Biological Variability of Serum and Urinary N-Telopeptides of Type I Collagen in Postmenopausal Women



Measurement of N-telopeptides of type I bone collagen (NTX) provides a specific indicator of the current level of bone resorption. The biological intrasubject variability of NTX in urine and serum was studied in 277 postmenopausal women, mean age, 63.6 years ± 10.2 (±SD) years. Second-morning void urine and serum specimens were collected at baseline and for two consecutive days to determine short-term variability (%CV). Long-term variability was determined by comparing NTX results at baseline and two consecutive months. Subjects were instructed to maintain current diet, lifestyle, and medications during the study. The median short-term %CV was 13.1% for urine NTX. This compared with 6.3% for serum NTX. Calculation of long-term %CV showed similar trends, with the %CV for NTX measured in serum (7.5%) lower than when measured in urine (15.6%). Using the least significant change (LSC) calculation, our data show that to be 90% confident that a decrease in NTX after initiation of antiresorptive therapy in an individual patient is not caused by variability alone, a 31% decrease in urine NTX and a 14% decrease in serum NTX is required. As reported changes in NTX caused by antiresorptive therapy are greater than these calculations; our results support the use of either specimen to measure NTX to monitor the effect of therapy.


Antiresorptive therapy for the prevention and treatment of osteoporosis results in a decrease of bone resorption markers within weeks and bone formation markers within months of therapy initiation.(1–4) Thus, bone resorption markers are an attractive approach to monitoring treatment response. The short-term decrease in bone resorption markers correlate with the long-term changes in bone mineral density (BMD).(1,2) Newer markers of bone resorption have been introduced that are more specific to the metabolic breakdown of bone collagen, such as deoxypyridinoline and the cross-linked N-telopeptides and C-telopeptides of type I bone collagen (NTX and CTX, respectively.)(5–9) These bone resorption markers are measured most often in urine.

Sources for variability in bone resorption markers include circadian rhythm, fluctuations in renal function, and dietary calcium intake. The concentration of these markers is corrected for urine dilution using creatinine concentration, introducing another potential source of variability. Serum-based markers of bone turnover tend to show less variability as compared with urine-based markers.(10) A serum-based test to measure NTX that utilizes the same antibody as the urine NTX test has been described recently.(11–16) Measurement of NTX in serum may offer bone resorption testing with decreased intrasubject variability.

The primary aim of this study was to evaluate the short-term and long-term intrasubject variability of urinary NTX. Variability was then related to age, hormone replacement therapy (HRT) use, and other determinants of bone turnover including BMD and diet. Variability of serum NTX was measured in the new immunoassay.


Two hundred seventy-seven healthy postmenopausal women participated at four clinical centers in this evaluation of intrasubject variability of NTX. Postmenopause was defined as at least 6 months from the last menstrual cycle. Follicle-stimulating hormone (FSH) levels were obtained in women less than 60 years old to confirm menopausal status. Enrollment was stratified by use or nonuse of HRT and by age group (45–54 years, 55–64 years, 65–74 years, and 75–85 years). Women with surgically induced menopause before the age 45 years were excluded, as were individuals diagnosed with diseases known to affect bone. Subjects included in the non-HRT group had not received HRT for at least 6 months before enrollment and had never taken bisphosphonates. Those in the HRT group took 0.625 mg Premarin (Wyeth Laboratories, Taplow, Berks) or an equivalent and 2.5–5 mg Provera (Upjohn Limited, Crawley, West Sussex, U.K.) or an equivalent on a regular basis for at least 6 months before their participation. All subjects had a fasting 21-member serum chemistry panel, which included measures of liver and kidney function, performed within 3 weeks before their participation in the study to provide assurance of general good health.

Short-term and long-term intrasubject variability was evaluated. Short-term variability was determined from second-morning void urine and fasting serum specimens collected once daily for three consecutive days. Long-term variability was determined from specimens collected at baseline and monthly thereafter for 2 months, providing three specimens per subject.

Urine specimens were shipped to a central reference laboratory and tested as received to simulate clinical testing. Serum specimens were frozen at −20°C before testing by a reference laboratory. NTX levels were measured by ELISA using a specific monoclonal antibody (1H11) directed against the N-telopeptide intermolecular cross-linked domain of type I collagen of bone (Osteomark NTX Urine and Osteomark NTX Serum, Ostex International, Inc., Seattle, WA, U.S.A.). Methods for measurement of NTX in urine and serum have been described elsewhere.(6,14) Concentration of NTX in patient specimens is calculated from a standard curve of known NTX concentrations and expressed as nanomolar bone collagen equivalents (nM BCE). Urine values are corrected for dilution by urinary creatinine analysis and results are expressed in nM BCE per millimolar creatinine (nM BCE/mM creatinine). Interassay and intra-assay variability for the urine and serum NTX assays are 4% and 8% and 6.9% and 4.6%, respectively.

BMD was obtained at the lumbar spine (L1–L4) and femoral neck using dual-energy X-ray absorptiometry (DXA; Hologic, Waltham, MA, U.S.A. and Lunar, Madison, WI, U.S.A.) at the baseline visit. Quality control and phantom cross-calibration for BMD measurements were provided by the University of Washington Osteoporosis Research Group (Seattle, WA, U.S.A.).

Also at baseline, subjects at the three U.S. sites completed the Food Frequency Questionnaire developed by Willet et al. from which daily nutrient intake is estimated.(17) The participant estimated exercise levels and provided self-reports of frequency and severity of hot flashes/night sweats. Additionally, smoking status; changes in diet; and the incidence of fracture, immobilization, or surgery were recorded to document changes to these variables throughout the study. Subjects were asked to maintain their current diet and lifestyle during the study. Each subject provided a medication summary that included prescription and over-the-counter medications taken within 6 months of study entry and was maintained throughout study participation. The protocol was reviewed and approved by the institutional review board at each site, and subjects provided written informed consent before study enrollment.

Table Table 1.. Subject Demographics by HRT Group (n = 277)
 Non-HRT group (n = 150)HRT group (n = 127)
  1. Results are given as mean ± SD.

Age (Years)64 ± 1063 ± 10
Years since menopause17 ± 1116 ± 10
Height (in.)63 ± 363 ± 3
Weight (lb)150 ± 32144 ± 25
BMD—lumbar spine (g/cm2)0.899 ± 0.1710.990 ± 0.164
BMD—femoral neck (g/cm2)0.702 ± 0.1270.732 ± 0.120
Smoking currently8%8%
Exercise regularly58%61%
Family history of osteoporosis (% positive)15%26%
Caucasian race89%94%


Subject demographics are presented as mean ± SD unless otherwise noted. A Wilcoxon rank–sum or Kruskal-Wallis rank test was used to test for differences in NTX by categorical variables. The relationship between continuous variables was analyzed with Spearman rank correlation coefficients and their associated test of significance. For each subject, both the short-term and the long-term intrasubject variability was estimated using the CV. CV is the SD multiplied by 100 and divided by the mean. Summary statistics for the short-term and long-term variability are reported as median values. CVs between groups were compared using f tests. Analyses were performed using the Statistical Analysis System (SAS Institute Inc., Cary, NC, U.S.A.). All statistical tests are two-sided. A p value less than 0.05 is considered statistically significant. A “signal-to-noise ratio” was computed using the percent difference in mean NTX between the HRT users and nonusers as the signal and the long-term variability as the noise. Then the signal-to-noise ratios were compared.


Baseline demographics of the 277 participants are presented by HRT use in Table 1. Combining the HRT and non-HRT groups, the mean (±SD) years since menopause was 16.3 ± 10.6 (age, 63.6 ± 10.2 years). The majority of women enrolled were white, exercised regularly, were either nonsmokers or former smokers and did not have a family history of osteoporosis.

Table Table 2.. Baseline Urine and Serum NTX Values
 Urine NTX (nM BCE/mM creatinine)Serum NTX (nM BCE)
 Mean ± SDRangeMean ± SDRange
  1. *p < 0.001; HRT group compared with non-HRT group, Wilcoxon rank-sum test.

HRT group (n = 127)30.6 ± 12.2*9–7314.0 ± 4.2*8.2–33.9
Non-HRT group (n = 150)56.4 ± 26.915–15417.8 ± 4.97.2–39.3

In both urine and serum, HRT users had significantly lower mean NTX values (30.6 nM BCE/mM creatinine and 14.0 nM BCE) compared with women not taking HRT (56.4 nM BCE/mM creatinine and 17.8 nM BCE; p = 0.0001; Table 2). There was no significant difference in baseline urine NTX values between the four sites or by age decade in either group. Serum NTX differed by site in both the HRT and the non-HRT groups (p = 0.0005 and 0.0009, respectively).

Intrasubject variability of urine and serum NTX

Intrasubject variability of urine NTX: Short-term variability (%CV) was measured from urine specimens collected over three consecutive days. Median short-term %CV for all subjects was 13.1%, with %CV ranging from 0% to 54% (Table 3). Short-term variability in urine NTX excretion showed weak yet significant negative correlations with serum calcium (r = −0.16) and creatinine (r = −0.13). Short-term variability also differed by race and smoking status. Nonwhites (n = 24) had higher median %CV than whites (18.1% vs. 12.8%), and subjects that reported to have never smoked or be former smokers had higher %CV (13.3%) than current smokers (9.6%). Recognized risk factors for osteoporosis such as age, family history of osteoporosis, HRT use, and BMD were not associated with variability (data not shown). Significant differences in %CV were found by site (p = 0.002), with the two clinical centers in the southern latitudes reporting higher median %CV (14.8% and 14.2%) than the northern latitude sites (11.7% and 10.4%).

Long-term variability (%CV) was measured from specimens collected at baseline and two-consecutive-month time points. Variability in urine NTX excretion ranged from 1.8% to 62.5% (median 15.6%). HRT use was associated with a higher %CV (17.2%) compared with nonuse (14.2%; p = 0.05; Table 3). Weak yet significant correlations were seen between long-term variability and dietary vitamin D intake (r = −0.17), serum calcium (r = −0.18), and calcium in the diet (with or without dietary supplements; r = −0.18 and r = −0.22, respectively; p ≤ 0.01). Unlike short-term variability, long-term variability was not related to latitude among the four sites.

Intrasubject variability of serum NTX: Median short-term %CV of serum NTX measurements was 6.3% and ranged from 0.3% to 30.4% (Table 3). No significant correlations were found between the short-term variability in serum and age, BMI, or diet intake variables, nor were significant differences found by race or exercise. As with the urine NTX measurement, never or former smokers had a higher median %CV (6.5%) compared with current smokers (5.5%; p = 0.03). None of the other osteoporosis risk factors or demographic parameters were associated with short-term variability. The clinical centers in the southern latitudes again had higher median %CV (6.3% and 8.7%) compared with those in the north (5.5% and 5.8%; p = 0.02 by site comparison).

Table Table 3.. Median Short-Term and Long-Term %CV Associated with HRT Use
 Urine NTXSerum NTX
 Short-term (%CV)Long-term (%CV)Short-term (%CV)Long-term (%CV)
All subjects13.
HRT group14.317.26.47.5
Non-HRT group12.314.26.27.6

Median long-term variability of serum NTX was 7.5% and ranged from 0.4% to 42.7% (Table 3). Significant yet moderate negative correlations were found between the long-term variability and BMI (r = −0.14), femoral neck BMD (r = −0.15), and serum creatinine (r = −0.19; p ≤ 0.02). The long-term variability did not differ by the osteoporosis risk factors measured or any of the other categorical variables. Again, differences were found by clinical site (p = 0.001), but the %CV did not follow the same pattern as found with the urine assay. The median %CV at the centers in the southern latitudes were 8.0% and 6.9% whereas in the north they were 8.8% and 6.0%.

Comparison of serum and urine NTX variability and application to clinical use

A signal-to-noise ratio can be developed to compare the two NTX measures to determine the effect that variability has on the ability to measure a change in bone resorption using urine or serum NTX. The “signal” refers to the effect of HRT, in this case the difference in mean levels of NTX between the two groups. The noise refers to the long-term variability, expressed as a median C V The ratio of signal-to-noise is similar for urinary and serum NTX, indicating similar diagnostic value (Table 4).

Table Table 4.. Signal-to-Noise Ratio of Urine and Serum NTX
NTX measurePercent difference in mean NTXLong-term %CVSignal to noise ratio


Short-term and long-term intrasubject variability in urine and serum NTX were evaluated in a large cohort of postmenopausal women (n = 277). Variability results reported here can be compared with those in previous publications.(10,13,18) Serum NTX had a lower short-term and long-term %CV (6.3% and 7.5%, respectively) compared with the median %CV for urine NTX (13.1% and 15.6%).

A tentative assumption was that the antiresorptive effect of estrogen would result in decreased variability in bone resorption. HRT use did not affect short-term variability but had an influence on long-term variability. Women taking HRT had a significantly higher long-term %CV for the urine NTX measure (17.2% vs. 14.2%). This was not true for serum NTX, in which the long-term %CV was virtually identical between the two groups. It is not known if the difference in long-term variability between the HRT and non-HRT groups for urine NTX is the result of noncompliance to therapy over the study period. Although subjects were questioned regarding compliance to HRT, pill counts were not performed. Nonetheless, higher long-term variability in this group may be a reflection of inconsistent HRT usage as discussed in numerous publications. The difference between the two groups in urine NTX but not in serum NTX may be a reflection of the greater variability in urine NTX overall.

The calculation of variability, especially long-term, includes assay variability. Intra-assay variability for the urine NTX is 8% and 5% for the serum NTX test. The long-term variability of 15.6% for urine NTX compares favorably with other commonly used quantitative urine laboratory tests such as urinary calcium/creatinine ratio (43.6%) and urinary hydroxyproline/creatinine (18.7%).(19) For serum NTX, the long-term variability was 7.5%, again comparable with other serum tests such as cholesterol (8.2%) and osteocalcin (12.7%)(10,19)

As indicated by the nearly identical signal-to-noise ratios for urine and serum NTX (Table 4), the effect of variability when using NTX to monitor a difference in bone resorption caused by antiresorptive therapy is similar whether using the urine or serum NTX measure. Previously published reports of changes in urine and serum NTX after the initiation of antiresorptive therapy provide another source of data from which the signal can be computed.(1,12) The percent difference in 6-month mean NTX values in postmenopausal women either treated with HRT plus calcium supplement or calcium supplement alone is nearly identical to our study (–43.4 for urine and −21.7 for serum). Using the long-term variability determined here as the noise, the signal-to-noise ratios derived from these data are 2.8 for urine and 2.9 for serum. As expected, the signal-to-noise ratio for patients treated with a bisphosphonate are greater because of the greater difference in mean NTX between the treated and untreated groups. Data provided by Greenspan et al. generate a signal-to-noise ratio of 3.8 for urine NTX and 4.0 for serum NTX.(20,21) Machado et al. reports a ratio for urinary NTX of 4.9 in a study of osteoporotic women randomized to either 10 mg/day alendronate plus calcium carbonate 500 mg/day elemental calcium or calcium only for 6 months.(10) In this study the ratio for DXA measurement was 1.3 for spine and 0.7 for hip and was 2.2 for osteocalcin.

An alternative approach more commonly used is to compare NTX values before and after initiation of antiresorptive therapy so that the percent difference is reflective of the mean change within a group rather than the difference between two groups. Again using the Chesnut and Greenspan data above, the signal-to-noise ratio for patients treated with HRT is 2.7 for urine and 3.3 for serum and for patients treated with bisphosphonate is 3.7 for urine NTX and 4.1 for serum NTX. The similarity in signal-to-noise ratios from these two methods provides a level of confidence in the analysis used in our data set.

Variation in the time of specimen collection did not play a role in the degree of variability in the urine NTX measure (data not shown). The majority of specimens (61% for short-term and 67% for long-term) were collected between 8 a.m. and noon, with only 4% collected in the afternoon (and the remainder before 8 a.m.). Analysis of those subjects with the highest %CV did not reveal differences in the time of collection.

As with other laboratory measures, both analytical and intrasubject variability should be taken into consideration when using quantitative values for clinical consideration. The least significant change (LSC) incorporates variability into laboratory test interpretation. Using our estimates of variability to compute the LSC, a decrease of 31% in urine NTX and 14% in serum NTX is required to achieve a 90% confidence level that a decrease between two sequential measurements after therapy initiation is clinically relevant and not because of variability alone. After 6 months of HRT, a mean decrease of 42% in urine NTX and 24% in serum NTX has been reported. A slightly larger decrease in NTX has been reported after alendronate therapy, 53% with the urine measurement and 26% in serum. These estimates of LSC are lower than reported by Harmon et al. because that study was over a longer period (6 months) and the estimate was based on a 95% confidence level and considered increases as well as decreases in NTX (i.e., two-tailed).(22)

In conclusion, our results indicate that short-term and long-term intrasubject variability is reduced for the serum NTX marker of bone resorption as compared with the urinary NTX and is comparable with other serum laboratory tests, that is, serum cholesterol. Interestingly, the signal-to-noise ratio is quite similar for both urine and serum NTX markers, and our LSC computations indicate that the levels of decrease observed in response to antiresorptive therapies for urine and serum bone resorption markers are greater than the changes caused by intrasubject variability. Such findings then address the concerns regarding within-subject variability in terms of clinical efficacy of bone markers for individual patients, particularly in terms of monitoring therapy, and reinforce the clinical utility of NTX to monitor response to antiresorptive therapy. We have been able to identify a number of sources of variability in urinary NTX measurements, including HRT use, diet, and clinic site. The variability of serum NTX is lower than for urinary NTX. The clinical utility of serum NTX deserves to be evaluated more extensively.


This study was supported by Ostex International Inc, Seattle, Washington, U.S.A.