A 2-yr Randomized Controlled Trial on Creatine Supplementation during Exercise for Postmenopausal Bone Health

ABSTRACT

Purpose

Our purpose was to examine the effects of 2 yr of creatine monohydrate supplementation and exercise on bone health in postmenopausal women.

Methods

Two hundred and thirty-seven postmenopausal women (mean age, 59 yr) were randomized to receive creatine (0.14 g·kg⁻¹·d⁻¹) or placebo during a resistance training (3 d·wk⁻¹) and walking (6 d·wk⁻¹) program for 2 yr. Our primary outcome was the femoral neck bone mineral density (BMD), with lumbar spine BMD and proximal femur geometric properties as the secondary outcomes.

Results

Compared with placebo, creatine supplementation had no effect on BMD of the femoral neck (creatine: 0.725 ± 0.110 to 0.712 ± 0.100 g·cm⁻²; placebo: 0.721 ± 0.102 to 0.706 ± 0.097 g·cm⁻²), total hip (creatine: 0.879 ± 0.118 to 0.872 ± 0.114 g·cm⁻²; placebo: 0.881 ± 0.111 to 0.873 ± 0.109 g·cm⁻²), or lumbar spine (creatine: 0.932 ± 0.133 to 0.925 ± 0.131 g·cm⁻²; placebo: 0.923 ± 0.145 to 0.915 ± 0.143 g·cm⁻²). Creatine significantly maintained section modulus (1.35 ± 0.29 to 1.34 ± 0.26 vs 1.34 ± 0.25 to 1.28 ± 0.23 cm³ (placebo), P = 0.0011), predictive of bone bending strength, and buckling ratio (10.8 ± 2.6 to 11.1 ± 2.2 vs 11.0 ± 2.6 to 11.6 ± 2.7 (placebo), P = 0.011), predictive of reduced cortical bending under compressive loads, at the narrow part of the femoral neck. Creatine reduced walking time over 80 m (48.6 ± 5.6 to 47.1 ± 5.4 vs 48.3 ± 4.5 to 48.2 ± 4.9 s (placebo), P = 0.0008) but had no effect on muscular strength (i.e., one-repetition maximum) during bench press (32.1 ± 12.7 to 42.6 ± 14.1 vs 30.6 ± 10.9 to 41.4 ± 14 kg (placebo)) and hack squat (57.6 ± 21.6 to 84.4 ± 28.1 vs 56.6 ± 24.0 to 82.7 ± 25.0 kg (placebo)). In the subanalysis of valid completers, creatine increased lean tissue mass compared with placebo (40.8 ± 5.7 to 43.1 ± 5.9 vs 40.4 ± 5.3 to 42.0 ± 5.2 kg (placebo), P = 0.046).

Conclusions

Two years of creatine supplementation and exercise in postmenopausal women had no effect on BMD; yet, it improved some bone geometric properties at the proximal femur.

Low bone mineral density (BMD), leading to osteoporosis, places a high burden on health care systems (1). Although exercise programs such as resistance training and walking have been shown effective for improving BMD (2,3), the short-term effects are small and may not be clinically relevant.

Creatine is a nitrogen-containing compound synthesized in the kidneys and liver or consumed in the diet primarily from red meat and seafood (4). When creatine combines with phosphate (as phosphocreatine), it provides a source of energy for cells, including bone cells (5). The addition of creatine to low serum cell culture medium increases the metabolic activity and differentiation of osteoblasts, the cells involved in bone formation (5). Creatine supplementation also reduces bone resorption (i.e., bone catabolism) in young boys with muscular dystrophy (6,7) and in older men during a supervised resistance training program (8).

We showed in a small preliminary study that creatine supplementation combined with supervised resistance training increased femoral neck BMD in postmenopausal women, with the increase in BMD reaching levels that approached those thought to clinically reduce fracture risk (9). The femoral neck is recognized as the most clinically relevant site, as there is significant trauma when it is fractured (10). A 5% increase in BMD is predicted to reduce fracture risk by 25% (11). Our previous study indicated an approximate 3.4% preservation of BMD at the femoral neck in a small group (n = 33) of postmenopausal women over 1 yr of creatine supplementation during a supervised resistance training program (9); therefore, we hypothesized that a longer intervention (i.e., 2 yr) may be necessary to achieve a clinically significant increase in BMD.

In addition to BMD, geometric properties of bone also affect bone strength. Geometric properties around the proximal femur, assessed using hip structural analysis, are good predictors for fracture (12,13). We recently showed that femoral neck section modulus, a predictor of bone bending strength, tended to increase (P = 0.061) in older men who supplemented with creatine during 1 yr of supervised resistance training (14). We also demonstrated that geometric properties around other bone sites (i.e., cross-sectional area (CSA) of the tibia) were improved with creatine supplementation and supervised resistance training programs in older men and postmenopausal women (15).

Our primary purpose was to determine whether creatine supplementation during a longer-term supervised resistance training and walking program (i.e., 2 yr) could improve BMD at the femoral neck in postmenopausal women. A secondary purpose was to determine the effects of creatine on BMD at the lumbar spine and on geometric properties at the proximal femur. We hypothesized that creatine supplementation would be beneficial for improving BMD and bone geometric properties.

METHODS

Study design

This two-site randomized controlled trial was approved by the Biomedical Research Ethics Boards at the University of Saskatchewan and University of Regina. Participants were informed of the risks and purposes of the study before written consent was obtained. The study complied with the World Medical Association Declaration of Helsinki—Ethical Principles for Medical Research Involving Human Subjects. The trial was registered with ClinicalTrials.gov (NCT02047864).

The study used a randomized, double-blind, placebo-controlled, parallel-group, repeated measures design at two centers: the University of Saskatchewan and University of Regina. Participants were randomized using a computer-generated allocation schedule to one of two groups (on a 1/1 ratio) for a 2-yr intervention: group 1—creatine monohydrate supplementation (n = 120, 0.14 g·kg⁻¹·d⁻¹ mixed with 0.14 g·kg⁻¹·d⁻¹ maltodextrin) and supervised resistance training plus partially supervised walking; or group 2—placebo (control; n = 117, 0.28 g·kg⁻¹·d⁻¹ maltodextrin) and supervised resistance training plus partially supervised walking.

Creatine supplementation

Participants were provided with the supplement in plastic Ziploc bags and instructed to mix supplements with water. The creatine monohydrate supplement (Creapure; AlzChem AG, Trosberg, Germany) was verified for purity (Cary Co., Addison, IL) and determined to be 99.9% pure with minimal contaminants (i.e., 44 mg·kg⁻¹ creatinine, 28 mg·kg⁻¹ dicyandiamide, and below the limit of detection for dihydrotriazine). On the days when participants performed resistance training, half of the supplement was consumed postexercise (approximately 5 min) under research assistant supervision, and the other half was consumed with a meal. On rest days when participants did not visit the resistance training facility, they consumed one-third of the supplement with each meal. Compliance was monitored by completing a supplement tracking log. The dose we chose (0.14 g·kg⁻¹·d⁻¹) was slightly higher than that used in our previous study (0.10 g·kg⁻¹·d⁻¹) where we observed an increase in femoral neck BMD in a small group of postmenopausal women (9) because this smaller dose was ineffective for increasing lean tissue mass or muscular strength, indicating a possible anabolic resistance in postmenopausal women.

All participants received a supplement of 500 mg of calcium and 10 μg (400 IU) of vitamin D per day to ensure that most participants were meeting Osteoporosis Society of Canada recommendations of 1200 mg·d⁻¹ for calcium and 20 μg·d⁻¹ for vitamin D (10).

A research assistant at the University of Saskatchewan with no other role in the study performed a stratified block randomization using a computer-generated allocation schedule. Randomization was performed using a fixed block size of eight (using a permuted block design with a computer random number generator). Participants were stratified as either 1–9 yr after menopause or >9 yr after menopause to account for differing rates of bone loss after menopause (16).

The allocation sequence was concealed from other research assistants enrolling and assessing participants. Participants were given their allocated supplement after completing all baseline assessments. Participants, researchers, and those involved in outcome assessment or resistance training supervision were blinded to the group assignment. Another research assistant oversaw all data entry during which they remained blinded to group allocations by coding of groups. Statistical analyses were performed blinded (by coding of groups).

Before codes were revealed at the end of the study, participants were surveyed to assess the effectiveness of blinding by asking if they thought they were on creatine or placebo.

Exercise training

Resistance exercises and brisk walking were performed in-lab and supervised 3 d·wk⁻¹. An additional 3 d of nonsupervised brisk walking was performed outside of the laboratory.

Exercises performed during resistance training included hack squat, hip abduction, adduction, flexion, and extension using a multihip machine, bench press, lat pull-down, shoulder press, hamstrings curl, quadriceps (knee) extension, biceps curl, triceps extension (presses), and back extension. Participants initially took part in a 2-wk familiarization process (two supervised sessions per week, 2 sets of 10 repetitions per exercise) where light resistance was used and proper form was emphasized. After the familiarization phase, participants took part in strength assessments and then enrolled in the resistance training program that involved two sets of eight repetitions for each exercise to muscular fatigue with 2-min rest between sets for each exercise at an intensity corresponding to a load equal to or >80% one-repetition maximum (1RM) (i.e., 80% of their maximal strength, for the bench press and hack squat) or the maximal amount of weight that could be lifted eight times (i.e., 8RM). Resistance was progressively increased by 2–5 kg once a participant could complete two sets of eight repetitions for an exercise. All exercises were performed on plate-loaded or weight stack machines (Lever Fitness or Hammer Strength, Winnipeg, Manitoba, Canada).

The walking program involved 20–30 min of brisk walking per session at an intensity corresponding to 70% of age-predicted maximum heart rate (220 − age). This duration was chosen as it would ensure participants were meeting the recommendations set by the Canadian Physical Activity Guidelines for Older Adults of approximately 150 min·wk⁻¹ of moderate- to vigorous-intensity activity (17). Participants were educated how to take a 15-s radial pulse count to ensure they were exercising at the proper intensity. Compliance was monitored by recording and tracking exercise in training logs.

Participants

Participants were recruited from January 2014 to July 2016, and all had completed the intervention by January 2019. The inclusion criteria were as follows: 1) postmenopausal women (no menstrual period in the past 2 yr, determined by questionnaire—if a participant reported being <2 yr postmenopausal, menopausal status was confirmed with blood testing for luteinizing hormone and follicle-stimulating hormone) and 2) considered at “low” and “moderate” risk of fracture according to the Osteoporosis Canada guidelines (10). Exclusion criteria were as follows: 1) preexisting kidney or liver abnormalities (determined by blood and urine tests), 2) high risk of fragility fracture according to Osteoporosis Canada guidelines (10), 3) taking bisphosphonates, parathyroid hormone, calcitonin, hormone replacement therapy, selective estrogen receptor modulators, or creatine monohydrate in the past 12 months, 4) currently taking corticosteroids, or any medication that affects bone mineral metabolism, 5) severe osteoarthritis, Crohn’s disease, or Cushing disease, 6) planning to travel during the study for greater than a 2-wk duration at one time, and 7) currently involved in resistance training (>20 min per session, more than twice per week). Participants were recruited via newspaper and e-mail advertisements and posters from the cities of Saskatoon and Regina, in the province of Saskatchewan, Canada.

Outcome assessments

All outcomes were assessed at baseline, at the end of year 1, and at the end of year 2. Falls and fractures were recorded for an additional year after completion of the study (i.e., 3 yr).

Height and mass were measured by a standard stadiometer and a calibrated electronic scale.

Measurements of outcomes from dual-energy X-ray absorptiometry are described in detail elsewhere (18). Briefly, BMD of the proximal femur (i.e., femoral neck, trochanter, Ward’s area, and total hip), lumbar spine (L1–L4 vertebrae), and total body, and whole-body lean tissue mass (excluding bone) were measured by dual-energy X-ray absorptiometry in array mode (QDR Discovery Wi; Hologic, Inc., Bedford, MD) using QDR software for Windows XP (QDR Discovery). DXA scanners were identical at both university centers.

Hip structural analysis (19) was used to assess geometric characteristics of the proximal femur using DXA at the narrowest part of the femoral neck, the intertrochanteric region, and the femoral shaft (i.e., 2 cm distal to the midpoint of the lesser trochanter). The following were assessed at each region: cortical thickness, subperiosteal width (SPW), bone CSA, which is equivalent to the cortical area, cross-sectional moment of inertia (CSMI), section modulus (Z), and buckling ratio. Higher CSMI and Z are associated with strength of bone in bending, and higher buckling ratio is associated with susceptibility of cortical bending under compressive loads (19).

Ultrasound (Sunlight, Omnisense, 7000S; BeamMed Ltd., Petah Tikva, Israel) was used at the distal radius and tibial shaft to assess bone speed of sound. A higher bone speed of sound predicts enhanced architectural arrangement of bone (20). This assessment was only conducted at the University of Saskatchewan site (n = 160), as it was the only site with the required equipment.

Strength in the lower and upper body was assessed by determining 1RM during hack squat and bench press, respectively, with methods as previously described (21).

Dynamic balance was measured as the time taken to perform backward tandem walking (i.e., toe to heel) over a distance of 6 m on a 10-cm-wide board that was raised 4 cm off the ground (18).

Walking speed was assessed over an 80-m course, with the instructions to walk at a fast pace (22).

Uncontrolled intervention factors

Dietary intake was assessed using a Food Frequency Questionnaire (Block 98.2 FFQ; Block Dietary Data Systems, Berkeley, CA). After the questionnaires were filled out, they were then sent out for analysis (www.nutritionquest.com; Nutrition Quest, Berkeley, CA).

Physical activity outside of study requirements was assessed using a Leisure-Time Exercise Questionnaire (23).

Fasting blood and urine analyses and 24-h urine analysis were completed at baseline, year 1, and year 2 to measure markers of kidney and liver function (e.g., creatinine clearance, urea, albumin, bilirubin, aspartate aminotransferase, alanine aminotransferase, alkaline phosphatase, urine specific gravity, protein, and microalbumin) and complete blood counts. Blood was centrifuged at 3200 rpm for 10 min at 20°C. All blood and urine analyses were carried out on a fully automated random access analyzer for clinical chemistry (Cobas C 311; Roche Diagnostic, Mannhein, Germany) at the Royal University Hospital (Saskatoon) and at the Regina General Hospital. Resultant values were reviewed by our physician coinvestigators. Any values outside the normal reference range were considered an adverse event.

Adverse events

Adverse events were collected on adverse event forms each time participants were in contact with the researchers. These included a description of the adverse event, relationship to the intervention (not related, unlikely, possibly, probably, or definite), whether it was serious (i.e., resulted in death, life-threatening, required hospitalization, or resulted in persistent disability) or nonserious, and its intensity (mild, moderate, severe, or life-threatening). We paid special attention to adverse events that might be a concern with creatine supplementation, such as kidney or liver adverse events, gastrointestinal problems, and muscle cramping. As a conservative approach to the safety of participants, we had physicians review adverse events, and they made decisions on whether to reduce the creatine/placebo dose based on potential kidney or liver-related adverse events (i.e., measurement of blood and/or urine parameters that might indicate liver/kidney problems).

Statistics

The sample size calculation was based on a clinically significant difference of 5% in BMD at the femoral neck, which was considered our primary outcome variable. This difference would be expected to reduce fracture risk by 25% (11). Based on the results of our previous 1-yr study (9), we expected that a 2-yr intervention could elicit differences between creatine and placebo groups at a clinically significant level of 5%. Baseline femoral neck BMD in postmenopausal women with similar exclusion criteria is about 0.739 g·cm⁻² with a standard deviation of 0.093 g·cm⁻² (18,24). A clinically significant difference of 5% would result in a BMD of 0.776 g·cm⁻². Using these predicted values for placebo and creatine groups (i.e., 0.739 and 0.776 with an SD of 0.093 g·cm⁻²), an alpha of 0.05, and a power of 80%, required a sample size of 101 per group (i.e., 202 total). In a previous 2-yr exercise training study of postmenopausal women from our laboratory, 15% were lost to follow-up (18); therefore, we required approximately 120 per group or 240 postmenopausal women for this multisite randomized controlled trial.

Reproducibility of all dependent variables was determined by repeated measures 3–7 d apart on 18 participants and expressed as percentage coefficient of variation (25), intraclass correlation coefficient (26), standard error of the mean (26), and minimal difference to be real (27). The percentage of participants from each group with a beneficial or detrimental change that exceeded the minimal difference to be real is presented in each results table.

All data were checked for outliers by first plotting variables at each time point in a scatter plot. Outlier data points were then checked to see if they were greater than 2.5× the median absolute difference (28), and any outliers were replaced by imputed values (see hereafter, this accounted for <0.2% of all data points). Baseline variables between groups were compared using independent samples t-tests. Our primary analysis was on an intent-to-treat basis; that is, all participants who were randomized were included in their original groups for analysis irrespective of their compliance with the exercise program or supplement. Separate analyses were conducted including only participants who completed the entire intervention (i.e., valid-completers analysis) and for participants who were deemed to have been adequately compliant with the intervention. Adequate compliance was chosen as consuming 50% of the creatine supplement and attending 60% of the resistance training sessions. These cut-offs were chosen because this level of creatine supplementation (i.e., 50% of 0.14 g·kg⁻¹·d⁻¹) would approximate the daily amount that is usually recommended (i.e., 5 g·d⁻¹), and the number of resistance training sessions (approximately 2 d·wk⁻¹) is the amount recommended by physical activity guidelines for older adults (17). Any missing data were assumed to be missing at random. This included data missing because of participants who were lost to follow-up (about 13% of all data points), outlier exclusion (<1% of all data points), poor positioning during hip scans (for hip structural analyses) (<1% of all data points), participants who missed testing appointments (1% of all data points), and participants who decided not to take part in fitness testing because of joint soreness (2% of data points for fitness testing variables). Multiple imputation was used to impute missing data for the intent-to-treat analyses (SPSS version 29, Chicago, IL). In order to protect against type I error, outcomes were grouped into “bone density” outcomes, “bone geometric” outcomes, and “fitness testing” outcomes, with each of these groups of variables analyzed by a 2 (group; i.e., creatine vs placebo) × 3 (time, i.e., baseline, year 1, year 2) multivariate analysis of variance (MANOVA), with repeated measures on the “time” factor. If multivariate tests were significant, univariate tests, with Bonferroni post hoc tests (to correct for multiple comparisons), were used to determine differences between groups across time points. Lean tissue mass, bone ultrasound outcomes, dietary variables, and physical activity levels outside of the program were assessed with a 2 (group) × 3 (time) ANOVA with Bonferroni post hoc tests to determine differences between pairs of means. Compliance to the intervention and number of falls and fractures over 3 yr were assessed with a between-groups ANOVA. Proportion of participants who were able to correctly guess which group they were in, differences in adverse events between groups, and differences between groups for percentage of participants who exceeded the minimal differences to be real were assessed by chi-square analysis. All results are presented as mean ± SD. All MANOVAs or ANOVAs were run with Statistica 7.0 (Statsoft, Chicago, IL). Significance was accepted at P ≤ 0.05.

RESULTS

At baseline, the creatine group (n = 120) had an age (mean ± SD) of 59.0 ± 5.6 yr, body mass of 73.3 ± 14.9 kg, and height of 164 ± 6 cm with corresponding values for the placebo group (n = 117) of 59.0 ± 5.7 yr, 71.8 ± 14.2 kg, and 163 ± 6 cm, respectively. All other baseline variables are presented in Tables 1–4. There were no differences between groups for any variables at baseline. Reproducibility of variables is presented in Supplemental Table 1 (see Supplemental Digital Content, http://links.lww.com/MSS/C857). Compliance (defined as the amount of supplement consumed relative to the amount prescribed) with the supplement was 56% for the creatine group and 60% for the placebo group (P > 0.05). Compliance with the calcium and vitamin D supplement was 70% and 68% for the creatine and placebo groups, respectively (P > 0.05). Both groups had 61% compliance (defined as the number of exercise sessions completed relative to the number prescribed) with the resistance training program, and compliance for the walking program was 65% and 60% for the creatine and placebo groups, respectively (P > 0.05).

TABLE 1.

BMD, bone speed of sound, and lean tissue mass at baseline, year 1, and year 2 for creatine and placebo groups.

	Creatine (n = 120)			Placebo (n = 117)			Time	Group × Time	% of Participants Exceeding Minimal Difference to Be Real
	Baseline	Year 1	Year 2	Baseline	Year 1	Year 2			Creatine	Placebo	Creatine	Placebo
									Increase	Increase	Decrease	Decrease
Femoral neck BMD (g·cm−2)	0.725 ± 0.110	0.717 ± 0.106	0.712 ± 0.100	0.721 ± 0.102	0.713 ± 0.098	0.706 ± 0.097	P < 0.0001	P = 0.84	23%	19%	57%	58%
Total hip BMD (g·cm−2)	0.879 ± 0.118	0.877 ± 0.112	0.872 ± 0.114	0.881 ± 0.111	0.880 ± 0.112	0.873 ± 0.109	P < 0.0001	P = 0.69	36%	32%	53%	54%
Trochanter BMD (g·cm−2)	0.658 ± 0.099	0.657 ± 0.096	0.656 ± 0.093	0.665 ± 0.092	0.662 ± 0.096	0.664 ± 0.095	P = 0.42	P = 0.63	36%	41%	51%	40%
Wards BMD (g·cm−2)	0.548 ± 0.131	0.544 ± 0.132	0.528 ± 0.116	0.542 ± 0.109	0.545 ± 0.118	0.519 ± 0.114	P < 0.0001	P = 0.40	18%	14%	58%	66%
Lumbar spine BMD (g·cm−2)	0.932 ± 0.133	0.931 ± 0.135	0.925 ± 0.131	0.923 ± 0.145	0.917 ± 0.145	0.915 ± 0.143	P = 0.003	P = 0.44	28%	32%	40%	47%
Total body BMD (g·cm−2)	1.087 ± 0.095	1.081 ± 0.096	1.076 ± 0.094	1.066 ± 0.117	1.063 ± 0.119	1.054 ± 0.119	P < 0.0001	P = 0.31	27%	23%	60%	65%
Distal radius speed of sound (m·s−1)a	4066 ± 116	4036 ± 102*	4064 ± 96	4051 ± 122	4058 ± 126	4038 ± 105	P = 0.16	P = 0.002	11%	9%	12%	19%
Tibia speed of sound (m·s−1)a	3856 ± 105	3871 ± 107	3844 ± 116	3857 ± 105	3863 ± 118	3843 ± 114	P = 0.0023	P = 0.79	17%	23%	26%	35%
Lean tissue mass (kg)	41.4 ± 6.2	43.3 ± 6.3	43.5 ± 6.3	40.3 ± 5.3	41.9 ± 5.2	42.1 ± 5.4	P < 0.0001	P = 0.11	88%	79%	8%	14%

All values are means ± SD.

^an = 81 creatine and 79 placebo group (data were only collected at the University of Saskatchewan).

*Year 1 < baseline and year 2 for the creatine group (Bonferroni post hoc, P < 0.05).

TABLE 4.

Physical activity score and dietary intake at baseline, year 1, and year 2 for creatine and placebo groups.

	Creatine			Placebo			Time	Group × Time
	Baseline	Year 1	Year 2	Baseline	Year 1	Year 2
Leisure physical activity score (arbitrary units)	27 ± 23	24 ± 23	27 ± 23	25 ± 22	26 ± 22	24 ± 22	P = 0.85	P = 0.49
Total energy intake (kcal·d−1)	1662 ± 511	1579 ± 525	1584 ± 562	1550 ± 515	1450 ± 530	1498 ± 562	P = 0.06	P = 0.86
Calcium intake (mg·d−1)a	805 ± 370	763 ± 348	756 ± 333	778 ± 374	711 ± 351	665 ± 335	P = 0.001	P = 0.34
Vitamin D intake (μg·d−1)a	148 ± 111	147 ± 133	149 ± 104	149 ± 117	150 ± 133	121 ± 101	P = 0.18	P = 0.12
Protein (g·d−1)	67 ± 22	63 ± 22	64 ± 22	64 ± 23	62 ± 23	60 ± 23	P = 0.067	P = 0.63

All values are means ± SD.

^aValues only include nutrients from dietary intake and do not include the supplements given during the study.

Participant flow through the study is presented in Figure 1. Loss to follow-up was similar between groups. Reasons for loss to follow-up are presented in the figure, with lack of time as the most common reason for withdrawing from the study. At the end of the study, 50% and 43% of participants in the creatine and placebo groups, respectively, were able to correctly guess which group they were in (P > 0.05), with the remaining participants either guessing the incorrect group or stating they were unsure.

FIGURE 1.

CONSORT flow diagram showing participant flow through the study.

BMD, Bone Ultrasound, and Lean Tissue Mass

No group – time interaction for the MANOVA assessing BMD outcomes in the intent-to-treat analysis was evident (P = 0.59, Table 1). Likewise, there was no group – time interaction for the valid-completers analysis (see Supplemental Table 2, Supplemental Digital Content, BMD, bone speed of sound, and lean tissue mass at baseline, year 1, and year 2 for creatine and placebo groups (valid completers), http://links.lww.com/MSS/C857) or the compliers analysis (see Supplemental Table 5, Supplemental Digital Content, BMD, bone speed of sound, and lean tissue mass at baseline, year 1, and year 2 for creatine and placebo groups (compliers), http://links.lww.com/MSS/C857). In the compliers analysis, there was a greater percentage of participants in the creatine compared with the placebo group who exceeded the minimal difference to be real for increases in total hip BMD (P = 0.0071), but there was also a greater percentage of participants who exceeded the minimal difference to be real for decreases in Wards BMD (P = 0.01) and lumbar spine BMD (P = 0.0007). There was also a greater percentage of participants in the placebo compared with the creatine group who exceeded the minimal difference to be real for increases in lumbar spine BMD (P = 0.012) (see Supplemental Table 5, Supplemental Digital Content, http://links.lww.com/MSS/C857).

A significant group – time interaction was found for distal radius speed of sound (P = 0.002) in the intent-to-treat analysis (Table 1) and the valid-completers analysis (see Supplemental Table 2, Supplemental Digital Content, BMD, bone speed of sound, and lean tissue mass at baseline, year 1, and year 2 for creatine and placebo groups (valid completers), http://links.lww.com/MSS/C857). The Bonferroni post hoc analysis indicated that the creatine group distal radius speed of sound decreased from baseline to year 1 (P < 0.05) but increased from year 1 to year 2 (P < 0.05). There were no differences between groups for percentage of participants who exceeded the minimal difference to be real for any of the bone ultrasound measures.

A significant group – time interaction was evident for lean tissue mass (P = 0.046) in the valid-completers analysis, with a greater increase in the creatine group (see Supplemental Table 2, Supplemental Digital Content, http://links.lww.com/MSS/C857). This was not significant in the intent-to-treat or compliers analyses (Table 1 and Supplemental Table 5, Supplemental Digital Content, http://links.lww.com/MSS/C857). The Bonferroni post hoc analysis from the valid completers indicated that both groups increased from baseline to year 1 and from baseline to year 2 (both P < 0.0001). For the valid-completers analysis, there was a lower percentage of participants from the creatine compared with the placebo group who exceeded the minimal difference to be real for decreases in lean tissue mass (P = 0.046; Supplemental Table 2, Supplemental Digital Content, http://links.lww.com/MSS/C857).

Bone Geometric Outcomes from the Proximal Femur

Geometric outcomes from hip structural analyses (i.e., the intent-to-treat analyses) are presented in Table 2. There was a significant group – time interaction from the MANOVA from the intent-to-treat analysis (P = 0.00003). Univariate tests indicated significant group – time interactions for section modulus (P = 0.0011) and buckling ratio (P = 0.001) at the narrow part of the femoral neck and cortical thickness (P = 0.017), SPW (P = 0.044), section modulus (P = 0.03), and buckling ratio (P = 0.045) at the femoral shaft. Bonferroni post hoc analyses indicated that in the placebo group, the narrow neck section modulus decreased from baseline to year 2 (P < 0.0001), and from year 1 to year 2 (P < 0.001), the narrow neck buckling ratio increased from baseline to year 2 (P < 0.05), the femoral shaft cortical thickness decreased from baseline to year 2 (P < 0.05), and the femoral shaft section modulus decreased from year 1 to year 2 (P < 0.05). Bonferroni post hoc analyses indicated that in the creatine group, femoral shaft SPW decreased from baseline to year 2 (P < 0.01), and femoral shaft buckling ratio decreased from baseline to year 1 (P < 0.05). There was a greater percentage of participants from the creatine compared with the placebo group who exceeded the minimal difference to be real for increases in CSA (P = 0.0013), CSMI (P = 0.0019), and section modulus (P < 0.0001) at the narrow part of the femoral neck, and cortical thickness (P = 0.0093) and section modulus (P = 0.022) at the femoral shaft. There was a greater percentage of participants from the placebo compared with the creatine group who exceeded the minimal difference to be real for decreases in section modulus of the narrow part of the femoral neck (P = 0.045) and the femoral shaft (P = 0.039), and for increases in SPW (P = 0.001) at the femoral shaft (Table 2).

TABLE 2.

Hip structural analysis measures at baseline, year 1, and year 2 for creatine and placebo groups.

	Creatine (n = 120)			Placebo (n = 117)			Time	Group × Time	% of Participants Exceeding Minimal Difference to Be Real
	Baseline	Year 1	Year 2	Baseline	Year 1	Year 2			Creatine	Placebo	Creatine	Placebo
									Increase	Increase	Decrease	Decrease
Narrow neck
Cortical thickness (cm)	0.17 ± 0.03	0.17 ± 0.03	0.17 ± 0.03	0.17 ± 0.03	0.17 ± 0.03	0.17 ± 0.03	P < 0.0001	P = 0.22	18%	14%	39%	40%
SPW (cm)	3.27 ± 0.32	3.29 ± 0.30	3.28 ± 0.27	3.26 ± 0.27	3.28 ± 0.29	3.29 ± 0.27	P = 0.10	P = 0.70	12%	9%	8%	5%
CSA (cm2)	2.77 ± 0.43	2.76 ± 0.42	2.74 ± 0.40	2.77 ± 0.43	2.74 ± 0.40	2.71 ± 0.40	P < 0.001	P = 0.25	26%‡‡	14%	43%	47%
CSMI (cm4)	2.44 ± 0.65	2.47 ± 0.59	2.43 ± 0.59	2.42 ± 0.56	2.42 ± 0.55	2.38 ± 0.49	P = 0.056	P = 0.67	27%‡‡	15%	27%	30%
Z (cm3)	1.35 ± 0.29	1.35 ± 0.26	1.34 ± 0.26	1.34 ± 0.25	1.33 ± 0.24	1.28 ± 0.23***,¶¶	P < 0.0001	P = 0.0011	35%‡‡‡	17%	41%‡	56%
Buckling ratio	10.8 ± 2.6	11.3 ± 2.4	11.1 ± 2.2	11.0 ± 2.6	11.2 ± 2.6	11.6 ± 2.7*	P < 0.0001	P = 0.011	27%	29%	10%	7%
Intertrochanteric
Cortical thickness (cm)	0.40 ± 0.07	0.40 ± 0.07	0.40 ± 0.07	0.40 ± 0.07	0.40 ± 0.07	0.40 ± 0.06	P = 0.13	P = 0.75	42%	39%	43%	44%
SPW (cm)	5.53 ± 0.51	5.67 ± 0.53	5.66 ± 0.54	5.48 ± 0.47	5.54 ± 0.45	5.55 ± 0.46	P < 0.0001	P = 0.065	49%	44%	18%	18%
CSA (cm2)	4.85 ± 0.87	4.86 ± 0.87	4.79 ± 0.81	4.86 ± 0.74	4.87 ± 0.73	4.86 ± 0.68	P = 0.031	P = 0.25	34%	44%	50%	39%
CSMI (cm4)	13.8 ± 4.2	13.7 ± 4.2	14.0 ± 3.9	13.6 ± 3.3	13.5 ± 3.3	13.7 ± 3.9	P = 0.14	P = 0.95	48%	49%	42%	38%
Z (cm3)	4.28 ± 1.11	4.32 ± 1.05	4.23 ± 0.98	4.24 ± 0.90	4.20 ± 0.92	4.22 ± 0.84	P = 0.35	P = 0.17	42%	46%	49%	44%
Buckling ratio	8.23 ± 1.78	8.31 ± 1.59	8.32 ± 1.59	8.19 ± 1.80	8.16 ± 1.72	8.24 ± 1.76	P = 0.27	P = 0.45	48%	48%	30%	34%
Shaft
Cortical thickness (cm)	0.58 ± 0.12	0.58 ± 0.11	0.59 ± 0.11	0.59 ± 0.12	0.58 ± 0.12	0.57 ± 0.11*	P = 0.26	P = 0.017	43%‡‡	29%	30%	41%
SPW (cm)	2.95 ± 0.24	2.93 ± 0.27	2.92 ± 0.24**	2.93 ± 0.22	2.92 ± 0.22	2.92 ± 0.21	P = 0.0072	P = 0.044	17%‡‡	37%	51%	40%
Shaft CSA (cm2)	4.18 ± 0.58	4.19 ± 0.58	4.18 ± 0.60	4.15 ± 0.55	4.16 ± 0.55	4.12 ± 0.55	P = 0.017	P = 0.17	46%	36%	38%	49%
Shaft CSMI (cm4)	3.54 ± 0.93	3.51 ± 1.00	3.48 ± 0.92	3.43 ± 0.72	3.46 ± 0.75	3.41 ± 0.72	P = 0.096	P = 0.25	31%	36%	53%	44%
Shaft Z (cm3)	2.31 ± 0.45	2.31 ± 0.46	2.31 ± 0.44	2.27 ± 0.35	2.28 ± 0.36	2.25 ± 0.35¶	P = 0.10	P = 0.03	45%‡	32%	38%‡	53%
Buckling ratio	2.76 ± 0.62	2.67 ± 0.55§	2.70 ± 0.61	2.76 ± 0.69	2.72 ± 0.63	2.77 ± 0.66	P = 0.029	P = 0.045	22%	26%	32%	25%

All values are means ± SD.

*Year 2 vs baseline (Bonferroni post hoc, P < 0.05).

**Year 2 vs baseline (Bonferroni post hoc, P < 0.01).

***Year 2 vs baseline (Bonferroni post hoc, P < 0.0001).

¶ Year 2 vs year 1 (Bonferroni post hoc, P < 0.05).

¶¶ Year 2 vs year 1 (Bonferroni post hoc, P < 0.001).

§ Year 1 vs baseline (Bonferroni post hoc, P < 0.05).

‡ Percentage of creatine participants different from percentage of placebo participants (chi-square, P < 0.05).

‡‡ Percentage of creatine participants different from percentage of placebo participants (chi-square, P < 0.01).

‡‡‡ Percentage of creatine participants different from percentage of placebo participants (chi-square, P < 0.001).

Group – time interactions for section modulus (P = 0.03) and buckling ratio (P = 0.049) at the femoral neck remained significant in the valid-completers analysis (see Supplemental Table 3, Supplemental Digital Content, hip structural analysis measures at baseline, year 1, and year 2 for creatine and placebo groups (valid completers), http://links.lww.com/MSS/C857), but the group – time interactions at the femoral shaft were no longer significant (see Supplemental Table 3, Supplemental Digital Content, http://links.lww.com/MSS/C857). Bonferroni post hoc analyses indicated that in the control group, narrow neck section modulus decreased from baseline to year 2 (P < 0.001) and from year 1 to year 2 (P < 0.05), and buckling ratio increased from baseline to year 2 (P < 0.01). There was a greater percentage of participants from the creatine compared with the placebo group who exceeded the minimal difference to be real for increases in CSA (P = 0.0013), CSMI (P = 0.0027), and section modulus (P < 0.0001) at the narrow part of the femoral neck, and for decreases in CSA (P = 0.007) at the intertrochanteric site. There was a greater percentage of participants from the placebo compared with the creatine group who exceeded the minimal difference to be real for decreases in section modulus (P = 0.025) at the narrow part of the femoral neck and for increases in CSA (P = 0.027) at the intertrochanteric site, and SPW (P = 0.011) at the femoral shaft (see Supplemental Table 3, Supplemental Digital Content, http://links.lww.com/MSS/C857).

There were no differences between groups in the compliers analysis from the MANOVA (see Supplemental Table 6, Supplemental Digital Content, hip structural analysis measures at baseline, year 1, and year 2 for creatine and placebo groups (compliers), http://links.lww.com/MSS/C857). There was a greater percentage of participants from the creatine compared with the placebo group who exceeded the minimal difference to be real for increases in cortical thickness (P = 0.018), and sectional modulus (P = 0.0009) at the narrow part of the femoral neck, and CSA (P = 0.005) at the femoral shaft, and for decreases in SPW (P < 0.0001) and buckling ratio (P < 0.0001) at the narrow part of the femoral neck, and cortical thickness (P = 0.048), CSA (P < 0.0001), CSMI (P = 0.015), SPW (P = 0.0009), and buckling ratio (P = 0.0035) at the intertrochanteric site. There was a greater percentage of participants from the placebo compared with the creatine group who exceeded the minimal difference to be real for increases in cortical thickness (P = 0.014), CSA (P = 0.002), CSMI (P = 0.0043), and sectional modulus (P = 0.036) from the intertrochanteric site, and for decreases in CSA (P = 0.012) at the femoral shaft (see Supplemental Table 6, Supplemental Digital Content, http://links.lww.com/MSS/C857).

Exercise Performance Measures

Exercise performance measures for the intent-to-treat analysis are presented in Table 3. There was a significant group – time interaction from the MANOVA from the intent-to-treat analysis (P = 0.0073). Univariate tests indicated a significant group – time interaction for walking time (P = 0.0008). The Bonferroni post hoc test indicated that time to walk 80 m decreased in the creatine group from baseline to year 1 and baseline to year 2 (both P < 0.0001). This group – time interaction remained significant in the valid-completers analysis (Supplemental Table 4, Supplemental Digital Content, exercise performance measures at baseline, year 1, and year 2 for creatine and placebo groups (valid completers); P = 0.013, http://links.lww.com/MSS/C857) with Bonferroni post hoc testing indicating the creatine group decreased from baseline to year 1 and baseline to year 2 (P < 0.001). In all analyses, both groups significantly improved bench press and squat strength (P < 0.0001) with no differences between groups (Table 3 and Supplemental Tables 4 and 7, Supplemental Digital Content, http://links.lww.com/MSS/C857). There were no differences between groups for any exercise measures in the compliers analysis (Supplemental Table 7, Supplemental Digital Content, http://links.lww.com/MSS/C857). There was a greater percentage of participants from the creatine compared with the placebo group who exceeded the minimal difference to be real for improvement in walking speed (decreased time to cover 80 m; P < 0.0001) in all analyses (Table 3 and Supplemental Tables 4 and 7, Supplemental Digital Content, http://links.lww.com/MSS/C857).

TABLE 3.

Exercise performance measures at baseline, year 1, and year 2 for creatine and placebo groups.

	Creatine (n = 120)			Placebo (n = 117)			Time	Group × Time	% of Participants Exceeding Minimal Difference to Be Real
	Baseline	Year 1	Year 2	Baseline	Year 1	Year 2			Creatine	Placebo	Creatine	Placebo
									Increase	Increase	Decrease	Decrease
Bench press (kg)	32.1 ± 12.7	44.0 ± 14.0	42.6 ± 14.1	30.6 ± 10.9	40.8 ± 10.5	41.4 ± 14	P < 0.0001	P = 0.20	67%	70%	5%	4%
Hack squat (kg)	57.6 ± 21.6	82.1 ± 28.8	84.4 ± 28.1	56.6 ± 24.0	80.3 ± 26.1	82.7 ± 25.0	P < 0.0001	P = 0.97	65%	69%	4%	2%
Balance (s)	43.9 ± 15.8	34.8 ± 12.6	31.3 ± 11.6	44.9 ± 19.0	35.6 ± 17.2	31.4 ± 12.1	P < 0.0001	P = 0.85	0%	3%	47%	48%
Time (s) to walk 80 m	48.6 ± 5.6	47.3 ± 5.5*	47.1 ± 5.4*	48.3 ± 4.5	47.8 ± 4.8	48.2 ± 4.9	P < 0.0001	P = 0.0008	4%	7%	19%‡‡‡	6%

All values are means ± SD.

*Years 1 and 2 vs baseline (Bonferroni post hoc, P < 0.0001).

‡‡‡Percentage of creatine participants different from percentage of placebo participants (chi-square, P < 0.001).

Uncontrolled Intervention Factors (Physical Activity Score and Dietary Intake)

Changes in physical activity and diet (outside of the intervention) between groups are shown in Table 4. There were no group – time interactions for physical activity or daily intake of calories, protein, calcium, or vitamin D.

Falls and Fractures

From baseline to 3 yr (i.e., 1 yr after the end of the intervention), reported falls were 30 in the creatine group and 19 in the placebo group (P = 0.24). There were four fractures in each group. Fractures in the creatine group included spine, patella, radius, and ribs; fractures in the placebo group included finger, toes (two), and ankle. None of these fractures were considered related to the exercise intervention.

Adverse Events

Kidney

Twenty kidney-related adverse events occurred in 15 participants in the creatine group, and 24 kidney-related adverse events occurred in 21 participants in the placebo group (P > 0.05). These included high urine albumin–creatinine ratio (two creatine participants), high urine microalbumin (three creatine and six placebo participants), hemoglobin in the urine (three creatine and eight placebo participants), low creatinine clearance (one creatine participant and 2 placebo participants), high blood creatinine (six creatine participants), low estimated glomerular filtration rate (four creatine participants), protein in the urine (six placebo participants), low urine creatinine (one placebo participant), kidney cysts (one placebo participant), and kidney infection (one creatine participant). All kidney-related adverse events were rated as “possibly related” to the intervention and “mild” in severity, except the kidney infection, which was rated as “moderate” in severity.

Liver

Nine liver-related adverse events were reported in seven participants in the creatine group, and six liver-related adverse events occurred in five participants in the placebo group (P > 0.05). These included high liver enzymes in the blood, that is, aspartate aminotransferase, alanine aminotransferase (six creatine and three placebo participants), bilirubin in urine (one creatine participant), and low albumin (one creatine participant and two placebo participants). All liver-related adverse events were rated as “possibly related” to the intervention and “mild” in severity.

Gastrointestinal

Thirteen gastrointestinal adverse events occurred in 13 participants in the creatine group, and 17 gastrointestinal adverse events occurred in 14 participants in the placebo group (P > 0.05). These included bloating, constipation, diarrhea, heartburn, nausea, and acid reflux. These were classified as possibly, probably, or definitely related to the intervention, and all were mild in severity except one case of heartburn (moderate), one case of constipation (moderate), two cases of diarrhea (moderate), two cases of nausea (moderate and severe) in the placebo group, and one case of upset stomach (moderate) and one case of heartburn (moderate) in the creatine group.

Muscle cramps

Two participants in each group reported muscle cramps during the study. These were rated as possibly or probably related to the intervention, and two were rated as “mild” and two as “moderate” (one in the creatine and one in the placebo group) in severity.

Other musculoskeletal adverse events

Other musculoskeletal adverse events reported (n = 172) during the intervention were rated as possibly, probably, or definitely related to the intervention (i.e., the exercise training), and all were rated as mild or moderate except for one rated as “severe” (tendonitis and pain in the Achilles tendon, plantar fascia, and anterior tibia). Most adverse events were related to muscle soreness or joint pain. One participant dropped a 5-lb weight on her toe, which caused bruising. Four participants had falls during the walking intervention, which caused injury (one swollen knee, rated as “moderate,” one neck and upper back pain, rated as “moderate,” one tailbone pain, rated as “moderate,” and shoulder pain, rated as “mild”).

Out of all reported adverse events, only one that was considered possibly, probably, or definitely related to the intervention was rated as “serious” (a fall off the treadmill causing a swollen knee) because the injury was judged to have resulted in impairment of performance of everyday activities. Three participants withdrew from the study due to adverse events: two in the creatine group (due to diarrhea and self-reported allergy to the metal from the exercise machines) and one in the placebo group (due to an injury when falling off a treadmill) (Fig. 1).

DISCUSSION

The main finding from this randomized controlled trial was that creatine monohydrate supplementation during a resistance training and walking program over 2 yr had no effect on BMD at the femoral neck, total hip, or lumbar spine. There was variation across groups for percentages of participants who exceeded minimal differences to be real for changes in BMD in the compliers analysis (e.g., greater percentage of participants in the creatine group exceeding the minimal difference to be real for increases in total hip BMD, but also decreases in Wards and lumbar spine BMD; Supplemental Table 5, Supplemental Digital Content, http://links.lww.com/MSS/C857). However, creatine supplementation preserved a number of geometric properties at the proximal femur compared with placebo (i.e., sectional modulus and buckling ratio at the narrow part of the femoral neck, and cortical thickness, SPW, section modulus, and buckling ratio at the femoral shaft). The preservation of section modulus and cortical thickness (in comparison with the decrease in the placebo group) would preserve strength in bending and compression, respectively, whereas the increased buckling ratio in the placebo group indicates an increased susceptibility of cortical bending under compressive loads (12,13). A decrease in SPW in the creatine group relative to the placebo group at the femoral shaft is also associated with reduced risk of fracture (12,13). Differences between groups for percentage of participants exceeding minimal differences to be real supported a positive effect of creatine on geometric properties at the narrow part of the femoral neck and femoral shaft (Table 2 and Supplemental Tables 3 and 6, Supplemental Digital Content, http://links.lww.com/MSS/C857) but not necessarily at the intertrochanteric region (Supplemental Tables 3 and 6, Supplemental Digital Content, http://links.lww.com/MSS/C857). Some differences at the intertrochanteric region indicated greater beneficial changes in the placebo group (e.g., cortical thickness, CSA, CSMI, and section modulus), and some indicated greater beneficial changes in the creatine group (e.g., SPW and buckling ratio) in the compliers analysis (Supplemental Table 6, Supplemental Digital Content, http://links.lww.com/MSS/C857). At the cellular level, creatine stimulates differentiation of osteoblasts (i.e., cells involved in bone formation) (5), but osteoclasts (cells involved in bone resorption) are also dependent on creatine kinase, the enzyme involved in breakdown of phosphocreatine (29), and therefore might also be responsive to creatine supplementation. We speculate that perhaps creatine supplementation stimulates remodeling of bone (i.e., both formation and resorption) to alter geometric properties and whether formation or resorption predominates may depend on the location of bone in the proximal femur.

The lack of an effect of creatine on BMD does not support our primary hypothesis, which was based on a smaller, shorter-duration study (i.e., 1 yr) in postmenopausal women, where we observed a significant preservation of BMD at the femoral neck with creatine compared with placebo during a supervised resistance training program (9). Differences between studies are likely not due to the dose of creatine, as it was higher in the current study. Characteristics of the women were similar across studies, as was the resistance training program; therefore, the difference between studies is most likely due to the smaller sample size in our 1-yr study (and therefore a chance statistical finding) or perhaps to the higher compliance with the creatine supplement in the previous study (i.e., 79% vs 56% in the current study). A lack of an effect of creatine (20 g·d⁻¹ for 5 d, 5 g·d⁻¹ for 23 wk) during supervised resistance training on hip or lumbar spine BMD in postmenopausal women was also observed in a shorter-duration study (i.e., 6 months) (30). Smaller doses of creatine supplementation (1–3 g·d⁻¹ for 1–2 yr) without resistance training were also ineffective for improving BMD in postmenopausal women (31,32). Collectively, creatine supplementation has a very minimal effect (if any) on BMD in postmenopausal women.

In contrast to its lack of effect on BMD, creatine supplementation preserved several geometric measures at the proximal femur in comparison with placebo, which could improve compressive and bending strength of bone and reduce susceptibility to cortical buckling under compressive loads in the creatine compared with the placebo group (12,13). Compared with placebo, creatine supplementation preserved section modulus and buckling ratio at the narrow part of the femoral neck, and cortical thickness, SPW, section modulus, and buckling ratio at the femoral shaft, all changes that are associated with decreased hip fracture risk in postmenopausal women (12,13). The changes observed in our study are however most likely below the threshold for clinical prevention of fracture. For example, a 1 SD change in geometric properties such as femoral neck or shaft cortical thickness, section modulus, and buckling ratio is associated with significant changes in fracture risk in longitudinal assessment of older women (followed for approximately 13 yr) (33). Our changes in these variables were much less than 1 SD, indicating that a longer duration of creatine supplementation would be necessary or that creatine supplementation may need to be combined with other therapies to significantly prevent fracture. In our smaller, shorter-term studies, we showed that 1 yr of creatine supplementation during supervised resistance training tended to increase (P = 0.061) section modulus at the narrow part of the femoral neck in older men (14). Using peripheral quantitative computed tomography (pQCT), we also found beneficial effects of creatine supplementation on bone geometry (i.e., increased total area of bone at the distal and shaft sections of the tibia) during a resistance training program in postmenopausal women and older men (15). Mechanistically, creatine supplementation may activate cells involved in bone formation (i.e., osteoblasts) (5) or may reduce bone resorption (8,34). Creatine may especially be effective when combined with exercise because exercise stimulates creatine uptake into muscle (35), leading to increases in phosphocreatine stores, which stimulates adenosine triphosphate resynthesis during short-term, high-intensity exercise, such as resistance training. Creatine supplementation is therefore more effective for increasing lean tissue mass when combined with resistance training than creatine supplementation alone without training (30). Increased lean mass may allow for increased mechanical stress on bone, stimulating a net bone formation, which may contribute to the improved geometric properties (36).

Our study indicated a beneficial effect of creatine supplementation combined with supervised resistance training for increasing lean tissue mass (in the valid-completers analysis) and walking speed but no effect above resistance training alone for improving hack squat or bench press strength. Previous meta-analyses show creatine supplementation combined with resistance training is more effective than resistance training alone for improving lean tissue mass, strength (37–39) and performance of functional tests (40,41) in older adults. Our previous study using pQCT indicated that creatine supplementation during supervised resistance training improved muscle quality (i.e., muscle density) of the lower limb (15), which is predictive of performance of functional tasks (i.e., timed-up-and-go) in older women (42); this may contribute to the enhanced walking speed with creatine supplementation in the current study. Our finding of increased walking speed agrees with a recent study in stroke survivors that found an improvement in 6-min walking distance from creatine supplementation and supervised resistance training (43). The lack of change for strength measures with creatine supplementation in the current study may be due to the large beneficial effect of resistance training alone on muscular strength, which may mask any small improvement due to creatine supplementation. Our hack squat strength measure is quite variable; this could also account for the difficulty in detecting any small effect due to creatine supplementation. We included this exercise as part of our training program, instead of leg press, which has a lower coefficient of variation in older adults (21), because we believe it provides a more direct stimulus to the hip area to improve bone status. Our previous studies indicated a lack of effect of leg press training on hip bone mineral (24) but a beneficial effect of hack squat training perhaps because of more direct loading of the hip during the hack squat (18).

An important aspect of our study was that a relatively large dose of creatine over 2 yr resulted in minimal adverse events, including markers of kidney and liver function. This agrees with previous studies of older women that had shorter duration and involved lower does (9,44). Many of the kidney-related adverse events in the creatine group were due to elevated blood creatinine levels or lower estimated glomerular filtration rate, which is estimated using levels of blood creatinine. Because creatinine is a metabolic by-product of creatine degradation, it is not surprising that blood creatinine levels were sometimes elevated, and this may not indicate kidney problems (45).

A limitation of our study was the relatively low compliance with the creatine and placebo supplements, low compliance to the exercise protocols, and high attrition rate. One contributing factor for the low compliance with the supplements was the conservative approach we took when dealing with kidney or liver adverse events. For example, if a participant experienced blood or urine markers of kidney or liver dysfunction outside the normal reference range, their supplement dosage was reduced.

CONCLUSIONS

In summary, 2 yr of creatine supplementation during a resistance training and walking program in postmenopausal women had no beneficial effects on BMD yet resulted in several improvements in bone geometry at the proximal femur (i.e., increased cortical thickness and section modulus, and reduced SPW and buckling ratio). These changes may be protective against hip fracture. Longer-term follow-up with larger sample sizes would be needed to confirm protection against hip fracture with creatine supplementation.