J. F. Strauss, III. 1987. The insulinlike growth factor, somatomedin-C, modulates low density lipoprotein metabolism by swine granulosa cells. Endocrinology 121:340–346. Villa-Godoy, A., T. L. Hughes, R. S. Emery, L. T. Chapin, and R. L. Fogwell. 1988. Association between energy balance and luteal function in lactating dairy cows. J. Dairy Sci. 71:1063–1072. Villa-Godoy, A., T. L. Hughes, R. S. Emery, W. J. Enright, A. D. Ealy, S. A. Zinn, and R. L. Fogwell. 1990. Energy balance and body condition influence luteal function in Holstein heifers. Domest. Anim. Endocrinol. 7:135–148. Journal of Dairy Science Vol. 86, No. 9, 2003