PCOS in bipolar women treated w/valproate or lithium
From MindLinc Wiki
Bottom Line: An open-label cross-sectional study of 38 women with bipolar disorder treated either with valproate or lithium revealed significantly more menstrual irregularities and a variety of less favorable endocrine and metabolic markers in women treated with valproate, despite not demonstrating any significant difference in weight.
Methods:
Design – Open-label cross-sectional (i.e., not blinded or randomized) ‘pilot’ study.
Setting – Referral outpatient clinic at University of Toronto , Sept 1999-Dec 2000.
Patient Population –
Inclusion criteria: females ages 18-50 with DSM-IV bipolar I or bipolar II disorder, “generally good health,” able to be “safely treated with anticonvulsants, antidepressants, or benzodiazepines.” If receiving valproate or lithium, needed to be on these for 2 months prior to blood sampling.
Exclusion criteria: currently on oral or injectable contraceptive meds, on antipsychotics (although 5 were on atypicals at time of blood sampling), any diagnosis of diabetes mellitus, dyslipidemia, any primary reproductive endocrine disorder (e.g., PCOS or nfertility), “assessed to be a suicide risk,” substance dependence in last 30 days, or “any known neurologic/medical disorders.”
Description of exposures: 18 patients taking valproate (average dose 1014 mg/day, SD 214; mean plasma level 71.4 mcg/mL; duration of treatment mean 28 mos [range 4-136 mos]) were compared with 20 patients taking lithium (average dose 976mg, SD 401; mean plasma level 0.92 meq/L; duration of treatment mean 24 mos [3-156 mos]).
Outcomes – Enrolled participants completed a menstrual history questionnaire and provided a fasting am blood sample during the follicular phase (day 4-7 of menstrual cycle, or at random for women with amenorrhea/oligomenorrhea [3/18 taking valproate, 0/20 taking lithium]). This sample was analyzed for FSH, LH, sex-hormone binding globulin (SHBG), androstenedione, dehydro-epiandrostenedione sulfate (DHEAS), testosterone, free testosterone, prolactin, TSH, total cholesterol, HDL, LDL, HbA1c, insulin-like growth factor 1 (IGF-1), insulin-like growth factor binding-protein 1 (IGFBP-1), fasting blood glucose, leptin. BMI was also measured (>25 = overweight, >30 obese). Menstrual history questionnaire was examined for 4 menstrual disorders: (1) amenorrhea (no menses for 6 mos), (2) oligomenorrhea (cycle length longer than 35 days for 6 mos), (3) prolonged cycles (?<35 to >35 days), and (4) irregular menstrual cycles (length 22-35 days, varying more than 4 days from cycle to cycle).
Analysis – Categorical data compared using chi-square test or Fisher exact test, continuous variables using 2-tailed unpaired Student t-tests.
Follow-up: none as this was a single-point-in-time cross-sectional study.
Validity Were the comparison groups clearly delineated and treated similarly outside of the putative exposure? The ‘control’ group was given the ‘alternative exposure’ of lithium so there was no true control or treatment-free group.
Did the groups have similar exposures to other possible determinants of outcome? Again, study was not randomized but patients were roughly similar (similar number of prior manic episodes, similar age). Weight was similar but standard deviation of weight was large suggesting skewed distribution with some markedly obese subjects (valproate-treated subjects mean weight 72.4kg [SD 56.44], mean BMI 31.14 [SD 53.13]; lithium-treated subjects mean weight 71.0kg [SD 62.58], mean BMI 29.21 [SD 60.89]). Of note, 3/18 valproate-treated subjects were amenorrheic or oligomenorrheic and so had blood drawn at random rather than in the ‘follicular phase’ of the cycle.
Were outcomes and exposures measured similarly in each comparison group? Yes, with the caveat of the amenorrheic and oligomenorrheic women in the valproate-treated group.
Was follow-up of sufficient duration? Time on drug varied considerably and there was no follow-up after initial blood draw.
Is there a reasonable temporal relationship between exposure and outcome? Again, time on drug varied considerably but exposure-outcome relationship is reasonable.
Is there a dose-response gradient? This was not analyzed or reported in this study.
Results What is the strength of the association between exposure and outcome?
See Table 3.
In summary:
- Of the ‘endocrine’ parameters measured, valproate-treated patients had significantly (p < 0.05) higher levels of free testosterone, androstenedione, estradiol, DHEAS; and significantly lower levels of SHBG, progesterone, LH, FSH, prolactin.
- Of the ‘metabolic’ parameters measured, valproate-treated patients had significantly (p < 0.05) higher levels of C-peptide and leptin, and significantly lower levels of IGF-1, IGFBP-1, and prolactin. Valproate-treated patients had nearly-significant trends toward higher total cholesterol, lower HDL, higher LDL, higher fasting glucose, higher triglycerides, and higher TSH. (Clinical relevance of this is, of course, a separate question.)
- There were no significant differences in absolute weight and BMI (see above).
- 9/18 valproate patients reported menstrual irregularities compared with 3/20 lithium-treated patients (p < 0.01), including one with amenorrhea and 2 with oligomenorrhea.
- There was a higher incidence of menstrual irregularities in patients who were overweight (65%) relative to patients not overweight (17%, p< 0.001).
- Overweight women with menstrual irregularities had a higher incidence of androstenedione elevation (50%) than normal-weight women with no menstrual irregularities.
How precise are the risk estimates for the exposure? The small sample size of this study makes any estimate of relative risk (i.e., for menstrual irregularities or for androgen elevation) preliminary at best.
What is the absolute level of risk from this exposure? See above for the preliminary findings from the study.
Comments
Strengths and Weaknesses of Study (internal and external validity)
Strengths – broad inclusion criteria, ‘real-world’ setting
Weaknesses – small sample size, cross-sectional design limits estimate of harm over time (i.e., no measurement of pretreatment baseline weights, no way to determine similarity of groups before exposure), open-label design, no ‘exposure-free group,’ permission of concomitant medications, lack of diagnostic interviews
Study in context of other available literature and/or current standard-of’care – The study is the most recent and most specific of its kind. Studies published previously had had mixed results about the presence of a valproate-PCOS link.
Should the study results cause avoidance of the exposure? These results should enter the risk-benefit calculus of the clinician when prescribing valproate over lithium.
Next steps for further study of this problem: The authors recommend a prospective multicenter trial which seems reasonable given the design limitations of this study.
Pathophysiologic considerations: The way in which valproate may affect the hypothalamic-pituitary axis or the body’s metabolic system is unclear and a matter of speculation (see page 32, right column, last paragraph). Valproate is known to cause weight gain but the weights and BMI of the subjects in this study did not differ significantly. The authors speculate that valproate may increase GH secretion which may increase insulin resistance. Other hypotheses involve the effect of valproate on progesterone, the LH/FSH ratio, and leptin.
Presented by Warren Kinghorn, MD, April 18, 2005
