Can Metacarpal Cortical Area Predict the Occurrenceof Hip Fracture in Women and Men Over 3 Decades

of Follow-Up? Results From the FraminghamOsteoporosis Study

DOUGLAS P. KIEL,1 MARIAN T. HANNAN, 1 KERRY E. BROE,1 DAVID T. FELSON,2 andL. ADRIENNE CUPPLES3

ABSTRACT

The purpose of this study was to determine if a single measurement of metacarpal cortical area could predictthe subsequent risk of hip fracture over a long-term follow-up period. Thirteen hundred eighty-six women and1014 men (mean age [6SD] 61 6 8 years) underwent posteroanterior hand radiography between 1966 and1970 as part of the Framingham Study. Measurements of cortical bone width (external width and medullarywidth) were made at the midpoint of the second metacarpal with a digital caliper to the nearest 0.1 mm. Hipfracture occurrence was ascertained on all survivors through December 1995. Surprisingly, in women, therewas no significant increase in hip fracture according to metacarpal cortical area measurements (per SDdecrease) in either age-adjusted (hazard ratio [HR]5 1.13; 95% CI, 0.94–1.35) or multivariate-adjustedmodels (HR5 1.06; 95% CI, 0.88–1.27). The same results were seen when considering only those women whowere>65 years of age at the time of their X-ray or when considering only the first 10 years of follow-up. Whenthe type of hip fracture was considered in women, after adjustment for other risk factors, there appeared tobe an association between metacarpal cortical area and intertrochanteric fracture risk (HR5 1.24; 95% CI,0.91–1.71) but not femoral neck fracture risk (HR5 0.93; 95% CI, 0.71–1.22). In men, the age-adjusted riskof hip fracture was increased modestly per SD decrease in metacarpal cortical area (HR5 1.38; 95% CI,1.02–1.87), and this remained true after adjustment for potential confounders. In this prospective cohort studywith up to 30 years of follow-up, metacarpal cortical area in men predicted hip fracture risk. In women, theonly association between metacarpal cortical area and fracture risk was observed for intertrochantericfractures and was not significant when adjusting for multiple potential confounders. We conclude that thisperipheral measure of bone status is not a potent predictor of hip fracture over a long period of follow-up. (JBone Miner Res 2001;16:2260–2266)

Key words: bone density, hip fractures, cohort study, men, women

INTRODUCTION

ALTHOUGH IT has been well established that measure-ments of bone mineral density (BMD) can predict hip

fracture risk, a recentmeta-analysis(1) suggested that very

few studies have examined the predictive ability of bonemass measurements made between ages 50 and 60 years topredict subsequent fracture, nor have there been many stud-ies in men, in whom fractures are less common. Finally, fewstudies have followed participants over long periods of

1Hebrew Rehabilitation Center for Aged Research and Training Institute and Harvard Medical School Division on Aging, Boston,Massachusetts, USA.

2Boston University Arthritis Center, Boston University School of Medicine, Boston, Massachusetts, USA.3Department of Epidemiology and Biostatistics, Boston University School of Public Health, Boston, Massachusetts, USA.

JOURNAL OF BONE AND MINERAL RESEARCHVolume 16, Number 12, 2001© 2001 American Society for Bone and Mineral Research

2260

time.(2) Thus, there are no large-scale prospective studiesexamining the ability of a single assessment of bone statusin middle-aged men and women to predict hip fractureoccurring many years later.

Radiogrammetry, an older technique of measuring bonestatus available before the development of modern densito-metric techniques, has the potential to provide insights intolong-term fracture risk prediction. There have been previousstudies evaluating the role of phalangeal bone density usingphotodensitometry,(3) as well as the role of metacarpal ra-diogrammetry.(4,5) Generally, the number of hip fractures inthese studies was small, results for men were limited, andthe number of years of follow-up ranged from 14 to 16years.

Therefore, we analyzed data from a long-standing longi-tudinal cohort study, the Framingham Heart Study. Wewished to determine the predictive usefulness of metacarpalcortical area in determining the risk for hip fracture whenmeasured by radiogrammetry in a group of middle-agedmen and women. We have shown this measurement to bepredictive of breast cancer, colon cancer, and the progres-sion of vascular calcification.(6–8)

MATERIALS AND METHODS

Study population

The Framingham Study began in 1948 with the primarygoal of evaluating risk factors for heart disease in apopulation-based sample of the town of Framingham, MA.The original cohort consisted of 5209 participants who were28–62 years old at the time of the first examination in1948–1951. These individuals (2873 women and 2336men) have participated in follow-up examinations every 2years since that time. At each examination, a medical his-tory is taken and a physical examination and a series oflaboratory tests have been performed.(9)

Between 1966 and 1970, at the time of biennial exami-nations 10–11, a posteroanterior radiograph of the righthand was taken as part of an osteoporosis study. Of the 2408women and 1791 men seen at either of those visits, 1394and 1018 unsolicited men and women, respectively, under-went posteroanterior hand radiography. Not all participantshad radiography performed because funding for the originalproject in 1967 was limited.

Assessment of bone mass

We used radiogrammetry to measure cortical bone massof the second metacarpal on the right hand. We chose thesecond metacarpal because it is one of the largest bones ofthe hand, has a more constant shape than the other meta-carpals,(10) and is approximately circular at the midshaft,with the medullary cavity nearly centered in the tubularbone cylinder.(10,11)

Two readers who were unaware of the hip fracture statusof the study assessed cortical bone mass according to astandard protocol. Hand radiographs were placed flat on alighted viewing box, and measurements of cortical externalwidth (W) and medullary width (w) were made halfway up

the second metacarpal with a digital caliper. Digital caliperswere calibrated to the nearest 0.01 mm, and measurementswere recorded to the nearest 0.1 mm. To assess intraob-server and interobserver reliability in the measurement ofcortical width, we gave 25 hand radiographs to each of thetwo readers twice for blinded readings. The intraobservercorrelation coefficients for external and medullary widthwere 0.99 and 0.94, respectively; the corresponding inter-observer correlation coefficients were identical. We used therelative metacarpal cortical area, calculated as 1003 (W2 2w2) 4 W2, as an indicator of bone mass but also examinedcombined cortical thickness (W 2 w).

Assessment of hip fracture

Osteoporotic hip fractures were defined as fractures of theproximal femur occurring as the result of a fall from stand-ing height or less and not as the result of metastatic canceror greater degrees of trauma such as a motor vehicle acci-dent. Only first hip fractures occurring after the date of theradiograph were included. The methods of hip fractureascertainment in this cohort have been described in detailpreviously.(12) Briefly, all hip fractures were ascertainedusing four sources: chart review completed on all partici-pants, hospital record reviews, direct questioning of partic-ipants at biennial examinations from 1988 to 1995, andtelephone contact with participants not attending examina-tions. Hospital discharge summaries have been used toconfirm fracture in.96% of cases. In the review of hospitalrecords, the location of the hip fracture is recorded (e.g.,femoral neck and intertrochanteric). The rates of hip frac-ture corresponded closely to age-specific rates obtainedfrom the National Hospital Discharge Survey.(12)

Other variables

Information was obtained on potential confounders mea-sured at the time of the radiograph and at previous exams.Age, weight, and height were taken from the exam at thetime of the radiograph. When this was not available, theaverage was taken from the examination before and afterthis exam or the closest examination available. Cigarette usewas determined by averaging the number of cigarettessmoked per day at baseline and previous examinations.Alcohol consumption, assessed at exam 2 and exam 7, wasdetermined by averaging the ounces of alcohol consumedper week at these two assessments (according to a formulareported by Gordon and Kannel(13)). For the 25 participantsmissing this information, alcohol information was takenfrom the closest examination after the X-ray. In the case of7 individuals with missing alcohol information, the gender-specific mean was assigned. Caffeine intake (computed inunits equal to the number of cups of coffee per day1 0.53the number of cups of tea per day) based on previous studiesof caffeine and hip fracture risk(14) was based on the averageintake at exam 4 and exam 12 (exams at which caffeine usewas assessed). For subjects with missing caffeine data,either information from examination 13 was used (2 menand 5 women) or the gender-specific mean (5 men and 12women) was used. For estrogen use at the time of the

2261METACARPAL CORTICAL AREA AND HIP FRACTURE

radiograph, women were categorized into three groups:never used estrogen, former users, and current users. Be-cause some of the women included in the analysis had notyet reached menopause (absence of menstrual periods for 1year) at the time of the hand radiograph (n 5 91), wecreated a variable that classified women as being postmeno-pausal or premenopausal. All aspects of this study wereapproved by the Boston University Institutional ReviewBoard.

Statistical analysis

Using t-tests for continuous variables andx2 tests forcategorical variables, we compared the characteristics of theparticipants at the time of the baseline hand radiographaccording to whether they later sustained a hip fracture ornot. Incidence of hip fractures per 1000 person-years wascalculated. To examine the association between metacarpalcortical area and fracture risk, Cox proportional hazardsregression(15) was used to model time to fracture, withcensoring due to death, loss to follow-up, or maximalfollow-up time (December 1995). Covariates were includedin proportional hazard models after verifying that the vari-ables did not violate the proportional hazards assumption.Models were first performed for men and women separately.Metacarpal cortical area was entered into the models as acontinuous variable and as age-specific quartiles. To obtainbetter control of age effects, subjects were classified accord-ing to age-specific quartiles using 2-year age groupingsfrom age 47 to 80 years. Two models were constructed: ageadjusted, and multivariate adjusted (age, weight, height,smoking, alcohol intake, caffeine intake, and, for women,menopause status and estrogen use). Results for fracture riskwere reported as the hazard ratios (HR) with 95% CIs foreach SD decrease in metacarpal cortical area. We alsocalculated incidence rates of hip fracture with their SEsacross age-specific quartiles. To determine if the predictivevalue of metacarpal cortical area changed with advancingage at the time of hand radiograph, we performed analysesstratified into two age groups:,65 years of age and$65years of age. Finally, to determine if the ability of metacar-pal cortical area to predict hip fracture differed according tothe type of fracture, we separately examined the risk offemoral neck fracture and intertrochanteric fracture amongwomen according to metacarpal cortical area.

RESULTS

Of the 2411 subjects who had hand radiographs, 11 (8women and 3 men) were not included in this analysisbecause of previous hip fracture. This resulted in 1386women and 1014 men with hand radiographs of whom 162(6.08/1000 person-years) and 41 (2.38/1000 person-years)sustained hip fractures, respectively. The median number ofyears from the time of the hand X-ray to the time of hipfracture was 16 years (range, 2–28 years) for women and 19years (range, 4–26 years) for men. As shown in Table 1,women who had incident hip fractures were older at base-line, weighed less, and had lower metacarpal cortical width

than women who did not sustain a hip fracture. Men whofractured were older at baseline and had lower corticalwidth than men without hip fracture; however, there wereno other differences.

Surprisingly, in women, there was no significant increasein hip fracture according to metacarpal cortical area mea-surement in either age-adjusted (HR5 1.13; 95% CI, 0.94–1.35) or multivariate-adjusted models (HR5 1.06; 95% CI,0.88–1.27; Table 2). To determine if an association betweenmetacarpal cortical area and hip fracture risk might be moreapparent in women who were older at the time of baselinehand X-ray, we stratified analyses into those aged$65years versus those,65 years of age at the time of handX-ray. The ability of cortical area to predict hip fracture riskwas not increased in either age group (HR5 1.06; 95% CI,0.80–1.42; and HR5 1.16; 95% CI, 0.92–1.46) in theyounger and older age groups, respectively. When we re-stricted our follow-up to the first 10 years after hand radio-graphs, there was still no association between metacarpalcortical area and the risk of hip fracture. When we repeatedthe analyses using combined cortical width (W2 w) insteadof metacarpal cortical area, results in men and women weresimilar to those using metacarpal cortical area.

When the type of hip fracture was considered in women,there appeared to be an association between metacarpalcortical area and intertrochanteric fractures that was onlystatistically significant at thep , 0.05 level in the age-adjusted models (Table 2). Incidence rates across age-specific quartiles are shown in Fig. 1. In contrast, there wereno suggestions of an association between metacarpal corti-cal area and femoral neck fractures. There were too few hipfractures in men to examine types of hip fracture.

In men, the age-adjusted risk of hip fracture was in-creased modestly per SD decrease in metacarpal corticalarea (HR5 1.38; 95% CI, 1.02–1.87; Table 2), and thisremained true after adjustment for potential confounders.Hip fracture incidence among men in the lowest quartile ofmetacarpal cortical area was 36% higher than among men inthe highest quartile (Fig. 2). The risk ratio for hip fracturefor men in the two lower age-specific quartiles of metacar-pal cortical area was 2.64 (95% CI, 1.34–5.21) comparedwith men in the two upper quartiles.

DISCUSSION

In this population-based prospective cohort study of menand women with up to 30 years of follow-up, a singlemeasurement of metacarpal cortical area had only a verymodest ability to predict later occurrence of hip fracture inmen. In women, metacarpal cortical area was not predictiveof hip fractures overall; however, there was a suggestionthat it may be predictive of intertrochanteric hip fracture butnot femoral neck fracture. To our knowledge, this is thelongest prospective study to examine a peripheral measure-ment of bone status and the risk of hip fracture and is thelargest study sample of men with such a long follow-up andlarge number of hip fractures.

It is well established that measurements of BMD canpredict fracture risk(1,3,16–37); however, the majority of stud-

2262 KIEL ET AL.

ies have measured areal BMD, focused on women withmean age older than our population, and have relativelyshort follow-up periods. The longest cohort study reportedin the literature had up to 25 years of follow-up and reporteda relative risk of hip fracture of 1.66 (95% CI, 1.13–2.46)per 1 SD decrement in BMD at the forearm at baseline.(21)

This cohort of 410 women with areal bone mineral mea-surements made in the early 1970s were followed for all

types of fracture up to 1994. Of the 213 fragility fracturesthat occurred during follow-up, 43 were hip fractures. An-other study of 131 Roman Catholic nuns, almost half ofwhom were premenopausal at the time of initial bone mineralcontent measurement, recorded 31 incident fractures duringfollow-up, and none of the fractures were hip fractures.

The only prospective study in men with long-termfollow-up involved the First National Health and Nutrition

TABLE 1. CHARACTERISTICS OF1014 MEN AND 1386 WOMEN IN THE FRAMINGHAM STUDY ACCORDING TO THEPRESENCE OR

ABSENCE OFHIP FRACTURE BETWEEN 1966AND 1995

CharacteristicsWomen with hip fracture(n 5 162; mean6 SD)

Women without hip fracture(n 5 1224; mean6 SD)

WomenAge at exam of X-ray (year) 656 7.24 606 8.08*Height at exam of X-ray (in) 62.406 2.78 62.156 2.38Weight at exam of X-ray (lb) 136.956 19.88 142.966 24.86*Cortical width (mm) 3.966 0.75 4.256 0.79*Cigarettes (no/day) 4.636 8.12 5.856 8.76Alcohol (oz/week) 2.576 4.01 2.446 3.60Caffeine (.2 cups/day) 65% 60%Menopausal 97% 93%Estrogen use:

Never 79% 74%Former 11% 11%Current 10% 15%

Men with hip fracture(n 5 41; mean6 SD)

Men without hip fracture(n 5 973; mean6 SD)

MenAge at exam of X-ray (year) 646 7.23 606 7.95*Height at exam of X-ray (in) 67.466 3.30 67.266 2.69Weight at exam of X-ray (lb) 170.836 25.35 172.216 25.47Cortical width (mm) 4.936 0.67 5.226 0.80*Cigarettes (no/day) 13.646 11.66 11.106 12.31Alcohol (oz/week) 7.396 6.78 6.436 8.00Caffeine (.2 cups/day) 59% 61%

*p , 0.05.

TABLE 2. ASSOCIATION BETWEEN RELATIVE METACARPAL CORTICAL AREA AND ALL HIP FRACTURES, INTERTROCHANTERIC

FRACTURES, AND FEMORAL NECK FRACTURES FOR1386 WOMEN AND ALL HIP FRACTURES FOR1014 MEN IN THE

FRAMINGHAM STUDY

Risk ratio for fracture (95% CI)

Age-adjusted Multivariate adjusted*

Women—all hip fracturesRelative metacarpal cortical area (per decrease in SD) 1.13 (0.94–1.35) 1.06 (0.88–1.27)

Women—intertrochanteric fractureRelative metacarpal cortical area (per decrease in SD) 1.34 (0.99–1.81) 1.24 (0.91–1.71)

Women—femoral neck fractureRelative metacarpal cortical area (per decrease in SD) 1.04 (0.80–1.35) 0.93 (0.71–1.22)

Men—all hip fracturesRelative metacarpal cortical area (per decrease in SD) 1.38 (1.02–1.87) 1.42 (1.05–1.92)

*Adjusted for age, weight, height average ounces of alcohol per week, average number of cigarettes per day, caffeine intake (0–2cups/day or.2 cups/day), menopause status, and use of estrogen at exam of X-ray (never, former, or current user).

2263METACARPAL CORTICAL AREA AND HIP FRACTURE

Examination Survey (NHANES I) and the epidemiologicfollow-up to this study. A subsample of this cohort (n 51437) was followed from the time of the radiographicabsorptiometric assessment of phalangeal bone density in1971–1975 for up to 21 years. A total of 26 hip fracturesoccurred during follow-up, and the risk of hip fractureincreased by about 70% for each 1 SD decrease in BMD.This is slightly greater than our estimate of 1.42; however,the 95% CIs for both estimates overlap with each other.

Thus, there are limited data regarding the predictability ofareal bone density measurements to predict hip fracture inmen and women beyond 20 years. The longest durationstudy using absorptiometric methods involved the forearmsite that is largely cortical bone and found that bone densitypredicted fragility fractures in general. We also measuredcortical bone in our study and were unable to show anoverall association between cortical width and hip fracturein women. We conclude that measurements of corticalwidth using radiographic morphometry of the metacarpalare not as useful as radial absorptiometric measurementsalthough both sites consist of primarily cortical bone.Whether longer-term follow-up of cohorts who have haddual-energy absorptiometry (DXA) measurements willyield results similar to our own is speculative.

Two prospective studies have specifically examined therelation between metacarpal morphometry and hip fracturerisk.(4,5) In the first,(4) 535 of 771 elderly persons aged$65years and having had X-rays performed in 1973–1974 wereable to be traced for the occurrence of hip fracture. Sixteenhip fractures in women and 7 hip fractures in men occurredduring 14 years of follow-up. Considering men and womentogether, metacarpal cortical index tended to predict hipfracture in this sample (p value for trend5 0.08 acrosstertiles of metacarpal cortical width). The second study useddata from the NHANES and its three follow-up studies andfound that in 1489 white women followed for an average of13 years, external cortical width predicted hip fracture.(5)

Our study differed from these two in several importantways. First, we had data from a larger cohort of both menand women, were able to examine longer duration offollow-up, had complete case ascertainment confirmed bymedical records, and were able to perform gender-specific

analyses as well as looking at the type of hip fracture inwomen.

We were surprised that overall, metacarpal cortical areadid not predict hip fracture in women and only modestlyincreased risk in men. One possibility is that the ability of asingle measurement of bone mass loses its predictive valueover a long duration of follow-up. However, even when werestricted our analysis to older women or to the first 10 yearsof follow-up, metacarpal cortical area was still not associ-ated with an increased risk of fracture. There was a modestability to predict intertrochanteric hip fractures comparedwith femoral neck fractures. The reasons for the differencein the ability of a measurement of cortical bone to predictthe two different types of hip fractures are not entirely clearbut have been described by other investigators.(38,39) Fur-thermore, in a study of the structural significance of corticaland trabecular bone in the proximal femur, using finiteelements analysis, Lotz and colleagues showed that corticalbone in the intertrochanteric region carried 80% of the loadduring gait and during a fall to the side, compared with only50% at the midneck.(40) In fact, there was a gradual transi-tion from the femoral head, where the majority of load wascarried by trabecular bone, to the trochanteric region, wherethe majority of load was carried by cortical bone. Thesefindings may help to explain the ability of cortical bonemeasurements to predict intertrochanteric hip fractures bet-ter than neck fractures. Because the percentage of mensustaining intertrochanteric hip fractures (51%) was greaterthan that in women (39%), this may explain why we wereable to show a significant association between cortical areaand hip fracture in men but only in women sustainingintertrochanteric fractures.

Our study also raises questions about the predictive abil-ity of a single measurement of cortical bone in a peripheralsite to predict hip fracture risk up to 30 years later. Simpleradiogrammetric measurements of the second metacarpal donot appear to perform as well as other measurements such asradiographic absorptiometry in the prediction of long-termhip fracture risk in women. In men, there is a modestincrease in risk for each decrement in metacarpal corticalarea. Because many of the recommendations about bonedensitometry screening are based on short-term follow-up

FIG. 2. Incidence rates (and their SEs) of hip fracture by age-adjusted quartiles of relative metacarpal cortical area in 1014 men inthe Framingham Study.

FIG. 1. Incidence rates (and their SEs) of intertrochanteric hip frac-ture by age-adjusted quartiles of relative metacarpal cortical area in1278 women in the Framingham Study.

2264 KIEL ET AL.

of women of advanced age, further long-term studies suchas this one are needed to be able to evaluate the use of bonemass testing in middle age to predict late-life fractures ofthe hip.

ACKNOWLEDGMENTS

We thank Dr. Harry Genant and Lisa McAllister for theirassistance in the measurement of metacarpal cortical thick-ness on the hand radiographs. This work was supported inpart by the National Institutes of Health (NIH) grant RO1AR/AG 41398, and AR 20613; NIH/NHLBI Contract N01-HC-38038; and an unrestricted grant from Merck HumanHealth.

REFERENCES

1. Marshall D, Johnell O, Wedel H 1996 Meta-analysis of howwell measures of bone mineral density predict occurrence ofosteoporotic fractures. BMJ312:1254–1259.

2. Stegman MR, Recker RR, Davies KM, Ryan RA, Heaney RP1992 Fracture risk as determined by prospective and retrospec-tive study designs. Osteoporos Int2:290–297.

3. Mussolino ME, Looker AC, Madans JH, Edelstein D, WalkerRE, Lydick E, Epstein RS, Yates AJ 1997 Phalangeal bonedensity and hip fracture risk. Arch Intern Med157:433–438.

4. Cooper C, Wickham C, Walsh K 1991 Appendicular skeletalstatus and hip fracture in the elderly: 14-year prospective data.Bone12:361–364.

5. Huang Z, Himes JH 1997 Bone mass and subsequent risk ofhip fracture. Epidemiology8:192–195.

6. Zhang Y, Kiel DP, Kreger BE, Cupples LA, Ellison RC,Dorgan JF, Schatzkin A, Levy D, Felson DT 1997 Bone massand the risk of breast cancer among postmenopausal women.N Engl J Med336:611–617.

7. Zhang Y, Felson DT, Ellison RC, Kreger BE, Schatzkin A,Dorgan JF, Cupples LA, Levy D, Kiel DP 2001 Bone mass andthe risk of colon cancer among postmenopausal women: TheFramingham Study. Am J Epidemiol153:31–37.

8. Kiel DP, Kauppila LL, Cupples LA, Hannan MT, O’DonnellCJ, Wilson PWF 2001 Bone loss and the progression of aorticcalcification over a 25 year period: The Framingham Study.Calcif Tissue Int68:271–276.

9. Dawber TR, Meadors GF, Moore FE 1951 Epidemiologicalapproaches to heart disease: The Framingham Study. Am JPub Health41:279–286.

10. Garn S, Rohmann C, Wagner B, Sascoli W 1967 Continuingbone growth throughout life: A general phenomenon. Am JPhys Anthropol26:313–317.

11. Garn S, Rohmann C, Nolan P Jr 1964 The developmentalnature of bone changes during aging. In: Birren JE (ed.)Relations of Development and Aging. Charles C. Thomas,Springfield, IL, USA, pp. 41–61.

12. Kiel DP, Felson DT, Anderson JJ, Wilson PW, MoskowitzMA 1987 Hip fracture and the use of estrogens in postmeno-pausal women. The Framingham Study. N Engl J Med317:1169–1174.

13. Gordon T, Kannel WB 1983 Drinking habits and cardiovas-cular disease: The Framingham Study. Am Heart J105:667–673.

14. Kiel DP, Felson DT, Hannan MT, Anderson JJ, Wilson PW1990 Caffeine and the risk of hip fracture: The FraminghamStudy. Am J Epidemiol132:675–684.

15. Cox DR 1972 Regression models and lifetables. J R Stat SocB34:187–220.

16. Schott AM, Cormier C, Hans D, Favier F, Hausherr E,Dargent-Molina P, Delmas PD, Ribot C, Sebert JL, Breart G,Meunier PJ 1998 How hip and whole-body bone mineraldensity predict hip fracture in elderly women: The EPIDOSProspective Study. Osteoporos Int8:247–254.

17. Pluijm SM, Graafmans WC, Bouter LM, Lips P 1999 Ultra-sound measurements for the prediction of osteoporotic frac-tures in elderly people. Osteoporos Int9:550–556.

18. Huang C, Ross PD, Wasnich RD 1998 Short-term and long-term fracture prediction by bone mass measurements: A pro-spective study. J Bone Miner Res13:107–113.

19. Huang C, Ross PD, Yates AJ, Walker RE, Imose K, Emi K,Wasnich RD 1998 Prediction of fracture risk by radiographicabsorptiometry and quantitative ultrasound: A prospectivestudy. Calcif Tissue Int63:380–384.

20. De Laet CE, Van Hout BA, Burger H, Weel AE, Hofman A,Pols HA 1998 Hip fracture prediction in elderly men andwomen: Validation in the Rotterdam study. J Bone Miner Res13:1587–1593.

21. Duppe H, Gardsell P, Nilsson B, Johnell O 1997 A single bonedensity measurement can predict fractures over 25 years. Cal-cif Tissue Int60:171–174.

22. Mussolino ME, Looker AC, Madans JH, Langlois JA, OrwollES 1998 Risk factors for hip fracture in white men: TheNHANES I Epidemiologic Follow-up Study. J Bone MinerRes13:918–924.

23. Gardsell P, Johnell O, Nilsson BE 1991 The predictive valueof bone loss for fragility fractures in women: A longitudinalstudy over 15 years. Calcif Tissue Int49:90–94.

24. Gardsell P, Johnell O, Nilsson BE, Gullberg B 1993 Predictingvarious fragility fractures in women by forearm bone densi-tometry: A follow-up study. Calcif Tissue Int52:348–353.

25. Cheng S, Suominen H, Era P, Heikkinen E 1994 Bone densityof the calcaneus and fractures in 75- and 80-year-old men andwomen. Osteoporos Int4:48–54.

26. Cummings SR, Black DM, Nevitt MC, Browner WS, CauleyJA, Genant HK, Mascioli SR, Scott JC, Seeley DG, Steiger P,Vogt TM, and the Study of Osteoporotic Fractures ResearchGroup 1990 Appendicular bone density and age predict hipfracture in women. The Study of Osteoporotic Fractures Re-search Group. JAMA263:665–668.

27. Cummings SR, Black DM, Nevitt MC, Browner W, Cauley J,Ensrud K, Genant HK, Palermo L, Scott J, Vogt TM 1993Bone density at various sites for prediction of hip fractures.The Study of Osteoporotic Fractures Research Group. Lancet341:72–75.

28. Hui SL, Slemenda CW, Johnston CC Jr 1989 Baseline mea-surement of bone mass predicts fracture in white women. AnnIntern Med111:355–361.

29. Lester GE, Anderson JJ, Tylavsky FA, Sutton WR, StinnettSS, DeMasi RA, Talmage RV 1990 Update on the use of distalradial bone density measurements in prediction of hip andColles’ fracture risk. J Orthop Res8:220–226.

30. Melton LJD, Atkinson EJ, O’Fallon WM, Wahner HW, RiggsBL 1993 Long-term fracture prediction by bone mineral as-sessed at different skeletal sites. J Bone Miner Res8:1227–1233.

31. Nordin BE, Chatterton BE, Walker CJ, Wishart J 1987 Therelation of forearm mineral density to peripheral fractures inpostmenopausal women. Med J Aust146:300–304.

32. Porter RW, Miller CG, Grainger D, Palmer SB 1990 Predic-tion of hip fracture in elderly women: A prospective study [seecomments]. BMJ301:638–641.

33. Wasnich RD, Ross PD, Davis JW, Vogel JM 1989 A compar-ison of single and multi-site BMC measurements for assess-ment of spine fracture probability. J Nucl Med30:1166–1171.

34. Wasnich RD, Ross PD, Heilbrun LK, Vogel JM 1987 Selec-tion of the optimal skeletal site for fracture risk prediction.Clin Orthop216:262–269.

2265METACARPAL CORTICAL AREA AND HIP FRACTURE

35. Kelsey JL, Browner WS, Seeley DG, Nevitt MC, CummingsSR 1992 Risk factors for fractures of the distal forearm andproximal humerus. The Study of Osteoporotic Fractures Re-search Group. Am J Epidemiol135:477–489.

36. Ross PD, Davis JW, Epstein RS, Wasnich RD 1991 Pre-existing fractures and bone mass predict vertebral fractureincidence in women. Ann Intern Med114:919–923.

37. Cleghorn DB, Polley KJ, Bellon MJ, Chatterton J, BaghurstPA, Nordin BE 1991 Fracture rates as a function of forearmmineral density in normal postmenopausal women: Retrospec-tive and prospective data. Calcif Tissue Int49:161–163.

38. Jergas M, San Valentin R, Black DM, Nevitt M, Palermo L,Genant HK, Cummings S 1995 Radiogrammetry of the meta-carpals predicts future hip fracture. A prospective study.J Bone Miner Res10:S1;S371.

39. Nevitt MC, Johnell O, Black DM, Ensrud K, Genant HK,Cummings SR 1994 Bone mineral density predicts non-

spine fractures in very elderly women. Osteoporos Int4:325–331.

40. Lotz JC, Cheal EJ, Hayes WC 1995 Stress distributions withinthe proximal femur during gait and falls: Implications forosteoporotic fracture. Osteoporos Int5:252–261.

Address reprint requests to:Douglas P. Kiel, M.D., M.P.H.

HRCA Research and Training Institute1200 Center Street

Boston, MA 02131, USA

Received in original form December 11, 2000; in revised formApril 23, 2001; accepted June 15, 2001.

2266 KIEL ET AL.