Risk factors for osteoporosis

The clinical manifestations of osteoporosis occur after many years, sometimes decades, of bone loss. Treatment of the established condition may prevent further bone loss or even increase bone mass but it cannot reverse spinal deformity or remove the sometimes persistent pain of existing vertebral fractures. Furthermore, the disruption of trabecular architecture associated with severe bone loss is unlikely to be reversed by any treatment currently available, so that bone strength may remain compromised despite restitution of bone mass. In contrast, osteoporosis can be prevented in the vast majority of cases, provided that those at risk are detected sufficiently early. A number of prospective studies have shown that fracture risk can be stratified according to bone mass and hence a single measurement of bone mass is of some predictive value (Gardsell et al., 1989; Hui et al., 1989); although other factors clearly contribute to fracture risk, measurement of bone mass represents the best method currently available for detecting those at risk from fracture. At present there is general agreement that, for a number of reasons, population screening cannot be justified and facilities for bone densitometry remain restricted to relatively few centres. Another possible approach to early detection of osteoporosis is the use of clinical and historical risk factors to predict bone mass and/or fracture; established and putative risk factors include: age, female sex, CaucasianiAsian race, late menarche, early menopause, low parity, low body weight, amenorrhoea, cigarette and alcohol consumption, immobilization, hyperthyroidism, corticosteroid therapy, low dietary calcium intake and malabsorption. Risk factors for low bone mass may differ between skeletal sites and are known to differ for fracture sites; for example, factors which affect the risk of falling and the associated protective neuromuscular responses are important for hip fracture but much less so for vertebral fracture. The ability of risk factors to predict bone mass and vertebral fracture risk has recently been evaluated by several groups and in this issue, Ribot er al. (1992) report the predictive value of historical and anthropometric risk factors for vertebral bone mineral density in 1565 peri and postmenopausal women. Women for this study were selected from a menopause clinic; those with a history of osteoporotic fracture or hormone replacement therapy were excluded as were those with a previous or current history of endocrine disease, malabsorption or corticosteroid administration. A supervised questionnaire was administered and vertebral bone mineral density assessed by dual photon absorptiometry. Using multiple stepwise logistic regression, weight, age at menopause, years since menopause, age at menarche and height were shown to be independent predictors of vertebral bone mass, the first three of these having the strongest predictive value. However, these variables accounted for only 30-35% of the observed variance in bone mass. At best, the regression model correctly classified 73% of women with low bone mass and 66% of those with normal bone mass; this degree of sensitivity and specificity is clearly insufficient to replace bone mass measurements, at least in prediction of bone mass itself, although if screening by bone densitometry were introduced, it could reduce the number of women requiring assessment by around 25%. Other groups have also reported that risk factor analysis performs poorly as a screening procedure for osteoporosis. Menopausal status was the only significant determinant of lumbar bone mass in 286 Dutch peri or post-menopausal women and, together with other risk factors, accounted for only 31% of the variance in bone mass (Elders et af., 1989). Slemenda er al. (1990), in a study of 124 perimenopausal women, found that the best possible prediction of bone mass by risk factor analysis provided correct classification of 68% of those with low bone mass and 77% of those with high bone mass; they concluded that such analysis could eliminate the need for bone densitometry in approximately 30% of the population. In their study only a limited number of risk factors were included and bone mass was assessed at the radius, spine and femur, prediction of bone mass being best for the radius. In a long-term, prospective study, Hansen et al. (1991) investigated the influence of risk factors on bone mass and post-menopausal bone loss in 121 healthy postmenopausal women. While high dietary calcium intake, lactation and oral contraceptive use were associated with high bone mass at menopause, risk factors were generally poor predictors of either bone mass or post-menopausal bone loss. The performance of risk factor analysis appears to be poorer when used to predict prevalent vertebral fracture. In a retrospective study of 663 post-menopausal women, 266 of whom had one or more vertebral fractures, Kleerekoper er al. (1989) reported that analysis of a large range of risk factors provided poor sensitivity and specificity for prediction of vertebral fracture; the optimal sensitivity of 63% was associated with a specificity of only 39%. Similar conclusions were reached by Cooper et al. (1991), in a multi-centre