Characterizing the Influence of Vitamin D Levels on IVF Outcomes

B. Rudick; S. Ingles; K. Chung; F. Stanczyk; R. Paulson; K. Bendikson


Hum Reprod. 2012;27(11):3321-3327. 

In This Article

Materials and Methods

This was a retrospective cohort study of 208 infertile women who underwent their first IVF cycle at University of Southern California (USC) Fertility Clinic from January 2006 to August of 2009. The study protocol was approved by the USC Institutional Review Board. Patients were excluded if they had previous IVF cycles at USC Fertility or if they underwent zygote intra-Fallopian tube transfer. Patient characteristics and cycle parameters were identified from patient medical records. Patient race was categorized according to self-identified race/ethnicity on their initial patient questionnaire.

All patients underwent IVF cycles using standardized regimens for pituitary down-regulation and controlled ovarian hyperstimulation. The particular protocol was chosen according to patient diagnosis and age. In general, good prognosis patients underwent a Leuprolide acetate down-regulation protocol (Lupron; TAP Pharmaceuticals, North Chicago, IL, USA; Porter et al., 1984), whereas those patients judged to have a poor prognosis underwent either an antagonist protocol with flexible ganirelix acetate start (Antagon; Organon, Inc., West Orange, NJ, USA) (Oliviennes et al., 1994), or a microdose flare protocol (Scott and Navot, 1994).

Controlled ovarian hyperstimulation was initiated with either recombinant FSH alone or in combination with human menopausal gonadotrophins (Menopur, Ferring, Inc., Suffern, NY, USA). Starting dose was selected on the basis of age, Day 3 FSH levels, and number of antral follicles, with adjustments made according to patient response. Serial monitoring of ovarian response was assessed by transvaginal ultrasound and serum estradiol (E2) assays. When two to three follicles reached or exceeded 17–18 mm, hCG (10 000 IU IM) was administered. Serum samples collected the day after hCG administration were stored at −20°C until assayed.

Transvaginal ultrasound guided oocyte retrieval was performed 34–35 h following hCG injection. Conventional insemination and/or ICSI was performed as indicated. Ultrasound-guided fresh embryo transfer was performed on Days 3 through 5 after egg retrieval. The number of embryos transferred depended upon embryo development and number of embryos available.

Luteal phase supplementation with vaginal micronized progesterone in capsules and oral estrace was started 3 days following egg retrieval. Clinical pregnancy was defined by the sonographic presence of a heartbeat at 7–8 weeks of gestation.

Vitamin D Status

Vitamin D status was measured by assessing circulating levels of 25(OH)D in frozen, never previously thawed serum samples using radioimmunoassay (RIA; DiaSorin, Stillwater, MN, USA; Hollis et al., 1993). Intra-and inter-assay coefficients of variation were 10.5 and 8.2%, respectively. Serum 25(OH)D was categorized according to clinically accepted ranges for vitamin D deficiency (<20 ng/ml), insufficiency (20–30 ng/ml) and replete (>30 ng/ml; Holick, 2007).

Statistical Analysis

Continuous data were summarized as the mean ± SD, or as median, 25th and 75th quartiles if highly skewed, and categorical data as percentage (%). Univariate analyses were carried out using the Kruskal–Wallis test for continuous outcomes and χ 2 tests for categorical outcomes.

Multivariable logistic regression was used to evaluate predictors of clinical pregnancy and of live birth. Vitamin D status (deficient, sufficient, replete) was included in the model as an ordinal variable. Race/ethnicity was coded using dummy variables, and effect modification was evaluated by including race-by-vitamin D cross-product terms in the model and conducting a likelihood ratio test. Since race was found to be an effect modifier, race-specific results were obtained by evaluating appropriate linear combinations of regression coefficients in the interaction model. Adjusted results are presented as predicted probability of the outcome (clinical pregnancy or live birth), adjusted to mean levels of covariates included in the model.

All models were adjusted for maternal age. Additional covariates were included in the final model if they confounded the relationship between vitamin D and treatment outcome, as evidenced by a change in race-specific regression coefficients of at least 15%. Variables evaluated as potential confounders included age, BMI, obesity, parity, diagnosis, previous treatment failure, stimulation protocol, season of transfer, number of embryos transferred and markers of embryo quality. Mild confounding effects (10–15% change in at least one race-specific regression coefficient) were observed for number of embryos, embryo quality (the average number of blastomeres among transferred embryos) and diagnosis of diminished ovarian reserve. Taken jointly, these three covariates resulted in substantial confounding (>25% change in one or more race-specific regression coefficients) and thus they were retained in the final model. Model fit was evaluated using the Hosmer and Lemeshow test (Hosmer and Lemeshow, 1980).

Power was calculated based on results of the Ozkan study, in which patients with the highest levels of vitamin D had a clinical pregnancy rate of 47% and those with the lowest vitamin D levels had clinical pregnancy rates of 20%. Thus, in order to detect a 27% difference in clinical pregnancy rates, with 80% power and alpha of 0.05, we required at least 102 patients. All P values are two sided and statistical significance was established as P < 0.05.

All analyses were conducted using Stata 11.0 (StataCorp, College Station TX).