Figure 2. Initial assessment – significant correlations of hormones with MQ symptom domains
However, given the large number (120) of correlations performed between hormone and symptom variables, a Bonferroni correction was applied with an experiment-wise significance level of p<.00045 required for interpretation of statistical significance. None of the hormones remained significantly correlated with a symptom domain after the Bonferroni correction.
Significant correlations between hormone levels and follow-up MQ symptom domains appear in Table 2. DHEA correlated significantly with the immunological domain (r=.65, p<.04). FSH correlated significantly with the neuropsychiatric (r=.56 p<.05) and musculoskeletal (r=.67, p<.013) domains. LH correlated significantly with the musculoskeletal domain (r=.59, p<.034). Free testosterone correlated significantly with the neuropsychiatric (r=.64, p<.019), musculoskeletal (r=.68, p<.01) and dermatologic (r=.57, p<.04) domains, and total testosterone correlated significantly with the immunological domain (r=.63, p<.028). Finally, TSH correlated negatively (r=-.62, p<.031) with the endocrinological domain. However, none of these correlations was significant after applying the Bonferroni correction for multiple comparisons, requiring p<.00045.
Table 2. Follow-up assessment
| Hormone | Symptom Domain Correlation(s) | Value, r | Value, p | Hormone Change |
| TSH | Endocrinological | -0.62 | <.031 | Decrease |
| LH | Musculoskeletal | 0.59 | <.34 | Increase |
| FSH | Neuropsychiatric | 0.56 | <.05 | Increase |
| Musculoskeletal | 0.67 | <.013 | Increase | |
| DHEA | Immunological | 0.64 | <.04 | Increase |
| Free Testosterone | Neuropsychiatric | 0.64 | <.019 | Increase |
| Musculoskeletal | 0.68 | <.01 | Increase | |
| Dermatologic | 0.57 | <.04 | Increase | |
| Total Testosterone | Immunological | 0.63 | <.028 | Increase |
Figure 3. Results of the follow-up assessment
Factor Analysis
A factor analysis of the MQ yielded 2 factors with eigenvalues > 1.0. The first factor had high loadings from the pulmonary, GI, cardiovascular, and immunological domains. The second factor had high loadings from all 3 neurologically related domains as well as the musculoskeletal and gynecological domains. The first factor correlated significantly with pregnenolone (r=.37, p<.019) and the second factor correlated significantly with TSH (r=.33, p<.034) [Fig 4].
Figure 4. A factor analysis of the MQ yielded 2 factors with eigenvalues > 1.0
A factor analysis of the hormone levels yielded 4 factors (Fig. 5). Factor 1 had high loadings from DHEA, progesterone, free testosterone, and total testosterone. Factor 2 had high loadings from FSH and LH. Factor 3 had high loadings from estrone and progesterone. The fourth factor had high loadings from pregnenolone and TSH. Factor 1 did not correlate with any of the symptom domains or with the symptom factors. Factor 2 correlated significantly with the endocrinological domain (r=-.47, p<.012). Factor 3 correlated significantly with the pulmonary (r=-.43, p<.035), musculoskeletal (r=-.43, p<.024), and genitourinary (r=-.49, p<.009) domains.
Figure 5. A factor analysis of the hormone levels yielded 4 factors
ANOVA and Regression Analysis
The one-way ANOVAs on the MQ symptom domains and the 2 MQ factors between the origin of menopause (age-related, surgery, disease) were not significant. The one-way ANOVAs on the MQ symptom domains and the 2 MQ factors between the younger (40-59) and older (>=60) subjects were not significant. A stepwise regression analysis to predict symptom domain factors from the 10 hormone levels was not significant. Similarly, a stepwise regression analysis to predict symptom domain factors from the 4 hormone level factors was not significant.
DISCUSSION
Our data show significant correlations with menopause symptomatology in a cohort of females ranging in age from 43 to 87-years, prior to taking estrogen. These hormonal correlations include DHEA associated with genitourinary; testosterone with all symptoms; FSH with pulmonary; TSH with both pulmonary and gynecological; estrone with musculoskeletal; and pregnenolone with both immunological and genitourinary. While we attempted to obtain data on a follow-up questionnaire only 13 patients responded following a mailed survey. Although this provided us with a small number for subsequent statistical analysis we did find significant correlations with FSH and immunological and dermatological symptoms as well as pregnenolone with immunological symptoms in this population. When we utilized a Bonferroni correction for multiple comparisons, none of these correlations were significant except for TSH, which correlated at r = .52 (p<.0001) with the pulmonary domain. We are confident that the significant correlations have clinical relevance because the Bonferroni