Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Published December 11, 2014 Female Traits, Ovary and Follicle Characteristics, and the Conditional Probability of Normal Oocyte Development After Superovulation of Beef Cows' R. C. Greer*, R. B. Staigmillert, and J. J. Pamsh" *Department of Agricultural Economics and Economics, Montana State University, Bozeman 59717; +Fort Keogh Livestock and Range Research Laboratory, ARS, U.S. Department of Agriculture, Miles City, M T 59301; and *Department of Meat and Animal Science, University of Wisconsin, Madison 53706. regressors not been added to the model a simple, single equation would have represented oocyte development response; given an oocyte at aspiration only one variable, cumulus quantity, was found to condition the probability of normal development directly. However, the complete model included four additional equations: 1) the probability that an oocyte was recovered at aspiration was conditional on the plane of nutritional treatment and progesterone concentration in follicular fluid; 21 cumulus quantity was conditional on the presence on a corpus luteum, follicle size, and progesterone concentration; 3) progesterone concentration was dependent on plane of nutrition; and 4) corpus luteum w&s conditional on plane of nutrition. The estimated model provided some insight into the complexity of oocyte development response and the role nutrition may Play. ABSTRACT: The proportion of transferable beef embryos obtained after superovulation, follicle aspiration, and in vitro maturation and fertilization has been small. To seek possible explanations, cows on different planes of nutrition were treated with exogenous gonadotropin and oocytes were isolated from their ovaries. The record for each oocyte included characteristics of the follicle, ovary, and cow from which it was obtained and the response to in vitro maturation, fertilizstion, and development. The sample was used to obtain estimates of the relationships among the variables. The logistic function with the probability of normal development as the dependent variable was the basic equation of the statistical model. When an explanatory variable was itself a result of the biological system, an equation explaining variation therein was added to the model. Had equations representing endogenous Key Words: Follicles, Ovaries, Nutrition, Oocytes Viability J. Anim. Sci. 1992. 70:263-272 Introduction at reasonable cost. This limitation rests with the female gamete. Large numbers of female gametes can be harvested from ovaries obtained at slaughter plants or from selected females. But these oocytes then must undergo a progressive series of biological changes Imaturationl before they can be fertilized and develop into embryos. Techniques for maturing,fertilizing, and developing these oocytes into transferable embryos are available (see reviews by Staigmiller, 1988; LeibGried-Rutledge et al., 1989). However, the success rate of such procedures is still relatively low, A major limitation in the application of embryo transfer technology to production of domestic animals is obtaining a large number of embryos 'Published with the approval of the Director of the Montana Agric. E-. Sta. Journal series no. 5-2097. Received December 5, 1990. Accepted August 20. 1991. 263 GREER ET AL. 264 ranging from 20 to 40%. The low success rate raises the question of whether there are qualitative differences in the harvested oocytes that affect potential for complete development into transferable embryos. Criteria for selecting 00cytes based on differences in oocyte quality are limited and rather subjective (Leibfried and First, 1979). Similarly, females, ovaq, and(or1 follicle characteristics that may influence the number or quality of oocytes are not well established (Fukui and Sakuma, 1980; Katska and Smorag, 1984; Leibfried-Rutledge et al., 1985; Grimes and Ireland, 1986). The objective of this study was to investigate the relationships among factors related to oocyte quality Ke., the complete biological integrity of an oocyte as measured by its ability to undergo early embryonic development in vitro after in vitro maturation and fertilization). The fundamental postulate for this study was that oocyte development is a result of systematic relationships within a biological system and thus can be represented by a system of equations. Data were individual records of 401 follicles aspirated from ovaries of 33 FSH-treated beef cows that were included in an experiment to determine the influence of nutrition on ovarian function. This data sample was used to estimate parameters of a model describing oocyte development response to variation in characteristics of the female, the ovary, the follicle from which the oocyte was aspirated, cumulus cells, and the oocytes themselves. Materials and Methods Data Sample The data were observations on various characteristics of recovered and fertilized oocytes, the follicles from which the oocytes were aspirated, the ovaries on which the follicles were located, and the cows from which the ovaries had been surgically removed. Data were collected on 401 follicles from the ovaries of 33 cows. In part the experiment was designed to determine the effect of nutrition and body condition on oocyte development. Thus a 2 x 2 factorial arrangement was implemented with 17 cows initially placed on a low plane of nutrition and 16 cows placed on a high plane of nutrition. Cows were maintained a t these feed levels for 60 d to provide treatment groups with either a low 0 or high 0 body condition. For the next 20 d (one estrous cycle, 2Mention of trade names or companies does not constitute an implied warranty or endorsement by the authors, USDA, or Montana Agricultural Experiment Station and referred to as the final nutrition phase) onehalf of the cows in each group were switched to the other plane of nutrition. The result was four treatment groups: LL (n = 91, LH, HL, and HH (n = 8 each). Estrous cycles were synchronized for the final nutrition phase by giving two injections of prostaglandin 12 d apart; the second injection was given at the time of the change in nutrition (d 01. On d 12 the cows received an implant of Norgestomet (Cera Laboratories, Overland Park, KSP and on d 15 through 19 received injections of pFSH (total dose = 15 mgl twice daily. Prostaglandin was injected on d 18,the implant was removed on d 19, and the ovaries were surgically removed on the morning of d 20. After ovariectomy, characteristics of the ovaries were recorded and the follicles 2 7 mm in diameter were aspirated. Follicles 1 7 mm were selected because they represented the population of follicles from which ovulations would have occurred had ovulation been permitted. After aspiration, recovered 00cytes (n 290) were matured and fertilized in vitro. Methods for handling ovaries and in vitro maturation were those described by LeibfriedRutledge et al. (19871, except that oocytes were incubated individually in 10-pL droplets so that the fate of each could be recorded. Maturation was based on expansion of the cumulus cells or absence thereof. Fertilization was as reported by Parrish et al. (19861 using semen from a bull of known fertility. After fertilization, oocytes were incubated for 48 h as described by LeibfriedRutledge et al. (1987) to permit embryonic development. Development was assessed by staining the embryos and counting the number of nuclei. Concentration of steroid hormones in follicular fluid was determined by RIA previously validated in this laboratory (Staigmjller et al., 1979; Moseley et al., 19841. 3 : Statistical Model "he 401 records in the data sample included observations on the variables defined in Table 1. Normal oocyte development is represented by the binary variable ND (indicating normal development), which is equal to 1 when development response was 8, 9, or 10, or equal to 0 otherwise. The first statistical tests considered traits and characteristics of each cow and the response to the plane of nutrition. Cow was the unit of observation and linear models with ordinary least squares methods were applied to test for systematic variation when the variable of interest was a continuous variable. When the variable of interest was a categorical variable, the binomial formula was applied. The statistic& analysis then turned to estimating parameters of a model describing relations OOCYTE DEVELOPMENT RESPONSE among management variables, cow traits, ovary and follicle characteristics, and their influence on oocyte recovery and development. Normal oocyte development was represented by the restricted variable ND equal to 1 when normal development occurred, or 0 when development did not occur. The variable ND was the dependent variable in the statistical model, and it was interpreted as the probability of normal oocyte development. The explanatory variables were factors hypothesized to condition the probability distribution of normal development, and the null hypotheses were that coefficients on the explanatory variables were equal to zero. The basic equation of the statistical model was the logistic function as follows: where NDi = probability of normal development for the ith oocyte with the observed sample values being 1 if normal development and 0 otherwise; Xij = values of the J traits, characteristics, and other factors hypothesized to condition the probability of normal development for the ith oocyte; a, = the unknown parameters to be estimated; and 6 = the ith random disturbance term. The logistic function was chosen because its sigmoid shape corresponds to a cumulative distribution function and is not subject to the shortcomings of a linear model (for a complete discussion see Fomby et al., 1984, or Kmenta, 1971). Because the function is intrinsically nonlinear, maximum likelihood procedures were applied to obtain parameter estimates White et al., 19901. Selection of the best fit model was based on t statistics associated with the coefficients, robustness of the coefficient estimate in terms of sensitivity to equation specification, residual variance, log of the likelihood Table 1. Data sample variables defrntions Variables Definitions CA cow w e , Yr; Plane of nutrition, L = low, H = high; Cow body condition score at end of nutrition phase, sum of two technicians' scores on a 1 (thin] to 10 0 scale; Weight of the ovary on which the follicle was located, g; 1 when CL was located on right or left ovary, respectively, = 0 otherwise, both variables = 0 then no CL was present; CL weight, g; Number of follicles 7 mm in diameter on the same ovary as FN; Number of follicles 2 7 mm in diameter on the same ovary as F"; 1 when follicle 0 was on R or L ovary, respectively, = 0 otherwise; Surface diameter of follicle, p ~ z n ; 1 when follicle was largest follicle from either ovary from the cow, = 0 otherwise; Volume of follicular fluid, & Concentration of progesterone in follicular fluid; ng/mL; Concentration of &adiol in follicular fluid, ng/mL; Concentration of androstenedione in follicular fluid, ng/mL; oocyte quality, 1 - uniform cytoplasm, 2 - Pycnotic, 3 = Degenerative or clear cytoplasm, cumulus quantity, 1 = Nude to very sparse cumulus cells, 2 = Corona only, 3 = Corona + 1 egg diameter cumulus, 4 = Corona + 2 egg diameter cumduq cumulus q w t y , 1 - Good, 2 = Pycnotic; Development response, 1 = Germinal vesicle intact, 2 = Germinal vesicle breakdown, 3 = Clumped chromatin, 4 = Metaphase plate a: or ID 5 = Pronuclei, 6 1 Fragmented, 7 1 2 nuclei, 8 = 3 to 4 nuclei, e = 5 to 8 nuclei, 10 = > 8 nuclei TRT BCS OWT CUR or LP, binary CLWT FL.7 FG7 FS(R or U,binary Fz DOMF, binary Fv p4 E2 A4 OQ,qucllitative CQN, qu&itf&tive CQL, qualitative DR, qualitative 265 R or L is the last letter of a variable symbol it designates side location of that variable. GREER ET AL. 266 function, and the logic of the specification. Given the basic equation, consideration of the variables (Table 1) that might influence ND led to the conclusion that the appropriate model was more than a simple, single equation. Among the possible explanatory variables were female traits and ovary and follicle characteristics that were themselves responses of the biological system. As such they were considered random variables and the model was expanded to include an equation for each endogenous explanatory variable. Thus, the complete statistical model was a set of simultaneous equations that expressed the various interdependent responses of the biological system leading to oocyte development (e.g., female traits and ovary, follicle, and oocyte characteristics, and the s’ubsequenteffects on the probability of normal oocyte development). occurrence of physiological events that happen through time that condition a response or generate an effect. Within the sample, all cows of a given age also had the same reproductive history-3-yr-old cows were primiparous, and so on. Thus, in this data sample, age, parity, or any other term describing reproductive history were synonymous; age was selected for its simplicity. Because it is the occurrence of discrete events that may be important, the analyses considered categorical age variables in addition to the seemingly continuous variable, age. It was hypothesized that cow body condition score (BCS) would vary with treatment group. The initial parameter estimates obtained with ordinary least squares procedures were Results and Discussion with SE of the coefficients in parentheses, R2 = .715, and SE of the estimate = 1.684. It was concluded that the estimated coefficients on LL and LH were significantly different from zero but not significantly different from each other. The estimated coefficient on HL was not significantly different from zero. Thus,it was inferred that the determinant of BCS was the initial plane of nutrition, H vs L. Although the plane of nutrition had changed for the LH and HL cows, the number of days on the alternative feeding level (the 2 O d final nutrition phase) was such a short period that & response to change in nutrient quantity had not yet been manifested in changed BCS. Because no other variables, including cow age, or interactions among variables were found to explain variation in BCS, the model judged to best fit the data sample was as follows: Cow Traits Sample means, SD, and minimum and maximum values for the quantitative variables in the data sample are presented in Table 2. Although all are not explicitly included in the subsequent detailed discussion of the results, all were included in the statistical analyses, first, to determine whether these variables were sigruficantly affected by the plane of nutrition, and, second, to determine whether these variables should be included as explanatory variables in the ND model. One variable deserves discussion; cow age is reported in years. Yet, simple time is probably not the important dimension, rather it is the Table 2. Variable means, standard deviations, and minimum and maximum values Variable 33 Cow@ CA, v BCSb o m ,8 o n ,8 n 7 R , No. FL7L. No. FG7R, No. FG7L, No. 401 Follicles - Fz, m, Pd,n g / d Ez, A4, n g / d Min. 4.80 11.24 11.55 9.87 30.87 26.58 6.33 5.82 2.19 3.00 4.64 5.00 19.17 19.69 4.36 4.79 3.00 10.00 6.00 3.01 17.00 23.44 20.04 11.25 552.21 675.46 62.12 1.78 2.71 460.41 1,049.90 69.68 4.33 3.71 5.00 7.00 1.00 1.00 7.00 40.00 27.00 .oo .oo = 13.750 BCSi - 4.7500LLi - 5.125OLHi + .125OHLi (.8183) (-8421) (.8421) = 13.812 - 4.9890Li (.56961 Max. SD Me- BCSi 80.00 113.00 15.00 17.00 23.00 4,250.00 10,679.00 524.00 40.1 1 aVariable symbols BS defined in Table 1. ’sum of two technicians’scores on a 1 (thin)to IO vat scale). with SE of the coefficient in parentheses, R2 = .712, and SE of the estimate = 1.635. Presence of a corpus luteum (CL)(CL= 1) was related to initial plane of nutrition and cow age. However, effects attributable to each could not be separated with the data sample. All 16 cows (including six 3-yr-olds) on the high plane of nutrition had a CL,whereas 7 of the 17 cows on the low plane of nutrition did not have a CL. Six of the seven cows without CL were 3-yr-01ds1leaving only two 3-yr-old cows on the low plane of nutrition that did have a CL. There was no difference in presence of a CL attributable to changing the plane of nutrition during the experimental period. Because the variables conditioning OOCYTE DEVELOPMENT RESPONSE the probability of a CL, cow age 3, and L were themselves binary, the estimated logit model parameters exactly reflected the proportions of 3-yr-oldcows having or not having a CL.Thus, the following conditional probability statements regarding the presence of a CL were derived from the data sample: RCL P(-CL P(CL P(-CL RCL I I I I CA3 CA3 CAO CAO IH) & & & & L) L) = = L) = L) = = .25 .75 3889 .1111 1.0 where L denotes a cow on initial low plane of nutrition; H denotes a cow on initial high plane of nutrition; - denotes not, the characteristic was not present; CA3 denotes a 3-yr-old cow; and CAO denotes a cow > 3 yr old. Results with respect to two other variables, CL location and dominant follicle location, were in part consistent with past results, and in part raised a question to be researched and a hypothesis to be tested with much larger data samples. Although the number of cows in the present sample is small, the possibility of a nutritional effect on relative right-left ovary activity was raised by the CL and dominant follicle distributions (Table 3). Previous studies with dairy cows found that ovulations occurred slightly but significantly more often on the right ovary than on the left ovary (Kidder et al., 1952). If the point estimate of the probability of right location is taken to be .56 and because only the initial plane of nutrition affected CL location (the final nutrition phase was started just before ovulation), then among the L cows the proportion 4 of 10 CL right ovary, although < .5, certainly was not surprising. However, among the H cows the 12 of 16 CL right ovary seemed disproportionate, and so the probability was calculated as follows: ):( (.56)l2(.44I4 = .0649. This probability and the probability of observing 12 or more from a sample of 16 k 2 (F) = (.56)k(.44)'6 - = .0984 12 were neither small nor large. Although they were in the range that would often lead to rejecting the null hypothesis, they were not considered persuasive evidence for or against a plane of nutrition affect. Consideration of the dominant follicle location distributions only increased curiosity. Again, among the L cows, whether LL or LH, and even among the HL cows, the distributions (Table 267 Table 3. Number of cows, by treatment, showing location of the corpus luteum (CL) of the previous cycle, and the location of the dominant follicle Dominant follicle location CL location* Treatmentb NO LL LH 4 3 HL HH CL L ~ R Right Left Right 3 2 2 6 4 4 5 6 0 3 2 2 - 3 4 5 8 - *Location of the CL was dependent only on the initial plane of nutrition, because the change to the second plane (beginning of the f i i nutrition phase) was after ovulation. bL = low plane of nutrition; H = high plane of nutrition. 3) were not surprising. However, the probability of observing the HH distribution, 8 out of 8 dominant follicle right, is very small c) (.56)* = .0097 and further focuses on the need for additional research to understand the mechanism that apparently makes the right ovary more actke and the possible role of plane of nutrition. Variation in other cow characteristics, CL weight, ovary weight, the number of follicles < or > 7mm, the total number of follicles and the proportion of oocytes/embryos that develoFed normally, was not explained by or attributable to treatment group, cow age, or BCS. Of course, ovary weight was highly correlated, .74, with the number of follicles > 7 mm, and correlated, .42, with the number of follicles < 7mm (i.e., heavy ovaries had a large number of follicles). In summary, it was the initial plane of nutrition, not the change in plane of nutrition during the experimental period that had important effects on BCS and presence of a CL. The interesting results that raised questions in need of further research were found when the previously reported probability of right ovary location, .56, was applied to the CL and dominant follicle location distributions. Although the power of any statistical test was low because of the small number of cows in the study, the calculated probabilities of observing the H distributions were small enough to raise the conjecture that the mechanism making the right ovary more active than the left is in fact responsive to plane of nutrition. Oocyte Developmeni For the analysis of oocyte development the unit of observation was the individual follicle. First, within the sample of 401 observations, 115 oocytes GREER ET AL. 268 developed normally; thus, the success rate, or the unconditional probability of normal oocyte development from this sample of aspirated follicles, was 115/401 = .287. Second, there were four distinct possible outcomes: 1) an oocyte was not recovered at aspiration (111 follicles);2) the oocyte was lost in the laboratory during the experiment (66 oocytes); 3) the oocyte developed normally (ND);or 41 the oocyte did not develop normally (109 oocytes). The objective was to determine whether plane of nutrition, cow traits, and ovary and follicle characteristics entered the logistic function as explanatory variables to condition the probability of normal development. Given that some of the possible explanatory variables were themselves responses of the system, and thus random variables, a potential estimation problem had to be considered. The problem was the possible correlation between a stochastic regressor and the equation disturbance term. When a regressor is correlated with the equation disturbance term the parameter estimators are biased and inconsistent, a problem that even maximum likelihood procedures will not resolve m e n t a , 1971). Correlation between the equation disturbance and explanatory variables representing female traits and ovary and follicle characteristics seemed likely because all were generated within the same system or unit of observation. The problem was solved by expanding the model to include an equation for each endogenous explanatory variable. Adding equations not only solves the estimation problem, but the explicit relations among system responses and among system responses and basic determinants furthers understanding of the complexity of the biological system. The best fit statistical model representing the relations leading to normal oocyte development included five equations and the implicit relation between the plane of nutrition treatment and the presence of a CL. Two of the equations along with the implicit relation represented the biological system in terms of the conditional probability that an oocyte was recovered at aspiration. The three remaining equations then represented the biological system in terms of the conditional probability that a recovered oocyte developed normally. The equations and parameter estimates are presented in Table 4. Through the testing of many specifications only two variables, concentration of progesterone in follicular fluid (P3 and plane of nutrition, were found to condition the probability that an oocyte was recovered at aspiration (Equation 2, Table 41. Because P4 w&s itself a response of the biological system, an equation explaining variation therein was added to the model. The distribution of P4 concentration was skewed to the right, and therefore the values were transformed to LnW4). With Ln(P41 the dependent variable, measures of equation fit improved, and the best fit specification included only the binary variables LL and LH [Equation 1, Table 4). The distinction between LL and LH in terms of P4 was the only time in the analyses with the follicle as the unit of observa- Table 4. Parameter estimates for the system of equations describing the relations among plane of nutrition, cow traits, follicle characteristics, and normal oocyte development Equation number Dependent variableB Ekplanatory variables Constant CQN34 P4 L LL LH -.86690 -.47492 (.0886) CL n b ~ncwc SEE^ 401 observations 1 6.4881 -4 LO88 lIe 2 224 observations 3 ORA 2.1023 w 6.2856 4 4 CQNW 5 ND -.000538 -1.1020 (.00013) L2784) -2 14.86 h4P -.83488 -37608 (.0879) (.1008) -2.8258 -.57054 .762 (.23)r .001039 (.000479) 1.2124 C.2798) .577 (.2d 1.2700 .13566 (.3385) LO5591 -137.01 (64F -145.36 (65s is the Ln of progesterone concentration; ORA equal 1. when an oocyte was recovered at aspiration; CQN34 equal 1. when cumulus quantity was 3 or 4; ND is normal development. bFZ is diameter of the follicle, millimeters. cLn[L) is Ln of the likelihood function for the logistic function. dSEE is standard error of the estimate for the linear equation estimated by ordinary least squares. eNumbers in parentheses me absolute value of the standard error of the coefficient. ‘Adjusted R2. gpercentage of correct predictions within the sample. OOCYTE DEVELOPMENT RESPONSE 269 Table 5. Number of follicles with progesterone (Pb] concentration greater than 1,000 ng/mL by treatment Treatments Total no. of follicles LL 146 LH 06 HL 98 62 HH *L = low plane of Follicles with P4 > 1,000 ng/mL No. Proportion 2 10 18 14 .014 .lo4 .la8 226 nutrition; H = high plane of nutrition. tion that the changing plane of nutrition during the experimental period had a significant effect on the result. The effect of plane of nutrition is shown by the number of follicles with P4 > 1,000 ng/mL for each treatment group mable 5). The low plane of nutrition not only decreased P4, but it also decreased the probability of recovering a n oocyte. Equations 1 and 2 in Table 4 represent the biological system in terms of the likelihood that an oocyte was recovered. The relations as well as the cow trait responses to plane of nutrition are depicted in a tree diagram in Figure 1. The rectangular node with the L represents the low initial plane of nutrition, a management choice variable, and the two smaller rectangles represent the LL and LH subsets. The oval nodes represent the system responses, and the arrows represent the causal relations; the + Table 6. Calculated conditional probability that an oocyte was recovered at aspirationa RORA = 1IL and PJ 100 200 300 .7205 .7095 400 .7870 .6753 .6635 .a14 .6302 .6267 .6141 S490 .4822 .4160 2 4 18 .lo61 .6984 500 600 700 600 000 1,000 1,500 2,000 2,500 4,000 5,500 1/1 + e-(2-1029- l.lMo RORA I 1) is oocyte recovered at aspiration. *Eq-tion: p4); ORA HORA = 1IH and P3 b o g e a t e r one. .8858 .a803 .a745 .a685 .a623 .a558 .a491 .a421 .8348 .a273 .7856 .7371 .6620 .4898 .3006 - .000536 Figure 1. Tree diagram of the initial cause-effect relations among plane of nutrition (L), cow trait, ovary and follicle characteristics, and the probability an oocyte was recovered at aspiration (ORA); DOMF is dominant follicle. or - immediately above the arrow indicate the direction of the effect. For instance L, compared with H, has a two-part effect on the probability that an oocyte was recovered at aspiration; L directly decreases the probability, whereas LL and LH by different amounts decrease the P4 concentration, which, given the negative sign of the estimated coefficient and smaller P4, then increases the probability that an oocyte was recovered at aspiration. Although the indirect effect of nutrition (decreasing P41 offset the direct effect, the net effect of L was to decrease the probability that an oocyte was recovered. The effects are shown by the calculated example conditional probabilities that an oocyte was recovered presented in Table 6. For a given P4 concentration there was a smaller probability that an oocyte was recovered when the cow had been on the low plane of nutrition than when the cow had been on the high plane of nutrition. These results point to the need for further investigation and understanding of the role and mechanisms through which nutrition affects physiological responses. The finding that the cow trait responses to the plane of nutrition experiment with L acting to decrease the probability of a CL and BSC were separate from oocyte development is made clear in Figure 1. Although there w ~ l scorrelation in the observed values of BCS and P4,when the effect of the common basic determinant plane of nutrition was removed there was not a discernible causeeffect relation, hence no arrow in Figure 1; similarly no arrows appear for other variables for which no cause-effect relation was discernable. 270 GREER ET AL. One further outcome, the probability that the oocyte was lost during the maturation, fertilization, or development periods after treatment with FSH,must be considered before explicitly considering the phenomenon of primary interest, normal oocyte development. To concentrate on the three remaining outcomes, lost during the experiment, ND, and not ND, it was decided to truncate the data sample, to delete from the sample those observations in which an oocyte was not recovered at aspiration. Thus, the sample became 290 observations, because the oocyte was lost during the experiment for 66 of the observations. To test for systematic loss of oocytes during the experiment, several specifications regressing the restricted dependent variable, oocyte lost during the experiment, on other variables in the data sample were estimated. None of the estimated coefficients were significantly different from zero nor were statistics measuring the fit of the regressions significnatly different from zero. Thus, it was concluded that the loss of oocytes during the experiment was a random occurrence not related to variables in the data sample, and the data sample was truncated again to remove those observations in which the oocyte was lost. Within the truncated sample, the unconditional probability of ND was 115/224 = 5134. Normal development was affected by cumulus quantity, which in turn was influenced by the presence of a CL, follicle size (FZ), and P 4 concentration. The model best representing the truncated data sample was a system of three equations, Equations 3, 4, and 5 in Table 4. Two were logistic functions representing the conditional probability of occurrence for the restricted dependent variables ND and CQN34, and the third represented P4 concentration. After testing many specifications, only one variable in the data sample was found to directly condition ND, Equation 5. The variable was a categorization of oocyte cumulus quantity, which was recorded in the data sample as one of four values, Table 1. During estimation there was no discernable difference in response or effect between cumulus quantity values l and 2, nor was there a discernable difference between cumulus quantity values 3 and 4. There waa, however, a difference between the two subsets. Thus, a binary variable was constructed (Le., CQN34 = 1 when cumulus quantity was a 3 or 4, = 0 otherwise). Because CQN34 was itself a response of the biological system, another equation, Equation 4, Table 4, was added to the model. Three explanatory variables, P4, CL, and FZ, entered the equation and thus acted to condition the probability that cumulus quantity was either 3 or 4. It should be remembered that CL was a binary variable, and the presence of a CL was itself a conditional probability, conditioned by the plane of nutrition treatment and cow age. It should be noted that CL included cows from both the high and low planes of nutrition and of different ages and both CL right and CL left ovary. The possibility of influence by ipsilateralcontralateral c o n f i i a t i o n of the CL and the follicle from which the oocyte was aspirated was tested by constructing variables representing the disjoint subsets. Estimation results did not support side or ipsilateral-contralaterrtl configuration as significant explanatory variables because there was not improvement in the measures of fit. Thus, the conclusion was that "simply" the presence of a CL was important in conditioning the probability of CQN34. The low plane of nutrition and inseparable cow age effect, however, had an indirect effect as they conditioned the probability that a CL was present and P 4 concentration. The second variable acting to condition the probability of CQN34 was a continuous variable, FZ, measuring the diameter of the follicle lin millimeters)from which the oocyte was aspirated. It should be noted that FZ and follicular fluid Table 7. Calculated probabilities cumulus quantity was 3 or 4 given corpus luteum (CL) and selected progesterone (P4)and follicle size (PZ)valuesa Zonditional probability, N o CL 100 200 300 400 500 600 700 800 900 1,000 1,500 100 200 300 400 500 600 700 800 900 1,000 1,500 .2262 .2772 .3347 .2449 .2ea .3582 .2646 .3207 .3824 .2854 .3437 .4072 .3070 .3675 .4325 .3295 .3920 .4582 3529 .4170 .4841 .4425 3770 .5100 .4017 .4682 .I5360 .4269 .5617 .4942 .5560 .6216 .e830 Conditional probability, CL was present 3770 .4425 .5105 .5773 .e417 .4017 .4683 5360 A024 .e653 .4269 .4942 .5617 3270 ,6880 .4525 .5201 .5871 3510 .7088 .4783 5460 .0120 .e742 .7308 SO43 .5716 A364 .e966 .75M .7181 .7696 .5302 ,5968 .6601 .5560 .e210 .MI30 .7386 .7875 .5815 ,6457 .7050 .7582 .8044 .e065 .e691 .7262 .7767 .8202 .7216 .7727 .8168 .8540 .8847 .1452 .1586 .1730 .1884 .2048 .2222 2407 .2602 .2807 .3021 .4212 - .1823 .1983 .2153 .a34 2525 .2726 .2937 .3157 3385 .3622 .4884 - OOCYTE DEVELOPMENT RESPONSE volume were highly correlated, r = .87. Regression results with either as a regressor were essentially identical. Thus, it could not be con cluded that one explained more of the variation than the other. The specification including size was selected because its measurement is obtained first, thus facilitating use as a decision variable. The inference from the positive coefficient was that as follicle size increased the probability that cumulus quantity was 3 or 4 increased. None of the variables in the available data sample were found to significantly explain variation in follicle size, thus no equation was added to the model. Including P4 as an explanatory variable in the CQN34 equation led to interesting inferences. Recall from earlier results with the full sample (401 observations) that as P4 increased the probability that an oocyte was recovered decreased. Now with the truncated sample and the positive estimated coefficient, it was inferred that as P4 increased the probability that cumulus quantity was 3 or 4 increased. From the two results, the inference was that truncating the sample by deleting those observations from which an oocyte was not recovered removed observations with large P4 values. Comparison of the sample means supported the inference;with 401 observations the mean P4 was 675 and with 224 observations, the mean P4 was 482. Again, with P4 entering the system a second time as an explanatory variable, the Ln(P41equation was reestimated with the truncated sample. There was very little difference in the estimated coefficients on LL and LH (Equations 1 and 3, Table 4); no other variables entered as significant explanatory variables, and the intercept was 271 smaller, reflecting the effect of sample truncation. Thus,the consistency of the estimates, the robustness, increased confidence that the structure of the relation was properly modeled. The influence of the three explanatory variables in conditioning the probability CQN34 is shown by the example calculations presented in Table 7; the unconditional probability in the truncated sample was 116/224 = .5179. For a given follicle size and P4 concentration the probability that CQN was 3 or 4 was greater when a CL was present than when there was no CL. In appearance, the ND equation (Equation 5, Table 4) was quite simple and straightforward. Because CQN34 was binary, there were only two conditional probabilities to be calculated: P(ND P(ND = 1 = 1 I I CQN34 = 01 = .3611 and CQN34 = 11 = .6552 This apparent simplicity, however, masks the complexity of the system because 1) cumulus quantity was conditional on the presence of a CL, FZ, and P4;2) P4 was dependent on LL and LH; 3) CL was conditional on L and an apparent, but inseparable, influence from cow age; 41 the estimates were conditional on an oocyte being recovered; and 5) the probability an oocyte was recovered was conditional on L and P4. The complexity of the total system, the various responses to the plane of nutrition followed by the effects on subsequent responses, may be represented by incorporating the results of Equations 3, 4, and 5 into the initial tree diagram, Figure 2. The objective of this research was to identify variables in the data sample that ostensibly \ + - Figure 2, Tree diagram of the cause-effect relations conditioning the probability of normal oocyte development; ORA is the probability that an oocyte was recovered at aspiration, DOMF is dominant follicle, FZ is follicle size (millimeters) and CQN34 is the probability that cumulus quantity was 3 or 4. GREER ET AL. 272 affected oocyte development response, variables that conditioned the probability of normal oocyte development. The variables and combinations of variables in the data sample were specified as independent regressors in the logit model. Had the research been done with the view that the variables in the data sample, other than ND,were predetermined variables the conclusion would have been that cumulus quantity alone conditioned the probability of normal oocyte develop ment after treatment with FSH. The model would have been a very simple single equation, a single explanatory variable model with no recognized plane of nutrition effect or effects from attendant traits, characteristics, and hormone levels. Rather, this research recognized that when regressors in the model were themselves results of the biological system additional equations explaining variation in the endogenous regressors should be specified and parameter estimates obtained; responses of the biological system should be traced back to the basic determinants that initiated and motivated the system. The system of equations selected as best representing the data sample revealed a complex system, a model consisting of five equations with a sixth relation implicit. The implicit relation was the plane of nutrition (and perhaps cow age1 effect on the presence of a CL.The five equations represented P 4 concentration and the conditional probabilities of recovering an oocyte, cumulus quantity, and ND. From the results, the high plane of nutrition clearly acted to increase the probability of normal oocyte development after treatment with FSH. The effect, however, was an indirect effect working through other system responses, the probability an oocyte was recovered at aspiration, the presence of a CL,P4 concentration, and cumulus quantity. Implications Research often focuses on one trait or characteristic: normal oocyte development. But, aa clearly shown here, if other response variables are considered, predetermined errors result. The more grievous errors are the errors in logic and understanding of the biological system; the probability of normal oocyte development is not simply conditioned by cumulus quantity. The plane of nutrition also plays an important role. Working toward complete specification provides insight and generates an additional hypothesis and a question relative right-left ovary activity is responsive to plane of nutrition, but what are the basic determinants of the loss of oocytes or of follicle size? Identifying and obtaining measurements on basic determinants whose values are generated outside of but initiate or motivate the system would seem to be a high research priority. Literature Cited Fomby, T. B., R. C. Hill, and S. R Johnson. 1984. Advanced Econometric Methods. Springer-Verlag, New York. M u & Y., and Y. Sakuma. 1980. Maturation of bovine oocyte cultured in vitro: Relation to ovarian activity,follicular size and the presence or absence of cumulus cells. Biol. Reprod. 22:669. Grimes, R. W., and J. J. Ireland. 1986. Relationship of macre scopic appearance of bovine ovarian follicles, concentrstions of steroids in follicular fluid, and maturation of oocytes in viko. Biol. Reprod 35:725. Katska, L.. and Z. Smorag. 1984. Number and quality of oocytes in relation to age of cattle. Anim. Reprod. Sci. 7:451. Kidder, H. E.,G. R. Barrett, and L. E. Casida. 1952.A study of ovulation in six families of Holstein-Friesians. J. Dairy Sci. 35:436. Kmenta, J. 1971.Elements of Econometrics. McMiUan Publishing Co.,New York. Leibfried, L., and N. L. First. 1979. Characterization of bovine follicular oocytes and their ability to mature In vitro. J. a. Sci. 48:76. Leibfkied-Rutledge,M. L., E. S. Critser, W. H.Eyestone, D. L. Northey, and N. L. First. 1987. Development potential of bovine oocytes matured in vitro or in vivo. Biol. Reprod. 36: 376. Leibfied-Rutledge, M. L,E. S. Critser, and N. L. First. 1985. Fertilization potential of follicular oocytes classified by stage of cycle and size of follicle. Theriogenology 23:753. Leibfied-Rutledge, M. L., E. S. Critser, J. J. Parris4 and N. L. First. 1989. In vitro maturation and fertilization of bovine oocytes. Theriogenology 31:61. Maseley, W.M., T. G. Dunn, C.C. Kaltenback, R.E. Short, and R B. Staigmiller. 1984.Negative feedback control of luteinizing hormone secretion in prepubertal beef heifers at 60 and 200 days of age. J. Anim. Sci. 58145. Parrish, J. J., J. SuskePanish, M. L. Leibfried-Rutledge, E. S. Wtser, W. H. Eyestone, and N. L. First. 1986. Bovine in vitro fertilization w i t h frozen-thawed semen. TheriogenolOgy 25591. Staigmiller, R. B. 1988. In vitro methods for production of *le oocytes. J. Anim. S c i 68(SUppl. 2):44. Staigmiuer, R B., R. E. Short, R A. Bellows, and J. B. Carr. 1979.Effect of nutrition on response to exogenous FSH in beef cattle. J. Anim. Sci. 48:1182. White, K.J., S.D. Wow, D. Whistler, and S.A. H a m 1990. Shazam; Econometrics Computer Program, Usera Reference M a n d (Version 6.2). McGraw-Hill Book Co.,New York.