Reduced kidney function is associated with BMD, bone loss and markers of mineral homeostasis in older women: a 10-year longitudinal study

The OPRA cohort is a single-center, longitudinal population-based cohort of Caucasian women, randomly selected from city files of Malmö from 1995 to 1999 [11]. Aiming to capture the average community-dwelling older woman, no exclusion criteria were applied. A total of 1604 women received an invitation on their 75th birthday, and 1044 (65%) attended at baseline with 715 and 382 returning for the 5- and 10-year follow-up, respectively. At age 80, reasons for non-attendance included the following: illness (31%); sheltered living (5%); no reason (53%); other reasons (social reasons, walking problems or migration (11%)). At age 85: illness (56%); sheltered living (7%); no reason (23%); other reasons (14%).

The analyses in this study are based only on data from women with eGFRcysC-values, corresponding to 981 at baseline, 685 at age 80 and 365 at age 85. Reasons for missing analytical values include the following: lack of serum, could not provide a blood sample, hemolysis and failed analysis; reflecting a random loss across the cohort and not a systematic error or selection.

The study was approved by the Regional Ethical Review Board in Lund and performed in accordance with the Helsinki declaration. Participants provided written informed consent.

Plasma-cysC from all visits was analyzed in batch in 2015 at the Department of Clinical Chemistry, Malmö, Skåne University Hospital using Cobas auto-analyzer (CV 2.2–1.1%). Estimated glomerular function rate (mL/min/1.73 m²) was calculated using the CKD-EPI_cysC-formula [7]. Comparative analyses were also performed with creatinine-based eGFR calculations using the CG- and MDRD-formulas, as previously described [6].

Using dual-energy X-ray absorptiometry (DXA), measured with a Lunar DPX-L (GE Lunar, Madison, WI), areal BMD (g/cm²) was determined at femoral neck (FN) and total body (TB) at Skåne University Hospital Malmö. Precision of DXA was assessed by duplicate measurements on healthy individuals. Precision error was 0.009–0.010 g/cm² at FN and 0.011–0.030 g/cm² at TB. No drifts in phantom-measured results were observed [12]. Osteoporosis was defined as a T-score ≤ − 2.5 at the FN and osteopenia as a T-score between − 1 and − 2.5. Rate of bone loss was calculated individually for every woman as annual percentage change in BMD or as absolute bone loss over 5 years (between 75 and 80, n = 685; between 80 and 85, n = 365) and 10 years (between 75 and 85, n = 365).

Non-fasting blood samples were collected between 8 a.m. and 1 p.m., centrifuged and stored at − 80 °C. Plasma calcium and alkaline phosphate (ALP) were analyzed using a Beckman synchron LX20–4 auto-analyzer or Cobas auto-analyzer (Roche). At age 75, serum levels of PTH were analyzed using Elecsys PTH immunoassay (Roche Diagnostics, Mannheim, Germany). The normal range spanned from 1.6 to 6.9 pmol/L with intra- and inter-assay CVs of 1.6 and 5.7%, respectively. At the 5- and 10-year follow-up, PTH was analyzed using Cobas auto-analyzer by Roche (intra-assay CV, 7–5%). To ensure inter-assay comparability between methods, consistency of PTH values was assessed by duplicate measurements [13]. Plasma phosphate from all visits was analyzed in batch according to routine methods. Serum 25(OH)D (vitamin D) was analyzed using liquid chromatography mass spectrophotometry (LC-MS) (inter-assay CV 3–6%). CRP was analyzed using Roche Diagnostics (Cobas) with a CV of 3.6–4.1%. All assays were performed according to routine methods. Bone turnover markers S-TRACP5b, serum bone-specific alkaline phosphatase (s-BALP) and total osteocalcin were analyzed as previously described [14].

Using standardized methods, weight (kg) and height (cm) were measured and BMI (kg/m²) calculated. Lifestyle-related information was available from questionnaires.

Analyses were performed using just the available data (Online resource 1a). Descriptive data is presented as mean with standard deviation or median with interquartile range as appropriate. To identify association between kidney function (eGFR) and BMD, multivariate linear regression analyses with eGFR (independent) and BMD (dependent) as continuous variables was performed, adjusted for weight (model 1) and weight, smoking and vitamin D (model 2). In each model, we present partial explained variation (unadjusted R ²) for eGFR. The analyses were also performed after excluding steroid (n = 29 at age 75; n = 41 at age 80; n = 22 at age 85) and bisphosphonate (n = 33 at age 75; n = 58 at age 80; n = 55 at age 85) users. A previous report from the OPRA cohort indicated that having ever used a potent estrogen (n = 49, 5%) has no effect on BMD at the age of 75 [15], and therefore, current users were not excluded from the analyses.

Participants were staged into categories of kidney function based on their eGFR alone as follows: (1) normal-mild reduction of kidney function, eGFR ≥ 60 mL/min/1.73 m² (CKD stages 1–2, “normal” kidney function), (2) mild-moderate reduction, eGFR from 45 to 59 mL/min/1.73 m² (CKD stage 3A, “intermediate” function) and (3) moderate-severe reduction, eGFR < 45 mL/min/1.73 m² (CKD stages 3B–5, “poor” function) [4].

Power analyses for bone density were performed prior to baseline inclusion; based on the supposition of 0.13 g/cm² SD in BMD, this study has over 80% power to detect a 0.056 g/cm² difference between equal groups at a 5% significance level. Differences in bone loss and characteristics according to kidney function were calculated using ANOVA (for normally distributed data) or Kruskal-Wallis (for non-normally distributed data). Differences between the women who maintained a normal versus poor kidney function throughout follow-up were compared using independent sample t test or Mann-Whitney U-test.

All data was analyzed with SPSS (IBM SPSS v22.0; Armonk, NY: IBM Corp.). A p value < 0.05 was considered nominally significant. When difference between three groups is compared, significance is presented as p values for trend.

Kidney function (eGFR by CKD-EPI_cysC) was positively associated with FN BMD at both 75 and 80 years old in the fully adjusted models (\( \widehat{\beta} \) = 0.001, p = 0.028 and \( \widehat{\beta} \) = 0.001, p = 0.001, respectively), but not in advanced old age (85 years). Overall, the contribution was small; however, the partial explained variation, 0.4% at 75 years old and 1.6% at age 80 years.

Excluding women on bisphosphonates did not substantially alter the results, but with exclusion of steroid users, the association at age 75 was lost. Notably, kidney function and TB BMD were inversely associated at all three ages, although abolished with weight adjustment (Table 2). Comparative analyses using the creatinine-based formulas were essentially the same (Online resource 2).

To compare differences in bone loss and markers of bone metabolism, women were stratified into three categories (normal, intermediate, poor) based on their kidney function. To evaluate age-related associations, bone loss was compared during two different 5-year periods (between ages 75–80 and 80–85). Individually calculated annual percentage bone loss differed between the three categories ((FN p = 0.001, Fig. 1); TB p = 0.014); a higher rate was apparent in women with intermediate (stage 3A) and poor (stages 3B–5) kidney function in the first time period, at both measurement sites. In the second period, bone loss did not differ at the FN (p = 0.264). However, a difference between the three categories was observed at TB (p = 0.044), and the lowest bone loss was seen in women with intermediate kidney function. During the first 5-year time period, analyzing absolute bone loss (g/cm²/5 years), the results were similar (data not shown).

Table 3 shows characteristics of women stratified by kidney function at each age (75, 80 and 85). Although women with reduced kidney function (i.e. intermediate and poor) tended to have higher TB BMD (p < 0.05), there was no difference in FN BMD or prevalence of osteopenia and osteoporosis at any age.

Markers of mineral homeostasis also differed by kidney function; at all ages, PTH was higher as kidney function decreased (p < 0.001), although mean values were well within normal range. However, phosphate and ALP only differed at age 75 (p < 0.001), although women with poor function developed higher calcium levels. Furthermore, at all ages, women with reduced kidney function had higher values for weight, the inflammation marker CRP and the bone formation marker osteocalcin (p < 0.001 for all). Other bone markers, BALP and TRACP5, did not differ with kidney function.

To understand the skeletal implications of kidney function over time, we compared women who had, and maintained, poor kidney function (stages 3b–5, n = 26) throughout the 10-year observation period to those with normal function throughout (stages 1–2, n = 52). Women who survived despite poor function had higher annual percentage bone loss at the FN over the 10 years, although variability precluded statistical significance (p = 0.097) (Fig. 2). Using absolute bone loss values, the results were largely similar.

Continuous poor kidney function was associated with a higher TB BMD (p < 0.05 all time points). FN BMD was higher, and correspondingly, prevalence of osteoporosis lower, although only significant at baseline (p = 0.039 and p = 0.017, respectively; Table 4). In addition, the women with continuous poor function had higher PTH (increasing even above normal range at age 85) and were 10 kg heavier (p < 0.05 and p < 0.001, respectively, for all time points). Differences in calcium, phosphate, CRP, BALP and osteocalcin were also observed.

To put our results into context and demonstrate the current knowledge base for kidney function and BMD or bone loss in elderly women, the findings from the literature review are summarized in Online resource 3 (in all 20 studies). A coherent picture of kidney function and BMD/mineral homeostasis is hampered by heterogeneity in study populations (wide age spans, ethnicity and number of participants). Only ten studies comprised more than 200 participants, and only four represent Caucasian women. Seven studies have been performed in patients with defined CKD, with a mean age considerably below 75 years. Furthermore, only three studies were longitudinal, and none of these had > 5 years of follow-up. Most studies used the creatinine-based MDRD or CG-formulas to estimate GFR, making comparisons difficult [16‐35].

Firstly, we acknowledge that participants might be healthier than women who declined to participate [42], although the random population-based sampling with a baseline attendance rate of 67%, which is comparatively very high, to an extent, reduces this. By the same token it might attract participants with an interest in osteoporosis and fracture. This limitation also applies to follow-up as the number is reduced from either mortality, indications of declining health or for various other reasons. In this study, this is mirrored in the eGFR values differing between attendees and non-attendees although, as previously reported not significantly between ages 75 and 80 (p = 0.123) but between ages 80 and 85 (p = 0.047) with eGFR as a predictor of mortality [6]. This is an inherent problem in all longitudinal studies of the elderly, which we have addressed by presenting both cross-sectional and longitudinal data. In addition, in an attempt to control for this, we have specifically evaluated those surviving, comparing those with good and poor function. Albeit, data needs to be interpreted with caution due to smaller numbers at the final follow-up, few other studies of same-aged women have been able to provide a similar analysis.

Secondly, the definition of normal kidney function in the “normal” older person is not without reservation since CKD staging may not fully apply. Reduced kidney function is a common state in the elderly, and many individuals in stage 3a without proteinuria might never reach stages 3b–5. We lack direct measurement, and data on urine protein would have been an advantage in this cohort. It is, however, plausible that cysC estimates of kidney function are more accurate compared to the creatinine-based equations used in most previous studies of older people.

In summary, kidney function is positively associated with FN BMD in elderly women, although the association attenuates as aging progresses. A reduction of kidney function is furthermore associated with bone loss and affected mineral homeostasis. We therefore conclude that reduced kidney function is harmful to bone mass in older women.