Changes in Collagen Type I C-Telopeptide and Procollagen Type I N-Terminal Propeptide During the Menopause Transition

Abstract

Context

Collagen type I C-telopeptide (CTX) and procollagen type I N-terminal propeptide (PINP) are reference bone resorption and formation markers, respectively.

Objective

To characterize CTX and PINP trajectories across the menopause transition (MT).

Methods

This 18-year longitudinal analysis of a community-based cohort from the Study of Women’s Health Across the Nation included 541 women (126 Black, 90 Chinese, 87 Japanese, 238 White) who transitioned from pre- to postmenopause. Multivariable mixed effects regression fit piecewise linear models of CTX or PINP relative to years from final menstrual period (FMP); covariates were race/ethnicity, body mass index (BMI), and age at FMP. In the referent participant (White, 52.46 years at FMP, BMI 27.12 kg/m²), CTX and PINP were stable until 3 years pre-FMP (premenopause). During the MT (3 years before to 3 years after the FMP), CTX and PINP increased 10.3% (P < .0001) and 7.5% (P < .0001) per year, respectively; MT-related gains totaled 61.9% for CTX and 45.2% for PINP. Starting 3 years post-FMP (postmenopause), CTX and PINP decreased 3.1% (P < .0001) and 2.9% (P < .0001) per year, respectively. Compared with the White participants, during the MT, Chinese participants had larger gains in CTX (P = .01), and Japanese women experienced greater increases in CTX (P < .0001) and PINP (P = .02). In postmenopause, CTX (P = .01) and PINP (P = .01) rose more in Japanese relative to White women.

Conclusion

CTX and PINP are stable in premenopause, increase during the MT, and decrease in postmenopause. During the MT and postmenopause, bone turnover change rates vary by race/ethnicity.

During the menopause transition (MT), alterations in bone turnover, characterized by negative remodeling balance and faster remodeling, lead to accelerations in bone mineral density (BMD) decline and loss of microarchitectural integrity (1-5). Bone remodeling is the mechanism by which old bone is resorbed by osteoclasts and replaced in situ with new bone by osteoblasts. This process occurs in basic multicellular units (BMUs) distributed across the skeleton. In women who have attained peak bone mass, areal BMD is stable until approximately 1 year before the final menstrual period (FMP) (4), as the amount of bone removed within each BMU roughly equals the amount replaced (ie, there is remodeling balance) (6). During the MT, bone resorption exceeds formation, leading to negative remodeling balance and bone loss (1, 6, 7). BMU activation frequency (the rate at which new skeletal sites of remodeling emerge) also increases during the MT. This results in more rapid bone turnover and damage to bone microarchitecture, including lower trabecular number, weaker trabeculae, larger trabecular spacing, conversion of trabecular plates to rods, diminished trabecular connectivity, erosion of the endosteal cortex, and cortical thinning (8). Faster increase in bone turnover during the MT is associated with subsequent fracture, independent of BMD or rate of BMD loss (9).

Biochemical bone turnover markers (BTMs) are compounds measured from blood or urine samples that reflect the total-body activity of osteoclasts (bone resorption markers) and osteoblasts (bone formation markers). BTM levels correlate with histomorphometric indices of bone turnover, and higher BTM levels are associated with faster bone loss and fractures (9-11). The International Osteoporosis Foundation (IOF) and International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) currently recommend using serum collagen type I C-telopeptide (CTX, a bone resorption marker) and serum procollagen type I N-terminal propeptide (PINP, a bone formation marker) as reference BTMs (12).

The longitudinal, contemporaneous trajectories of CTX and PINP from pre- to postmenopause have not been described. Cross-sectional studies report that CTX and PINP levels were greater in post- vs premenopausal women; however, cross-sectional designs cannot discern the onset of MT-related changes in bone turnover (1, 13-17). One longitudinal analysis of urinary collagen type I N-telopeptide (u-NTX), another bone resorption marker, revealed that it increased ∼40% from pre- to postmenopause (3). However, to discern when and how the balance of resorption and formation changes during the MT, a longitudinal analysis of bone resorption and formation markers is necessary (1).

The primary objective of this study, therefore, was to characterize the longitudinal trajectories of CTX and PINP before, during, and after the MT, using stored biological samples and data from the Study of Women’s Health Across the Nation (SWAN). Specifically, we assayed CTX and PINP using banked serum samples, and modeled repeated measures of these BTMs in relation to years before or after the FMP. We examined the associations of CTX and PINP with time from FMP because menopause-related changes in bone physiology are more strongly tied to time relative to FMP than to clinical bleeding patterns or to chronological age (3-5, 18). Our secondary objective was to examine the possible effects of race/ethnicity, body mass index (BMI), and age at FMP on rates of CTX and PINP change.

Methods

SWAN is an ongoing, multicenter, multiracial/multi-ethnic, longitudinal cohort study of the MT and aging. At the SWAN baseline visit, participants were 42 to 52 years old, had an intact uterus and ≥1 ovary, and were in premenopause (no change from usual menstrual bleeding) or early perimenopause (less predictable menstrual bleeding, but at least once every 3 months). Women were recruited at 7 clinical sites, with each recruiting White women plus women from one other racial/ethnic group: Black (Boston, Chicago, Detroit, Pittsburgh), Chinese (Oakland), non-White Hispanic (Newark), and Japanese (Los Angeles). The full SWAN cohort included 3302 women. The SWAN Bone Cohort included a subset of 2365 participants from 5 sites (excluding Chicago and Newark). Since study inception in 1996, one baseline and 16 follow-up visits have occurred at a median (p25, p75) of 1.1 (1.0, 1.4) years between consecutive visits. Seventy-four percent of surviving women participated in the most recent follow-up visit (conducted between March 2015 and March 2018). All participants provided written informed consent, and each site obtained institutional review board approval.

Study Sample

Starting with the SWAN Bone Cohort (N = 2365), we derived our study sample by excluding those who did not have a natural menopause and known FMP date, censoring participants at first use of a bone-active medication, and excluding those who did not have banked sera collected before 10 Am from ≥2 study visits during the interval spanning 4 years before to 4 years after the FMP (the period when we expected CTX and PINP to increase most rapidly). The resulting study sample consisted of 541 women, among whom we conducted 3340 repeated measurements of CTX and PINP (Fig. 1). Bone-active medications were defined as drugs that are beneficial (hormone therapy, calcitonin, calcitriol, bisphosphonates, denosumab, parathyroid hormone) or detrimental (oral or injectable glucocorticoids, aromatase inhibitors, gonadotropin releasing hormone agonists, or anti-epileptics) to bone health.

Figure 1.

Derivation of study sample. Bone-active medications were drugs that are beneficial (hormone therapy, calcitonin, calcitriol, bisphosphonates, denosumab, parathyroid hormone) or detrimental (oral or injectable glucocorticoids, aromatase inhibitors, gonadotropin releasing hormone agonists, or anti-epileptics) to bone health. Abbreviation: FMP, final menstrual period.

Outcomes

We measured CTX and PINP, the analysis outcomes, using stored serum samples acquired at fasting blood draws; SWAN protocol stipulated that phlebotomy be performed before 10 Am, but a small proportion of samples did not fulfill this requirement. To minimize preanalytical variability in CTX measurements resulting from diurnal variation (12, 19-21), we only used sera collected before 10 Am. Specimens were stored at a central repository at −80 °C and did not undergo prior free-thaw cycles. Based on sample availability, we assayed CTX and PINP at least 2 and up to 10 times per participant, as follows: 1) at least 2 and up to 8 times during the interval spanning 4 years before to 4 years after the FMP (when we expected the greatest changes in CTX and PINP); 2) at the first (earliest) study visit that occurred prior to 4 years before the FMP (if sample was available); and 3) the last (latest) study visit that occurred subsequent to 4 years post FMP (if sample was available). Stored serum samples from the SWAN baseline visit are reserved and were, therefore, not available to the current project. We refer to the first instance at which we assayed CTX and PINP as our study’s initial visit.

The Bone Biochemistry Laboratory at the University of Sheffield performed all CTX and PINP assays using electrochemiluminescence (Cobas e411 analyzer, Roche Diagnostics, Germany). Assays for the same individual were performed in a single batch. Based on duplicate measurements from 200 randomly selected women, intra-assay coefficients of variation were 3.4% (CTX) and 3.5% (PINP). Inter-assay coefficients of variation were 5.1% for CTX and 3.3% for PINP.

Primary Predictor

The primary exposure was the number of years before or after the FMP that the BTMs were obtained. SWAN defined the FMP date as the last reported occurrence of menstrual bleeding at the visit immediately preceding the first visit at which the woman was classified as postmenopausal (had experienced ≥12 consecutive months of amenorrhea). We computed years from FMP using the FMP date and the collection dates of serum for CTX and PINP measurements. Using this construct, the FMP date equals time 0. Values for years from FMP are negative if BTMs were measured before the FMP date and positive if they were measured after the FMP (eg, FMP −3 years means 3 years prior to, and FMP +3 years means 3 years after the FMP date).

Covariates

Covariates included self-designated race/ethnicity (Black, Chinese, Japanese, or White), BMI, and age at FMP. SWAN measured body weight and height using calibrated scales and stadiometers. We calculated BMI as weight in kilograms divided by the square of height in meters.

Data Analysis

Using means (SD) for normally distributed continuous variables, medians (p25, p75) for continuous variables with skewed distributions, and counts (percentages) for categorical variables, we summarized the characteristics of the SWAN Bone Cohort and the current study sample. Because CTX and PINP values were not normally distributed, we tested whether their values in Black, Chinese, or Japanese women were significantly different than in White women using the Wilcoxon-Mann-Whitney test (separate tests compared values in each racial/ethnic minority group with those of White participants).

We characterized the trajectories of CTX and PINP relative to years from the FMP using a 3-step approach (4). In Step 1, we examined the shape of the CTX and PINP trajectories from pre- to postmenopause using LOESS plots of natural log-transformed CTX [ln(CTX)] or natural log-transformed PINP [ln(PINP)] as functions of years before or after the FMP. These plots suggested that both BTM trajectories were piecewise linear, with 3 linear segments and 2 inflection points where the CTX and PINP trajectory slopes changed (one at FMP −3 years and a second at FMP +3 years).

In Step 2, we formally tested whether the slopes of ln(CTX) or ln(PINP) were significantly different before and after each candidate inflection point identified in Step 1. Specifically, we used mixed effects linear regression (without covariate adjustment) to fit piecewise linear growth curves to repeated measures of ln(CTX) or ln(PINP) as functions of years from FMP, using linear splines with knots at FMP −3 years and FMP +3 years. To account for within-woman correlations, we included random effects for the intercept and 3 slopes (before FMP −3 years [premenopause], FMP −3 years to FMP +3 years [the MT], and after FMP +3 years [postmenopause]). For both BTMs, slope changes were statistically significant at each of the 2 candidate knots (all 4 P values for slope change <.0001).

After confirming that there were statistically significant differences in slope between successive linear segments, in Step 3, we quantified the annualized rates of change in ln(CTX) and ln(PINP) during each piecewise segment and examined whether race/ethnicity, BMI, or age at FMP influenced the rates of ln(CTX) or ln(PINP) change in each segment. Specifically, to the mixed effects linear regression models derived from Step 2, we added, as fixed effects on the intercept and 3 slopes, race/ethnicity, BMI at the initial visit, age at FMP, and SWAN study site. When constructing models, we stipulated the referent participant as White with overall sample mean values for age at FMP (52.47 years) and BMI (27.12 kg/m²). Thus, we first report annualized changes in ln(CTX) and ln(PINP) for the referent individual, followed by how these slopes vary by race/ethnicity, age at FMP, or BMI. For ease of interpretation, we also converted these slopes to percent change in each raw (non-log-transformed) BTM per year, and report both versions in the text.

Based on repository sample availability, participants had between 2 and 10 BTM assessments. To examine whether differences in sample availability among participants within the analysis sample would yield different findings, we conducted sensitivity analyses in which we re-ran the above models, restricted to participants who had at least the median number (≥7) of BTM measurements.

Results

Participant Characteristics

The study sample consisted of 126 Black (24%), 90 Chinese (16%), 87 Japanese (16%), and 238 White (44%) women. At SWAN baseline, relevant characteristics of the analysis sample (N = 541) and parent SWAN Bone Cohort (N = 2365) were similar (Table 1).

Table 1.

Sample descriptives^a for the SWAN Bone Cohort and current study sample

	SWAN Bone Cohort	Study sampleb
	N = 2335	N = 541
Time-fixed variables
Race/ethnicity
Black	658 (28)	126 (24)
Chinese	245 (11)	90 (16)
Japanese	266 (11)	87 (16)
White	1166 (50)	238 (44)
Age at FMP, yearsc	52.16 (2.80)	52.46 (2.35)
Characteristics at the SWAN baseline visit
Age, years	46.34 (2.67)	46.02 (2.47)
Time from FMP, yearsc	−5.81 (3.35)	−6.44 (2.06)
BMI, kg/m2	27.75 (6.98)	26.77 (6.46)

Abbreviations: BMI, body mass index; FMP, final menstrual period; u-NTX, urine collagen type I N-telopeptide.

^a Mean (SD) for continuous, normally distributed variables (age, time from FMP, BMI, age at FMP); median (p25, p75) for continuous, nonnormally distributed variables (u-NTX); count (%) for categorical variables.

^b Study Sample: SWAN Bone Cohort participants with a known FMP date, did not use a bone-active medication during the course of the SWAN study, and had banked serum samples collected before 10 Am from ≥2 study visits between FMP −4 years and FMP +4 years. See “Methods” and Fig. 1 for sample derivation.

^c Negative numbers indicate the number of years prior to the date of the FMP. Age at FMP and time from FMP at the SWAN baseline visit were calculated retrospectively after a study participant reported ≥12 consecutive months of amenorrhea and the FMP date identified. See “Methods” for details of FMP dating.

At the initial visit (the first visit at which study sample participants had available samples for CTX and PINP measurements), average age was 47.81 years, mean time from FMP −4.66 years, and average BMI 27.12 kg/m². BMI values ≥30 kg/m² were present in 51% of Black, 4% of Chinese, 5% of Japanese, and 30% of White participants. Median CTX and PINP levels were 269 pg/mL and 42.9 ng/mL, respectively. Median CTX values were similar among Black (287 pg/mL) and White (298 ng/mL) women (P = .3); compared to values in White women, CTX levels were lower in Chinese (256 pg/mL, P = .003) and Japanese (239 pg/mL, P = .0001) participants. Like CTX, median PINP levels were similar in Black (45.8 ng/mL) and White participants (45.2 ng/mL) (P = .3), but, in comparison to White women, were lower in Chinese (36.5 ng/mL, P < .0001) and Japanese (39.1 ng/mL, P = .001) women (Table 2).

Table 2.

Sample descriptives^a for the study sample at initial study visit^b

	Study sample N = 541
Age, years	47.81 (2.64)
Time from FMP, yearsc	−4.66 (2.38)
BMI, kg/m2	27.12 (6.56)
CTX, pg/mL
All	269 (194, 368)
Blackd	287 (188, 385)
Chinesee	255 (181, 326)
Japanesee	239 (171, 291)
White	298 (211, 399)
PINP, ng/mL
All	42.9 (33.6, 54.9)
Blackd	45.8 (35.3, 60.8)
Chinesee	36.5 (29.3, 45.5)
Japanesee	39.1 (28.4, 50.4)
White	45.2 (35.6, 55.8)

Abbreviations: FMP, final menstrual period; CTX, collagen type I C-telopeptide; PINP, procollagen type I N-terminal propeptide.

^a Mean (SD) for continuous, normally distributed variables (age, time from FMP, BMI); median (p25, p75) for continuous, nonnormally distributed variables (CTX, PINP); count (%) for categorical variables.

^b Initial study visit defined as the first SWAN study visit at which participants had banked serum sample to assay CTX and PINP. NOTE: Because banked serum samples from the SWAN baseline visit were not available, we do not have CTX and PINP measurements from this time point.

^c Negative numbers indicate the number of years prior to the date of the FMP. Time from FMP at the initial study visit was determined retrospectively after a study participant reported ≥12 consecutive months of amenorrhea and the FMP date identified. See “Methods” for details of FMP dating.

^d P > .3 for comparison with values among White women (Wilcoxon-Mann-Whitney test).

^e P < .003 for comparison with values among White women (Wilcoxon-Mann-Whitney test).

In total, we had 3340 repeated assessments of CTX and PINP among the 541 participants. Sera for BTM measurements were collected over a period spanning FMP −10 to FMP +13 years. The median (p25, p75) number of CTX and PINP assessments was 7 (5, 8), and median (p25, p75) length of follow-up 13.6 years (10.2, 14.3). Baseline characteristics were similar between those who had less than the median number of samples vs those with greater than or equal to the median number of samples (data not shown).

Annual Rates of Change in CTX in the Referent Participant, and the Effects of Race/Ethnicity, BMI, and Age at FMP

The first row of data in Tables 3 shows the rates of change (slopes) in ln(CTX) during premenopause, the MT, and postmenopause for the referent participant; we summarize these slopes below. For ease of interpretation, we also present the CTX slopes as percent change per year in the raw (non-log-transformed) BTM.

Table 3.

Annual rates of change^a in natural log-transformed CTX [ln(CTX)] relative to FMP date, and the influence of race/ethnicity, BMI, and age at FMP on ln(CTX) change rates

	Annual slopes for ln(CTX)a during premenopause, the MT, and postmenopause (95% CI)
	Premenopause	MT	Postmenopause
	Before FMP −3 yearsb	FMP −3 to FMP +3 years	After FMP +3 yearsb
Reference slopec,f
White	+0.004 (−0.014, +0.022)	+0.098 (+0.089, +0.108)	−0.032 (−0.042, −0.022)
Differences in slopes vs reference slopes by race/ethnicity, BMI, or age at FMPd,e,g,h
Race/ethnicity
Black	+0.010 (−0.020, +0.041)	+0.01 (−0.015, +0.017)	+0.010 (−0.008, +0.028)
Chinese	+0.033 (−0.003, +0.068)	+0.023 (+0.005, 0.042)	−0.009 (−0.027, 0.008)
Japanese	−0.028 (−0.060, +0.005)	+0.051 (+0.032, +0.069)	−0.026 (−0.046, −0.001)
Per kg/m2 increment in BMI	−0.001, (−0.003, +0.001)	−0.002 (−0.004, −0.001)	+0.001 (−0.001, +0.002)
Per 1-year increment in age at FMP	−0.003 (−0.008, +0.003)	+0.002 (−0.001, +0.005)	−0.001 (−0.004, +0.001)

Abbreviations: BMI, body mass index; FMP, final menstrual period; ln(CTX), natural log-transformed collagen type I C-telopeptide.

^a Rates estimated using repeated measures, mixed effects linear regression with years before or after the FMP (modeled using 3-piece linear splines with knots at FMP −3 years and FMP +3 years) as primary predictor, and ln(CTX) as outcome. Models adjusted for race/ethnicity, BMI at the initial visit (first visit with CTX and PINP measurements), age at FMP, and study site. Random effects for the intercept and slopes for each spline segment were included to account for within-woman correlation between repeated observations.

^b Premenopausal slopes estimated using CTX measurements from FMP −10 to FMP −3 years. Postmenopausal slopes estimated using CTX measurements from FMP +3 to FMP +13 years.

^c Reference slopes represent change in ln(CTX) per year among White women with overall sample-average values of BMI (27.12 kg/m²) and age at FMP (52.47 years).

^d BMI at initial visit.

^e To calculate annual change in ln(CTX) for a woman who has non-reference characteristics (ie, is not White, or does not have sample-average values of BMI or age at FMP), add the difference in ln(CTX) slope for a given characteristic to the reference ln(CTX) slope.

^f In the referent sample, change in ln(CTX) can be converted to annual % change in the raw CTX using the following formula: 100*(exp^{(reference slope)—}1). Thus, for the referent participant, annual % change in raw CTX was +0.4% (95% CI −1.4 to +2.2%) per year in premenopause, +10.3% (95% CI 9.3 to 11.3%) per year during the MT, and −3.1% (95% CI −4.1 to −2.1%) per year in postmenopause.

^g To obtain annual % change in raw CTX in women with a specific race/ethnicity apply the following formula: 100*(exp^{[reference slope + difference in slope]} − 1), where difference in slope is that for the specific racial/ethnic group. Here, we present the CTX slopes (in % change per year) for non-White women that were significantly different than those for White participants. Compared to rates of CTX change in White women, CTX increased faster during the MT in Chinese and Japanese women, and decreased faster in postmenopause in Japanese women. Specifically, CTX changed by +13.0% (95% CI, 11.9 to 14.0%) per year in Chinese women during the MT, + 16.0% (95% CI, 14.9 to 17.2%) per year in Japanese women during the MT, and −5.6% (−6.5 to −4.7%) per year in Japanese women in postmenopause.

^h To obtain annual % change in raw CTX in women with a specific BMI, or age at FMP, apply the following formula: 100*(exp^{[reference slope + difference in slope]} − 1), where difference in slope is that for the specific characteristic.

CTX did not change significantly in premenopause (P = .6). During the MT (FMP −3 through FMP +3 years), ln(CTX) increased by 0.098 per year (P < .0001), corresponding to a 10.3% (95% CI, 9.3 to 11.4%) per year rise in raw CTX. Consequently, the cumulative CTX gain over the 6-year MT period was 61.9%. From its peak at the end of the MT, ln(CTX) decreased by 0.032 per year (P < .0001) in postmenopause, equivalent to a 3.1% (95% CI, 2.1 to 4.1%) decrease per year in CTX.

Table 3 also presents the differences (relative to the reference sample) in slopes for ln(CTX), depending on race/ethnicity, BMI, and age at FMP. To obtain the rate of ln(CTX) change for women who were not White and did not have sample-average BMI and age at FMP characteristics, we add the effect size estimates for race/ethnicity, BMI, or age at FMP to the referent slope.

Relative to White participants, Chinese women experienced faster increases in CTX during the MT (0.122 per year in ln(CTX) or 13.0% per year in raw CTX, P = .01). Chinese-specific slopes did not differ from the referent slope in pre- or postmenopause. In Japanese women, CTX increased faster (0.149 per year in ln(CTX) or 16.0% per year in CTX, P < .0001) during the MT and decreased faster in postmenopause (0.058 per year in ln(CTX) or 5.6% per year in CTX, P = .01). CTX slopes for Black women were not significantly different from those for White women during any MT stage (P > .8 for each comparison). Figure 2 depicts the model-predicted trajectories of CTX relative to years from the FMP in each of the 4 race/ethnicity groups represented in this study.

Figure 2.

Model-predicted trajectories of CTX in relation to the date of the FMP, by race/ethnicity. Values assume the sample mean values for age at FMP (52.47 years) and baseline BMI (27.12 kg/m²).

Women with greater BMI experienced a lesser increase in CTX during the MT: each 1 kg/m² increment in BMI was associated with a 0.002 per year slower rise in ln(CTX) (P < .0001) or 0.2% per year slower gain in raw CTX. BMI was unrelated to CTX slopes in pre- or postmenopause (P > .1 for each comparison). Age at FMP was not associated with rates of CTX change in any MT phase (P > .1 for each comparison).

Annual Rates of Change in PINP in the Referent Participant, and the Effects of Race/Ethnicity, BMI, and Age at FMP

Analogous to Table 3, Table 4 presents the slopes for ln(PINP) during each MT stage (premenopause, the MT, postmenopause) in the referent participant, and differences in ln(PINP) slopes (vs the referent) by race/ethnicity, BMI, and age at FMP. As with the CTX slopes above, the text below describes the PINP slopes using annual change in ln(PINP) and annual percent change in raw PINP.

Table 4.

Annual rates of change^a in natural log-transformed PINP [ln(PINP)] relative to FMP date, and the influence of race/ethnicity, BMI, and age at FMP on ln(PINP) change rates

	Annual slopes (scaled by 100) for ln(PINP)a during premenopause, the MT, and postmenopause (95% CI)
	Premenopause Before FMP −3 yearsb	MT FMP −3 to +3 years	Postmenopause After FMP +3 yearsb
Reference slopesc,f
White	−0.005 (−0.024, +0.013)	+0.073 (+0.063, +0.083)	−0.029 (−0.040, −0.019)
Differences in slopes vs reference slopes by race/ethnicity, BMI, or age at FMPd,e,g,h
Race/ethnicity
Black	+0.031 (−0.001, +0.062)	−0.015 (−0.032, +0.002)	+0.002 (−0.016, 0.021)
Chinese	−0.000 (−0.037, +0.037)	+0.015 (−0.005, +0.034)	−0.002 (−0.020, +0.017)
Japanese	+0.004 (−0.029, +0.038)	+0.028 (+0.003, +0.042)	−0.023 (−0.045, −0.002)
Per kg/m2 increment in baseline BMI	−0.000 (−0.002, +0.002)	−0.003 (−0.003, −0.001)	+0.001 (−0.001, +0.002)
Per 1-year increment in age at FMP	+0.003 (−0.003, +0.008)	−0.002 (−0.005, +0.001)	−0.001 (−0.004, +0.002)

Abbreviations: BMI, body mass index; FMP, final menstrual period; MT, menopause transition; PINP, N-terminal propeptide of type I collagen.

^a Rates estimated using repeated measures, mixed effects linear regression with years before or after the FMP (modeled using 3-piece linear splines with knots at FMP −3 years and FMP +3 years) as primary predictor, and ln(PINP) as outcome. Models adjusted for race/ethnicity, BMI at the initial visit (first visit with CTX and PINP measurements), age at FMP, and study site. Random effects for the intercept and slopes for each spline segment were included to account for within-woman correlation between repeated observations. Slopes for log-transformed PINP were converted to annualized percentage change for ease of interpretation. Reported annual slopes of ln(PINP) are scaled by 100.

^b Premenopausal slopes estimated using PINP measurements from FMP −10 to FMP −3 years. Postmenopausal slopes estimated using PINP measurements from FMP +3 to FMP +13 years.

^c Reference slopes represent change in ln(PINP) per year among White women with overall sample-average values of BMI (27.12 kg/m²) and age at FMP (52.47 years).

^d BMI at initial visit.

^e To calculate annual change in ln(PINP) for a woman who has non-reference characteristics (ie, is not White, or does not have sample-average values of BMI or age at FMP), add the difference in ln(PINP) slope for a given characteristic to the reference ln(PINP) slope.

^f In the referent sample, change in ln(PINP) can be converted to annual % change in the raw PINP using the following formula: 100*(exp^{(reference slope)} − 1). Thus, for the referent participant, annual % change in raw PINP was −0.5% (95% CI, −2.3 to +1.3%) per year in premenopause, +7.5% (95% CI, 6.5 to 8.6%) per year during the MT, and −2.9% (95% CI, −3.9 to −1.9%) per year in postmenopause.

^g To obtain annual % change in raw PINP in women with a specific race/ethnicity apply the following formula: 100*(exp^{[reference slope + difference in slope]} − 1), where difference in slope is that for the specific racial/ethnic group. Here, we present the PINP slopes (in % change per year) for non-White women that were significantly different than those for White participants. Compared to White women, Japanese participants had faster increase in PINP during the MT and faster decline in PINP in postmenopause. Specifically, in Japanese women, PINP changed by +10.0% (95% CI, 8.9 to 11.1%) per year during the MT, and −5.1% (−6.1 to −4.1%) in postmenopause.

^h To obtain annual % change in raw CTX in women with a specific BMI, or age at FMP, apply the following formula: 100*(exp^{[reference slope + difference in slope]} − 1), where difference in slope is that for the specific characteristic.

In each midlife stage, the direction of change for PINP was the same as for CTX, but PINP increased more slowly than CTX during the MT. In the referent women, PINP did not change significantly in premenopause (P = .5). Log-transformed PINP increased by 0.073 per year (P < .0001) during the MT. This corresponds to a 7.5% (95% CI, 6.5 to 8.6%) per year rise in PINP, amounting to a total gain of 45.2% during the 6-year MT interval. Starting in postmenopause, PINP began to decrease, from its end-of-MT high, by 0.029 per year in ln(PINP) or 2.9% (95% CI, 1.9 to 3.9%) per year in PINP (P < .0001).

Compared to rates of PINP change in White women, PINP increased faster (0.095 per year in ln(PINP) or 10.0% per year in raw PINP, P = .04) during the MT and decreased faster (0.052 per year in ln(PINP) or 5.1% per year in PINP, P = .01) in postmenopause in Japanese women. PINP slopes for Black and Chinese women did not differ from those of White women in any stage (P > .1 for each comparison). Figure 3 shows model-predicted PINP trajectories in relation to FMP time, by race/ethnicity.

Figure 3.

Model-predicted trajectories of PINP in relation to the date of the FMP, by race/ethnicity. Values assume the sample mean values for age at FMP (52.47 years) and baseline BMI (27.12 kg/m²).

In women of all racial/ethnic groups, the 95% CI of annual percent change in CTX during the MT did not overlap with that of PINP (CIs for the referent group are presented in the text above [data not shown for nonreferent groups]). This indicates that rates of CTX and PINP change during the MT interval were significantly different, with gain in CTX outpacing gain in PINP. In contrast, the 95% CIs of the annual percent changes in CTX and PINP in postmenopause were essentially the same, indicating that both BTMs declined at similar rates during this period.

During the MT, greater BMI related to slower PINP rise, with each 1 kg/m² increment in BMI associated with a 0.002 per year slower increase in ln(PINP), or 0.2% per year slower rise in PINP (P < .0001). PINP change rates did not differ by BMI in pre- or postmenopause (P > .1 for each comparison). Age at FMP was not associated with PINP slope in any midlife period (P > .1 for each comparison).

Sensitivity Analyses

To examine whether differences in repository sample availability among participants influenced our findings, we re-ran models that were restricted to participants with greater than or equal to the median number (≥7) of BTM measurements. CTX and PINP slopes in this subset of participants (N = 285, observations = 2222) were essentially identical to those in the full analysis sample (data not shown).

Discussion

Using data from a diverse sample of women, this study described the longitudinal patterns of change, from pre- to postmenopause, in currently recommended markers of bone resorption (CTX) and bone formation (PINP). In the referent group (White women with sample-average characteristics), CTX and PINP did not change significantly during premenopause. At the start of the MT (spanning the period from 3 years before to 3 years after the FMP), CTX and PINP rose, reflecting an acceleration in bone turnover. During the MT, gain in CTX (10.3% annually) outpaced gain in PINP (7.5% annually), consistent with progressive osteoclast hyperactivity and negative bone balance (3, 22), with balance defined herein as total-body balance between bone resorption vs formation. In postmenopause, CTX and PINP decreased by 3.1% and 2.9% per year, respectively, consistent with bone turnover slowing from its peak rate at the end of the MT. Postmenopausal CTX and PINP decline rates were similar, indicating that bone resorption continued to exceed bone formation. Compared to White participants, Chinese women demonstrated a faster CTX increase during MT. In Japanese women, relative to White participants, CTX and PINP rose more during the MT and decreased more in postmenopause. Greater BMI was associated with less CTX and PINP gain during the MT.

To our knowledge, prior studies of CTX and PINP in relation to menopause are confined to cross-sectional comparisons of BTM levels in samples of different pre- or postmenopausal women (13-17). In these cross-sectional analyses, mean CTX and PINP levels were 32% to 86% and 17% to 53% greater in post- vs premenopausal samples (13-17). Our current longitudinal study adds to this existing knowledge base. First, it quantifies the within-individual changes in BTM levels from pre- to postmenopause. Second, it elucidates the temporal course of CTX and PINP change. Specifically, we demonstrate that CTX and PINP begin to increase 3 years before the FMP, gain rapidly during the ensuing 6 years, and peak 3 years after the FMP. Prior SWAN analyses of the MT-related trajectories of BMD, trabecular bone score, and hip strength found that the period of fastest bone loss falls within the interval of rapid CTX and PINP rise (4, 5, 18).

One prior SWAN analysis described the longitudinal trajectory of a different bone resorption marker (u-NTX) from pre- to postmenopause (3). Like CTX, u-NTX, did not change significantly in premenopause, increased rapidly during the MT, and decreased slowly in postmenopause. A comparison of results from the current vs previous SWAN analyses suggest that CTX may be a more sensitive marker of bone resorption than u-NTX. In the present analysis, the within-woman rise in CTX, from pre- to postmenopause, was approximately 60%. In contrast, the former SWAN analysis reported a within-individual pre-to-postmenopause increase in u-NTX of roughly 40% (3). While the current and former study samples were not identical, they had similar racial/ethnic composition and baseline characteristics. Further supporting CTX’s greater sensitivity to change relative to u-NTX are clinical trial data showing that CTX declines more than u-NTX in response to bisphosphonate therapy (23). In some studies, CTX also had lower intra-individual variability than u-NTX (24). Thus, compared to u-NTX, CTX appears better suited to estimate longitudinal changes in total-body bone resorption.

Our results reveal inter-racial/ethnic differences in rates of CTX and PINP change during the MT and postmenopause. These distinctions are partly concordant with previously observed racial/ethnic differences in BMD decline over the same time frames (4). Compared to White participants, CTX increased faster during the MT in Chinese participants. Japanese women experienced faster MT-related gains in CTX and PINP, with the difference in CTX slope more than double the difference in PINP slope. Thus, relative to White women, bone balance during the MT was more negative in Chinese and Japanese women, consistent with higher, contemporaneous Chinese- and Japanese-specific rates of BMD loss (4). Black and White participants had similar rates of CTX and PINP increase during the MT, which does not parallel the lower MT-related BMD loss rate observed in Black vs White women (4). Possibly, these incongruous findings could be related to inter-racial/ethnic differences in bone mineralization, which would not be captured by PINP, a marker of type I collagen synthesis (25). Using bone histomorphometry, a previous study showed that Black women had longer bone formation periods than White women; more prolonged bone formation can result in greater secondary mineralization of bone (26). Dual-energy x-ray absorptiometry (DXA) scans quantify the bone mineral content within a region of interest, a composite of the amount of collagen matrix synthesized and the degree of bone mineralization. If bone mineralization during the MT were greater in Black women, then decline in densitometric BMD would be slower. To test this hypothesis, future studies could examine the trajectory of a marker that captures the mineralization phase of bone formation (ie, bone-specific alkaline phosphatase).

In addition to rates of CTX and PINP change, race/ethnicity was associated with BTM level. Our observed racial/ethnic differences in CTX and PINP values at the initial visit are relevant to ongoing efforts to establish normative BTM reference ranges. The National Bone Health Alliance recommends establishing country-specific reference ranges for CTX and PINP to aid their clinical interpretation (27). For females, the “normal” population consists of healthy, premenopausal women, presumed to have optimal bone health (ie, low bone turnover and stable BMD) (6). Few studies have described the reference intervals for CTX and PINP in U.S. women, and those that did were not powered to test for differences by race/ethnicity (28, 29). In our community-based sample of U.S. women, CTX and PINP levels at the initial visit, obtained before bone turnover accelerated, were lower in Chinese and Japanese women relative to Black and White participants, underscoring the importance of considering race/ethnicity-specific reference intervals.

We acknowledge this study’s limitations. First, the SWAN Bone Cohort did not include a specific sample of non-White Hispanic participants, because the site that recruited women from this racial/ethnic group did not participate in the bone protocol. Second, our study sample size was large enough to detect racial/ethnic differences in CTX and PINP slopes, but not sufficiently large to test whether BMI and age at FMP modified the rates of CTX or PINP change within a racial/ethnic group. Third, because our study focused on characterizing the trajectories of currently recommended BTMs, we did not include markers of other phases of the bone remodeling cycle (eg, sclerostin [a regulatory molecule], tartrate-resistant acid phosphatase type 5b [a marker of osteoclast number], or bone-specific alkaline phosphatase [a marker of bone mineralization]). Follow-up studies can include these markers to examine how other facets of bone remodeling change across the MT. Fourth, some investigators have suggested that CTX is more stable in plasma than in serum (6). As plasma samples were not available to us, we measured CTX using banked sera. We maximized analyte stability by promptly storing serum, maintaining samples at −80 °C, and not thawing them until assay performance. Additionally, physical activity can affect BTM levels, but we were unable to adjust for this variable as SWAN did not collect physical activity data at all study visits. Lastly, not all participants in the analysis sample had the full complement of 10 BTM measures; however, results were unaltered by restriction to a sample with at least 7 BTM observations.

In conclusion, this longitudinal analysis newly describes the temporal patterns of change in currently recommended markers of bone resorption (CTX) and formation (PINP). We uncovered that the MT-related rise in CTX and PINP occurs 3 years prior to the FMP. During the subsequent 6 years, CTX and PINP increase rapidly, with gain in CTX outpacing gain in PINP. Both BTMs reach their respective peaks 3 years post FMP, after which they slowly decrease. Our study also revealed racial/ethnic differences in rates of CTX and PINP change during the MT and in postmenopause. The mechanisms that underlie these racial/ethnic differences in bone remodeling, and how these distinctions may relate to fracture risk await elucidation.

Abbreviations

BMD

bone mineral density

BMI

body mass index

BMU

basic multicellular unit

BTM

bone turnover marker

CTX

collagen type I C-telopeptide

FMP

final menstrual period

MT

menopause transition

PINP

procollagen type I N-terminal propeptide

u-NTX

urinary collagen type I N-telopeptide

SWAN

Study of Women’s Health Across the Nation

Funding

The Study of Women’s Health Across the Nation (SWAN) has grant support from the National Institutes of Health (NIH), DHHS, through the National Institute on Aging (NIA), the National Institute of Nursing Research (NINR) and the NIH Office of Research on Women’s Health (ORWH) (Grants U01NR004061; U01AG012505, U01AG012535, U01AG012531, U01AG012539, U01AG012546, U01AG012553, U01AG012554, U01AG012495, and U19AG063720). This analysis was supported by NIH through the National Institute of Arthritis and Musculoskeletal and Skin Diseases (Grant R01AR075729).

Disclosures

A.S., A.S.K., F.G., and G.A.G. have nothing to disclose. R.E. receives consultancy funding from Immunodiagnostic Systems, Sandoz, Samsung, CL Bio, Biocon, Takeda, UCB; meeting presentations for Pharmacosmos, Alexion, UCB and Amgen; and grant funding from Alexion.

Data Availability

Some or all datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.