Calcified Tissue International

, Volume 83, Issue 2, pp 93–100

Grip Strength May Facilitate Fracture Prediction in Perimenopausal Women with Normal BMD: A 15-Year Population-Based Study


    • Bone and Cartilage Research UnitUniversity of Kuopio
    • Department of Orthopedics and TraumatologyKuopio University Hospital
  • Toni Rikkonen
    • Bone and Cartilage Research UnitUniversity of Kuopio
  • Marjo Tuppurainen
    • Bone and Cartilage Research UnitUniversity of Kuopio
    • Department of Obstetrics and GynecologyKuopio University Hospital
  • Jukka S. Jurvelin
    • Department of Clinical Physiology and Nuclear MedicineKuopio University Hospital
  • Esko Alhava
    • Department of SurgeryKuopio University Hospital
  • Heikki Kröger
    • Bone and Cartilage Research UnitUniversity of Kuopio
    • Department of Orthopedics and TraumatologyKuopio University Hospital

DOI: 10.1007/s00223-008-9155-0

Cite this article as:
Sirola, J., Rikkonen, T., Tuppurainen, M. et al. Calcif Tissue Int (2008) 83: 93. doi:10.1007/s00223-008-9155-0


The aim of the present study was to investigate the ability of grip strength measurements to predict fracture risk in perimenopausal women according to bone mineral density (BMD). A random sample of 971 perimenopausal women from the Kuopio Osteoporosis Risk Factor and Prevention study cohort was measured with dual-energy X-ray absorptiometry (DXA) at the femoral neck and grip strength with a pneumatic squeeze dynamometer in 5-year intervals from baseline (1989–1991). Fractures during the 15-year follow-up were recorded based on self-reports and validated from medical records. In the total sample and in osteopenic or osteoporotic women (T score < −1, n = 284) grip strength was not significantly associated with fracture-free survival rate (P = nonsignificant in Cox regression). In women with normal baseline BMD (N-BMD, T score > −1, n = 687) the lowest grip strength quartile had a significantly lower fracture-free survival rate in the Cox proportional hazard model (P = 0.005, hazard ratio [HR] = 2.0). In the multivariate Cox regression model, T score and grip strength were the only significant predictors of 15-year fracture-free survival in the N-BMD group and a risk index (RI) was formed according to HRs of these two variables. High RI (0–5 points) was associated with significantly lower 15-year fracture-free survival rate (P = 0.001, HR = 0.137) in the N-BMD group. In contrast, 5-year T score was no better a predictor of fractures in the baseline N-BMD group (P = 0.04, HR = 0.36). In conclusion, grip strength predicts 15-year fracture-free survival in perimenopausal women with N-BMD, while 5-year DXA does not seem to be any better a predictor of fracture risk. DXA measurements could be coupled with simple and cost-effective grip strength measurements. Controlling BMD in women with N-BMD could be abandoned.


Dual-energy X-ray absorptiometryFracture riskGrip strengthOsteoporosisPopulation-based studyPostmenopause

Osteoporotic fractures result in significant morbidity and mortality [1, 2]. The risk of fractures is greater among women with high bone loss rate, although bone mineral density (BMD) changes explain only 20% of fractures among the elderly [3]. Accordingly, bone structure may be an even more important determinant of individual fracture risk [4, 5]. The methods for quantification of bone quality, however, are not universally adopted. Thus, planar BMD remains the best clinical indirect means for determination of bone fragility.

The diagnosis of osteoporosis is based on central spine and hip dual-energy X-ray absorptiometric (DXA) measurements. Several risk factor–based clinical decision rules have been developed to aid in directing patients to BMD measurements (e.g., SCORE, ABONE) [6] and to evaluate individual risk of future fracture (e.g., fracture index [FI]) [7]. While the sensitivity of these clinical decision rules is fairly good, the specificity is at most moderate [6]. Most recently, a score for evaluation of bone fragility risk (FRAX) was introduced, which combines BMD with other fracture risk factors for prediction of 10-year probability of fractures for both men and women [8].

A significant part of patients will score normal BMD (N-BMD) by DXA (T score > −1) and still develop fragility fractures in later postmenopausal years. The follow-up protocols, such as timing of control BMD measurements, for patients with N-BMD have not been generally accepted, nor is it clear whether these patients’ bone density should be measured again at all. However, some nonosteoporotic patients are at significant risk of developing low-trauma energy fractures in due course, which indeed may represent the majority of all fracture patients [9].

The use of muscle strength measures and a variety of physical activity scales in predicting fractures, BMD, bone loss, and falls has been studied recently; good muscle strength and its maintenance seem to be associated with decreased fracture incidence, higher BMD, as well as lower bone loss rate [1012]. However, the usefulness of simple grip strength measurements in concordance with DXA measurements in the prediction of low-trauma energy fractures has not been studied. Whether such simple and inexpensive measurements would prove useful in the evaluation of long-term fracture risk would aid in the follow-up of postmenopausal patients for fragility fractures.

The aim of the present study was to investigate the clinical usefulness of simple grip strength measurement in the prediction of long-term 15-year fracture risk among perimenopausal women according to BMD.

Materials and Methods

Study Population

The study population was formed based on the prospective Kuopio Osteoporosis Risk Factor and Prevention (OSTPRE) study cohort. The OSTPRE cohort was established in 1989 by selecting all women born in 1932–1941 and resident in Kuopio Province, Finland (n = 14,220) [13]. The baseline postal inquiry, including questions, e.g., about health disorders, medication, use of hormone replacement therapy (HRT), gynecological history, nutritional habits, calcium intake, physical activity, alcohol consumption, smoking habits, and anthropometric information, was sent to these women at baseline in 1989. The 5-year (in 1994–1997) and 10-year (in 1999–2001) follow-up questionnaires were sent to the 13,100 women who responded at baseline, with response rates of 11,954 (5-year) and 11,537 (10-year). The present study was performed retrospectively by stratifying the study group of interest, described below in detail.

Of the 13,100 respondents in 1989, 11,055 (84.4%) were willing to undergo DXA densitometry. A subsample from this cohort was selected for central bone density measurements since the whole population was not resourced for measurement protocols. Accordingly, a densitometry sample of 3,686 women (33.3%) was selected for the measurements, of which 3,222 (87.4%) women were actually willing to undergo baseline densitometry. Of these, the randomly selected population-based sample consisted of 2,025 women and the remaining 1,197 women formed a nonrandom part, which was stratified for other study purposes or labeled as having a high-risk profile (i.e., experienced menopause within 2 years, had certain diseases/medications affecting bone, had multiple behavioral risk factors, were selected for a HRT + vitamin D trial, or were included in an additional rheumatoid arthritis sample [14, 15]). This selection was also partly done for ethical reasons in the case of high-risk subjects precluding population-based follow-up. In all, 1,873 women of the random part underwent the 5-year and 10-year bone density measurements. Serial valid measurements for lumbar spine and neck of femur were recorded for 1,438 women at both baseline and follow-up measurements. Accordingly, severe bone deformities, including spondyloarthritis (also osteophytes), scoliosis, and severe compression fractures, among other inaccuracies, were excluded by a systematic manual review of densitometry reprints by the study group physicians.

Hysterectomized women (for whom it was impossible to define menopausal status) and premenopausally bilaterally ovariectomized women (n = 467) were additionally excluded. All remaining women were postmenopausal or had undergone menopausal transition during the 10-year follow-up period; i.e., they were perimenopausal. The beginning of menopause was defined in this study as 12 months’ amenorrhea [16]. The beginning of amenorrhea was based on self-reports about the last natural periods in the inquiries, and no hormonal samples were collected. Accordingly, the final study population consisted of a random sample of 971 naturally postmenopausal women.

The selection process of the present study population has been outlined with a figure in our previous report [17].

Grip Strength Measurements

Grip strength was measured with a hand-held pneumatic squeeze dynamometer (Martin Vigorimeter; Medizin-Technik, Tuttlingen, Germany) and taken to be the mean of three successive attempts. Grip strength measurement was performed on the isolated dominant hand (the elbow was not allowed to contact any body part). Otherwise, the position of the hand was not uniformly controlled. At baseline dynamometers were of different calibers and no cross-calibration was performed, making the absolute grip strength values (kPa) incomparable. Accordingly, for the purposes of the present study, the study population was divided into quartiles of grip strength at each measurement. The use of grip strength quartiles has been validated previously in the present cohort [17]. The quartiles were, at baseline, 1st (<54 kPa), 2nd (54–64 kPa), 3rd (64–74 kPa), and 4th (>74 kPa) and, at 5-year follow-up, 1st (<60 kPa), 2nd (60–69 kPa), 3rd (69–82 kPa), and 4th (>82 kPa). The intraclass correlation coefficient (ICC) of the grip strength measurements has been shown to be 0.87 for absolute grip strength values and 0.84 for quartiles (< 0.001 in two-way mixed effects reliability analysis) [17].


Fractures during the 15-year follow-up period (1989–2005) were recorded based on questions (baseline, 5-year, 10-year, and 15-year follow-up postal questionnaires) on whether the respondent had suffered a low-trauma energy fracture during the follow-up and, if so, the type, mechanism, circumstances, and treatment of the fracture. All self-reported fractures were validated by cross-checking radiological reports from medical records by study group physicians. However, rib fractures were accepted without radiological evidence if the clinical diagnosis in the medical records was clearly and uniformly a rib fracture. The false-positive rate in self-reported fractures was 16.5% and the false-negative rate was 21.6% [18]. For the purposes of the present study, only low-trauma energy fractures were accepted.

Other Variables

The use of HRT (as an adjustment variable) was calculated based on the self-reported use of estrogen-containing tablets and patches during the follow-up taken for menopausal symptoms. The self-reported use of HRT compared to national medical prescription records has shown good correlation (97.8% of HRT users were true users) and described in detail previously [19]. Other adjustment variables of interest are listed under “Statistical Methods.”

Bone Mass Measurements

DXA measurements of the anteroposterior spine (L2–L4) and femoral neck were carried out using the same Lunar DPX scanner in both baseline and 5-year measurements with the imaging and analysis protocols provided by the manufacturer (Lunar, Madison, WI) and described earlier [14]. The measurements were carried out at Kuopio University Hospital by specially trained nurses. Quality standards were tested on a daily basis. The short-term reproducibility of this method has been shown to be 0.9% for lumbar spine and 1.5% for femoral neck BMD measurements [20]. The long-term reproducibility (coefficient of variation) of the DXA instrument for BMD during the study period, as determined by regular phantom measurements, was 0.4% [15]. For the purposes of the present study, N-BMD, osteopenia, and osteoporosis were defined according to World Health Organization (WHO) criteria with femoral neck T score.

Statistical Methods

Statistical analyses were carried out using the Statistical Package for Social Sciences (SPSS version 11.5; SPSS, Inc., Chicago, IL) for Windows. The annual BMD changes at both measurement sites were calculated according to the following formula:
$$ \left( {{\text{BMD at follow-up }} - {\text{ BMD at baseline}}} \right)/{\text{duration of follow-up}}$$
and reported as a percentage of baseline BMD. Kaplan–Meier curves were used to evaluate the fracture-free survival rate, and Cox’s proportional hazard model was used to determine the corresponding hazard ratios (HRs) and statistical differences. A risk index (RI) for fracture risk was computed by rounding the HR of each significant factor to the nearest integer in the multiple Cox regression model and summing these integers. In this model the value of the control group was set to zero. A corresponding statistical approach has been validated and used in previously introduced RIs [21].

Adjustment for age, baseline weight, baseline height, use of HRT during the follow-up, duration of menopause, and bone-affecting diseases/medication (yes/no) was used when appropriate. The selection of bone-affecting diseases/medication (used in adjustments) was described previously by Kröger et al. [14]. Diseases were renal disease, liver disease, insulin-dependent diabetes, malignancies, rheumatoid arthritis, endocrine abnormalities (parathyroid/thyroid glands, adrenals), malabsorption (including lactose malabsorption), total/partial gastrectomy, postovariectomy status, premenopausal amenorrhoea, alcoholism, and long-term immobilization. Medications included corticosteroids, diuretics, cytotoxic drugs, anticonvulsive drugs, anabolic steroids, calcitonin, bisphosphonates, and vitamin D. The percentage of women with bone-affecting diseases did not differ significantly between the study population and total population samples (39% vs. 45%, respectively; > 0.100).


The primary hypotheses for the present study were (1) grip strength measurements may be used for prediction of fractures in perimenopausal women and this effect may depend on BMD and (2) DXA follow-up measurements may be facilitated with simple grip strength measurements and associated risk factor–based information.

Table 1 represents the characteristics of the study population. The women were aged 48.0–59.5 years (mean 53.3) at baseline, 53.2–65.6 years (mean 59.1) at 5-year follow-up densitometry, and 57.1–70.2 years (mean 63.9) at the 10-year follow-up measurement. In perimenopausal women with normal baseline BMD (T score > −1) 17.2%/28.4% were osteopenic (–2.5 < T score < −1) and 0%/0.5% osteoporotic (T score < −2.5) in 5-year/10-year follow-up measurements, respectively. In the total population 271 women (27.9%) suffered a fracture during the 15-year follow-up. Of these, the most common fracture locations were wrist (33.6%) and malleolar ankle (15.5%) fractures.
Table 1

Baseline and follow-up characteristics of the study population (n = 971)

Continuous variables



Age (years)



Years since menopause



Body mass index



Nutritional calcium intake






Age at beginning of menstruation



Baseline T score (FN)



Baseline T score (lumbar spine)



Category variables



No HRT use during follow-up



No regular alcohol intake



No smoking



Regular physical activity



Mother’s fracture


Grip strength alone did not predict 15-year fracture-free survival in Cox’s regression model in the total population sample (P = nonsignificant [NS]). The study population was further divided according to baseline femoral neck (FN) T score (WHO classification) into a lowered BMD group (T score < −1, n = 284) and N-BMD group (T score > −1, n = 687). Of the baseline osteopenic or osteoporotic women (lowered BMD group) 36.3% (n = 103) and of the baseline normal BMD women 24.5% (n = 168) suffered a low-trauma energy fracture during the 15-year follow-up (P = 0.001 in chi-squared test for any low-trauma energy fracture). In the FN osteopenia or osteoporosis group (T score < −1) grip strength was not predictive of 15-year fracture-free survival rate (P = 0.871, HR = 1.05, 95% confidence interval [CI] 0.58–1.91, between the lowest and highest grip strength quartiles in Cox proportional hazard model). The only significant variable predicting fragility fractures among osteopenia/osteoporosis groups was fracture history (P = 0.030, HR = 0.57, 95% CI 0.35–0.95) in multivariate Cox regression (all other multivariate model variables listed in Table 2).
Table 2

Construction of RI for prediction of perimenopausal fractures in women with normal baseline BMD (T score > −1): adjusted Cox proportional hazards modela





HR (95% CI)

Index coefficientb

Grip quartile





1.96 (1.22–3.13)






1.20 (0.73–1.99)






1.11 (0.69–1.80)







T-score tertilec





2.50 (1.59–3.91)






2.20 (1.41–3.43)







RI = G(0–2) + T(0–3)

aAdjusted for fracture history, body mass index, age, years since menopause, use of HRT (yes/no), alcohol intake, smoking, nutritional calcium intake, and bone-affecting diseases/medications (covariate P = NS in Cox regression)

bIndex coefficient = HR rounded to nearest integer, coefficient of control group set to zero [21]

cT-score tertiles in women with normal baseline BMD (T score > −1): 1, −1 to −0.44; 2, −0.44 to 0.24; 3, >0.24

Figure 1 depicts the association of baseline grip strength measurement with 15-year fracture-free survival rate for women with normal FN baseline BMD in the Cox proportional hazard model. Accordingly, women within the lowest grip strength quartile had a two times lower fracture-free survival rate during the 15-year follow-up in comparison to the highest quartile (P = 0.007, HR = 2.0, 95% CI 1.2–3.1). The Cox regression analysis on control (5-year) grip strength quartiles produced essentially the same results for 15-year fragility fractures (HR = 2.1, P = 0.02, 95% CI 1.2–3.9; data not shown). In the covariate Cox regression model only grip strength (P = 0.010, HR = 1.96, 95% CI 1.22–3.13) and baseline T score (< 0.001, HR = 2.50, 95% CI 1.59–3.91) remained statistically significant predictors of fracture-free survival at both baseline and 5-year follow-up. Other covariates in these models included age, weight, height, use of HRT during the follow-up, duration of menopause, alcohol intake, smoking, fracture history, nutritional calcium intake, and bone-affecting diseases/medication (yes/no), with no significant association to 15-year fracture-free survival (> 0.100 in Cox multivariate regression).
Fig. 1

Fracture-free survival according to grip strength quartile in women with N-BMD. Cox proportional hazards model. HR = 2.0 (95% CI 1.2–3.1), P = 0.005. For adjustment variables, see Table 2, GS, grip strength

Table 2 shows the construction of a predictive model (RI) for fracture-free survival for women with normal FN baseline BMD. Accordingly, for this Cox proportional hazard analysis only statistically significant variables in the previous multivariate model (i.e., grip strength and baseline T score) were included. The baseline T score was further devided into tertiles among the women with T score > −1 (see Table 2 for thresholds). The statistical basis for RI model integer formation was as described in “Materials and Methods” and reported previously [21]. The RI points ranged 0–5 based on the individual factor-related HRs in the multivariate Cox regression model. Table 3 represents the distribution of follow-up fractures in the present study population according to baseline RI and 5-year follow-up FN T score. Accordingly, 43% of the women with five RI points and 50% of women with follow-up T score < −2 suffered a low-energy fracture during the follow-up.
Table 3

Distribution of follow-up fractures according to RI (baseline to 15-year follow-up fractures) and follow-up T score (5-year to 15-year follow-up fractures) in women with normal start point BMD


Follow-up fracture (BL–15 years) (%)
























Follow-up (5-year) T score

Follow-up fracture (5–15 years) (%)



Over −1



−2 to −1



Under −2






Figure 2 depicts the predictive ability of baseline RI for 15-year fracture-free survival in the Cox proportional hazard model in women with normal baseline FN BMD. Accordingly, RI was inversely highly predictive for fracture-free survival rate (P = 0.001, HR = 0.14, 95% CI 0.04–0.46). Figure 3 shows the predictive ability of follow-up (5-year measurement) FN T score for 15-year fracture-free survival. The lowest FN T-score limit was set to −2 (considered as treatment threshold by National Osteoporosis Foundation) for statistical purposes because of the small number of women with T score < −2.5 (WHO limit for osteoporosis). Accordingly, women with lower follow-up FN T score had shorter fracture-free survival (P = 0.04), while the survival percentage at the end did not differ from that of RI.
Fig. 2

Fracture-free survival according to RI (at baseline measurement) in women with normal baseline BMD. Cox proportional hazards model
Fig. 3

Fracture-free survival according to follow-up (5-year measurement) T score in women with normal baseline BMD. Cox proportional hazards model


The present 15-year population-based follow-up study evaluated the usefulness of grip strength measurement in conjunction with central DXA in prediction of perimenopausal fractures. Accordingly, grip strength proved to be highly predictive of fracture-free survival rate in perimenopausal women with N-BMD and was the only predictive variable among these women for fractures in addition to T score. Furthermore, the predictive value of 5-year BMD measurement for fractures was no better in comparison to grip strength assessment in conjunction with DXA at baseline. In women with osteopenic or osteoporotic BMD or in the total population, grip strength was not predictive of fractures. For baseline osteopenic and osteoporotic women, the only significant variable predicting 15-year fragility fractures was fracture history.

The strengths of the present study include its prospective and population-based nature, large base population, as well as long-term follow-up interval. Bone mass and grip strength measurements were performed under the supervision of trained personnel, which may suppress occasional confounders due to measurement errors. Additionally, all the self-reported fractures were validated from the medical records by study group physicians. Finally, comprehensive adjustment, including a variety of bone-affecting diseases, was used in the multivariate models, which should have weakened any possible bias caused by varying sampling fractions.

A possibility of uncontrolled confounding is always present in epidemiological studies. The grip strength measurements were performed with different-caliber pneumatic squeeze components during the baseline and follow-up measurements, and no cross-calibration was performed between the measurement devices. Accordingly, we used grip strength quartiles of each respective measurement point, which was aimed at producing comparable estimation of actual grip strength. Previously, the use of quartiles has shown good correlation (ICC) between the different grip strength measurements in reliability analysis (comparable to interobserver reliability) [17]. In addition, we used mean grip strength as the predictive variable and not maximal grip strength. Previously, however, it was shown that maximum grip strength and average grip strength may be equally consistent [22].

Although the study sample of 2,025 women was randomly selected from the base population of 14,220 women, the final study sample of 971 women presented a relatively small part of the original OSTPRE cohort and the random sample. Accordingly, this heavy selection process includes the risk that the final sample may not be 100% representative of the underlying population and, in this sense, may not be considered random. However, the selection was partly done due to reasons, such as DXA measurement errors, ethical questions, etc., which may not be totally avoided. The selection procedure may additionally suppress the power to detect small effects of different factors on the risk of fractures. Nevertheless, the occurrence of bone-affecting diseases and medications was similar between the study group and the total population samples, which suggests that the study population would be representative of the base population. Furthermore, the definition of “menopausal transition” in the present study was based purely on self-reports according to amenorrhea without information on hormonal levels. Self-reports, however, have proved to be quite accurate in this matter [23].

Previously, it was suggested that muscle strain affects bone mass site-specifically [24, 25]. Accordingly grip strength should not be considered a holistic indicator of muscle performance. In our previous study, however, a strong association of grip strength measurements with physical performance as well as self-rated health was reported; and it may be concluded that changes in grip strength also moderately reflect the overall changes in muscular fitness among the perimenopausal population [26]. This encourages us to further study the use and clinical applications of easily reproducible and cost-effective peripheral muscle strength measurements in the evaluation of overall postmenopausal muscle performance. Unfortunately, we did not have other follow-up muscle strength measures (e.g., quadriceps strength) available. Accordingly, any conclusions over associations between fractures and general muscle performance may not be directly drawn based on the present results. Indeed, grip strength measurements may be affected by comorbidities, e.g., hand joint arthrosis.

It has been suggested that muscle strength may be positively associated with higher BMD as well as lower bone loss and fracture rate [1012]. To our best knowledge, no previous study has evaluated the predictive value of grip strength measurements in conjunction with BMD measurements on long-term occurrence of fragility fractures in the perimenopausal population. According to the present study, grip strength predicts 15-year occurrence of fragility fractures in perimenopausal women with N-BMD (T score > −1) but not in women with osteopenia or osteoporosis (T score < −1). This may be explained by the low start point BMD in osteopenic and osteoporotic women, which could increase the fracture risk to the extent that it outweighs the predictive power of grip (muscle) strength (possibly associated with other qualities of bone). Furthermore, the finding that T score, in addition to grip strength, still predicts fractures among perimenopausal women with N-BMD (T score > −1) stresses the continuous association of BMD with bone fragility. Nevertheless, it has been suggested that fracture risk is profoundly affected, besides BMD, by bone microstructure, i.e., bone quality [27]. Hence, women with normal baseline BMD and additional follow-up fractures may also not have normal bone architecture. On the other hand, the present study further showed that among osteopenic and osteoporotic women the fracture risk was affected by fracture history independently of BMD.

In terms of cost benefit, the number of central DXA measurements should be minimized. The results of the present study may aid in directing patients away from control DXA for women with N-BMD. According to the present study, 5-year DXA did not seem to be a better predictor of 15-year fracture risk in comparison to simple grip strength measurements at baseline. This should be of special interest since the follow-up protocols for women with N-BMD have not been generally adopted, which leaves these women prone to fragility fractures in due course. However, it is not only osteoporotic women who are at increased risk of low-trauma energy fractures: A recent study of Pasco et al. concluded that the burden of fragility fracture originates in women with nonosteoporotic BMD [9]. In the present study the percentage of fractures was, naturally, greater among baseline osteopenic or osteoporotic women than in the normal baseline BMD group, whereas the majority of fractures still occurred among women with normal baseline BMD and not in women with lowered baseline BMD.

In conclusion, grip strength seems to predict low-trauma energy fractures in women with N-BMD (T score > −1). Among these perimenopausal women, future fracture risk may be determined without controlling BMD in due course. Instead, simple grip strength measurements in conjunction with DXA measurements could be used, which, based on the present study, is no worse at predicting fractures than second BMD measurement. Accordingly, DXA measurements should be accompanied with simple and cost-effective grip strength assessment. The present results also indirectly add to the fact that BMD is not a full measure of fracture risk, but a combination of BMD and bone structure is?


The study was financially supported by the Academy of Finland.

Copyright information

© Springer Science+Business Media, LLC 2008