Fracture Risk in Patients with Parkinsonism: A Population-based Study in Olmsted County, Minnesota

Transcription

1 Fracture Risk in Patients with Parkinsonism: A Population-based Study in Olmsted County, Minnesota OLOF JOHNELL, L. JOSEPH MELTON, III, ELIZABETH J. ATKINSON, W. MICHAEL O'FALLON, LEONARD T. KURLAND Summary In this population-based retrospective cohort study, the Olmsted County, Minnesota residents first diagnosed with Parkinson's disease during were matched by age and sex to an equal number of control subjects from the community. Fractures were assessed through review of each subject's complete (inpatient and outpatient) medical records. At the time of diagnosis, County residents with parkinsonism were no more likely to have a history of selected fractures than control subjects (% in each group). Subsequently, % of cases and % of controls experienced one or more new fractures during 9 person-years of follow-up (p =.). The greatest increase in risk was seen for proximal femur fractures, confirming previous case-control studies. By years after diagnosis, an estimated 7% of the parkinsonism cohort had experienced a new hip fracture. The pattern of fractures that was observed suggested that the increased risk was due more to specific types of falls than to disuse osteoporosis. Introduction The link between Parkinson's disease and an increased risk of limb fractures has usually been attributed to a greater likelihood of falling [-]. Evidence for this association is based largely on the fact that parkinsonism is often noted among patients with hip fractures [], and series of patients with both conditions have been assembled to address certain clinical management problems [7-]. However, most patients with Parkinson's disease are quite old [] and fractures are common among the elderly in any event []. Thus, it cannot be established from these observations that parkinsonism patients are really over-represented among the victims of hip fracture. One population-based casecontrol study of hip fractures in Rochester, Minnesota suggested that parkinsonism was not significantly more common than expected in this group []. However, a more recent casecontrol study from the North-east United States documented a nine-fold increase in hip fracture risk among patients with parkinsonism [], while a second, larger study in Malmo, Sweden found that patients with signs of parkinsonism were five times more frequent among those with hip fractures than among a group of matched controls []. Patients with a particular condition, like parkinsonism, are relatively uncommon in any series of hip fractures, however. In the Swedish study, for example, signs of parkinsonism were seen in only.% of cases and.% of controls. In addition, estimates of effect in case-control studies may be subject to biases related to the selection of subjects. Consequently, we undertook a cohort study to quantify the risk of fractures among unselected Age and Ageing 99;:- Downloaded from at Pennsylvania State University on September 7,

2 FRACTURE RISK IN PATIENTS WITH PARKINSONISM parkinsonism patients from the community. Documentation that fractures are an important practical problem in patients with parkinsonism is needed before embarking on an extensive research programme to develop an intervention to reduce fractures in this complex clinical setting. Methods Olmsted County, Minnesota is well suited for studies of disease associations such as this because comprehensive medical records for the residents are available for review and because these records are accessible through a centralized index of diagnoses made by essentially all medical care providers used by the local population []. This medical records linkage system (the Rochester Epidemiology Project) permitted retrieval of all diagnoses of Parkinson's disease made among County residents as a result of inpatient or outpatient care, death certification or autopsy. Criteria for inclusion in the study included any one or more of the following: (a) diagnosis of parkinsonism by a qualified neurologist; (b) demonstration of all three major manifestations of parkinsonismresting tremor, bradykinesia, and rigidity; (c) demonstration of exaggerated glabellar reflex, reduced facial expression, and unilateral or significantly asymmetrical bradykinesia and rigidity; or (d) diagnosis from histological findings characteristic of idiopathic Parkinson's disease [7]. Patients who developed parkinsonian symptomatology while taking tranquilizing drugs, and in whom the symptoms persisted for months or longer while the patient continued the drugs, were classified as having drug-induced parkinsonism. In contrast, patients in whom the parkinsonian manifestations first appeared while taking dopamine antagonistic agents, and in whom the symptoms persisted for months or longer after discontinuing the neuroleptic agent, were classified as having idiopathic Parkinson's disease. Patients who developed a transient parkinsonian syndrome while taking neuroleptic drugs but who were not noted to have parkinsonian symptoms on subsequent evaluations were excluded. These criteria resulted in an incidence cohort of cases of parkinsonism newly diagnosed among Olmsted County residents in the period [7]. An age- and sex-matched control for each patient was selected from a list of all Olmsted County residents seen for medical care at the time the parkinsonism was diagnosed. Each year about threefourths of the local population is seen at one of the Clinic facilities, and almost all local residents will have at least one Mayo Clinic contact within any given -year period. Additional residents are seen by the other providers who do not attend Mayo Clinic. Consequently, the Rochester Epidemiology Project medical records linkage system provides an enumeration of the population from which samples can be drawn and, since a large proportion of the population will have at least one medical contact over a period of time, the controls approximate a sample of the general population matched by age and sex to the index cases. Controls were selected from the same institution that diagnosed each index case and from registration numbers closest to that of the index patient. Because registration numbers are unique to each individual and are assigned at the initial visit for each patient, this approximately matches for the duration of documented clinical history prior to the index date (the date of diagnosis for cases and, for controls, the diagnosis date of the matched case). Cases and controls were then followed forward in time through their linked medical records in the community (retrospective cohort study) until death or the most recent clinical contact. For each subject, all inpatient and outpatient medical records at any local provider of health care were searched for the occurrence of specific fractures. Mayo Clinic records, for example, contain the details of every inpatient hospitalization at its two large affiliated hospitals (St Mary's and Rochester Methodist), every outpatient or office visit at the Clinic, the emergency rooms, nursing homes or private homes, as well as radiographic reports and pathology reports, including autopsies. Fractures at eight selected sites were recorded whether they occurred before or after the index date. The records contained the clinical history and the radiologist's report of each fracture, but the original radiographs were not available for review. The diagnosis of vertebral fracture was accepted on the basis of a radiologist's report of compression or collapse of one or more thoracic or lumbar vertebrae. Ascertainment of the fractures of interest is believed to be complete except for vertebral fractures, some of which are never diagnosed. The influence of parkinsonism on fracture incidence was evaluated using four basic methods of analysis. First, the frequency of fractures in the parkinsonism cases was compared directly with that in the control subjects. The duration of observation prior to the index date was approximately equal for each matched case and control pair. Similarly, for this analysis, each member of the case/control pair was followed from the index date to the earlier of the two dates of last follow-up. This censoring assured that the interval of follow-up for the two members of each pair was identical. The proportions of cases and controls with fractures prior to the index date or after the index were compared using the McNemar test. Downloaded from at Pennsylvania State University on September 7,

3 O. JOHNELL ET AL. Fracture prevalence per was calculated as of the index date. Fracture incidence per person-years of observation was then calculated for cases and controls from the index date to the earliest date of last follow-up. The ratios of fracture prevalence rates and fracture incidence rates determined the relative risk of fractures for those with parkinsonism compared to control subjects. Because of the censoring, the denominators of these rates are the same so only the numerators are given in the tables. Ninety-five per cent confidence intervals (9% CI) around the relative risks were calculated assuming that the number of fractures per a fixed number of personyears follows a Poisson distribution; the ratio of the fracture incidences follows an F-distribution []. We also calculated standardized morbidity ratios (SMR) separately for parkinsonism cases and controls, comparing the number of fractures observed at selected skeletal sites (based on the first fracture of a given type per person) to the number expected over the period of observation. Expected numbers were derived by applying age- and sex-specific incidence rates from the general population for these fractures to the age- and sex-specific person-years of follow-up in the case and control cohorts. Incidence rates from the general population of Rochester were available for fractures of the proximal femur [9], distal forearm [], proximal humerus [], pelvis [] and vertebrae []. Ninety-five per cent confidence intervals for the SMRs were calculated assuming that the expected rates are fixed and the observed fractures follow a Poisson distribution. In the third method of analysis, the cumulative incidence of new fractures was projected for up to years following the index date, using product-limit life-table methods []. Cumulative incidence curves for cases and controls were compared using the logrank test statistic []. The same methods were used to assess survival, with expected death rates derived from West North Central United States whites in 97. Finally, the proportional hazards model [] was used to determine the relative influence of various clinical characteristics on subsequent fracture risk among those with Parkinson's disease. The dependent variable was time until the first new fracture, and the independent variables were the clinical characteristics. Results At the time of diagnosis of Parkinson's disease or the comparable date for control subjects (mean age, 7. years for parkinsonism patients and 7. years for control subjects, range to 9 years for both groups), % of patients and an identical % of control subjects had experienced one or more of the fractures of interest during an average interval covered by medical record documentation of 7. years (median,.7 years). Prior to the index date, fractures were about equally common in the two groups at most skeletal sites (Table I), with the exception of vertebral fractures (p =.) and proximal humerus fractures (p =.) in women with parkinsonism. Among men with parkinsonism, vertebral fractures were less common than expected, but this did not quite reach statistical significance (p =.7). None of the fractures in either patient group occurred within days of the index date. Altogether, there were pairs (cases and controls) of men and 7 pairs of women. Following the index date, the patients with parkinsonism were followed for 9 person-years (mean,. years per patient; median,. years), while the control subjects were followed for person-years (mean,. years per subject; median,. years). During this period of observation, fractures were significantly more common among parkinsonism patients than control subjects (p =.) when both sexes were combined (Figure ), despite the fact that survival was much reduced among the patients with parkinsonism (only % survived years compared to an expected 9% among members of the general population of like age, sex and race). Fractures were also significantly more common in patients with Parkinson's disease when men and women were considered separately (Figure ). By years after the index date, an actuarially estimated % of men and % of women with parkinsonism had experienced one or more of these fractures (9% for both sexes combined) compared with % and % of male and female control subjects, respectively (% for both sexes combined). When follow-up was censored to be the same for cases and controls, there were 9 personyears of observation in each group (mean,. years per subject, median,. years). Forty-six (%) of the patients with parkinsonism and 7 (%) of the control subjects experienced one or more of the fractures of interest (p =.) Downloaded from at Pennsylvania State University on September 7,

4 FRACTURE RISK IN PATIENTS WITH PARKINSONISM Table I. Number of fractures observed at selected sites on or before the index date and after the index date among Olmsted County, Minnesota, residents with Parkinson's disease (PD) and control subjects On or before index date After index date Fracture site PD Controls RP 9% CI PD Controls RR 9% CI Men Ribs Ankle Proximal tibia Total Women Ribs Ankle Proximal tibia Total Both sexes combined Ribs Ankle Proximal tibia Total > > >.->.->.- -> -> >.->.-. -> >.->.-. ->.-. RP = relative prevalence and RR = relative risk in the two groups; CI = 9% confidence interval around the ratio. during this comparable follow-up. When cases and controls were compared directly (Table I), there was a statistically significantly increased relative risk of fractures in women, in men, and in both sexes combined, with especially high relative risks for proximal femur fractures (p <. considering both sexes combined). The fracture experience of parkinsonism cases and controls was also compared with that expected from the general population. In these analyses, patients with an earlier history of each specific fracture were deleted from the analysis for that fracture site; and, among those free of such a history, the number of first fractures of each type was compared with that expected. As shown in Table II, the observed numbers of fractures of each type were consistently greater than expected among the patients with parkin- Downloaded from at Pennsylvania State University on September 7,

5 O. JOHNELL ET AL. Ymn fotartig Indax (tat* Figure. Cumulative incidence of any new fracture after the index date among Olmsted County, Minnesota residents with parkinsonism and controls. sonism, and four of the five SMRs were statistically significantly elevated. None of the SMRs for controls was significantly different from the expected value of. except for fractures of the vertebrae. After deleting the four cases (one of whom had had two hip fractures) and five controls who had had a hip fracture prior to the index date, the cumulative incidence of a new proximal femur fracture by years after the diagnosis of parkinsonism was 7% (9% in men and % in women), compared to only 9% (% in men and % in women) in controls (p <.). As shown in Figure, hip fractures increased steadily following the diagnosis of parkinsonism. In a multivariate analysis, only age was an independent predictor of time to the first hip fracture, while age and sex were predictors of time to any of the fractures studied. YsaI * to****!) tnda * <*" Figure. Cumulative incidence of hip fractures after the index date among Olmsted County, Minnesota residents with parkinsonism and controls. Discussion The risk of fracture following the diagnosis of parkinsonism was quite high in this populationbased study, but most patients were elderly (% were years old or over), and such individuals might be expected to experience fractures frequently on the basis of age-related pathology besides Parkinson's disease. Indeed, fractures were common among the age- and sex-matched controls. However, there appeared to be a clear excess of hip fractures among those with parkinsonism that was not evident prior to diagnosis. The -fold increase in risk might have been exaggerated by the dearth of hip fractures among the controls; if they had experienced the. fractures expected on the basis of population hip fracture incidence rates instead of the one hip fracture actually Table II. Observed (Obs) fractures after the index date at selected sites in comparison with expected numbers (Exp) and standardized morbidity ratios (SMRs) among Olmsted County, Minnesota residents with Parkinson's disease and control subjects Parkinson's disease Control Downloaded from at Pennsylvania State University on September 7, Fracture site Obs Exp SMR 9% CI Obs Exp SMR 9% CI

6 FRACTURE RISK IN PATIENTS WITH PARKINSONISM 7 observed, the relative risk estimate would have been.. This is similar to the five-fold increase in risk seen in the Swedish hip fracture casecontrol study described in the Introduction []. Fractures at other sites were not increased to such an extent and, overall, the relative risk was doubled. Because fractures in the elderly are due to the interplay of bone fragility and trauma [, 7], these excess fractures might have resulted from a greater tendency to fall or from disuse osteoporosis caused by decreased activity related to Parkinson's disease. As noted in the Introduction, patients with Parkinson's disease seem predisposed to falling [, ], but falls are very common among elderly individuals generally; it has been estimated that a third of --year-olds fall each year [7]. However, parkinsonism patients are particularly over-represented among those with multiple falls [, ]. The mechanism of this effect is unclear and may vary from case to case. Possibilities include gait abnormalities that increase the likelihood of falling or diminished reflexes that reduce the ability to recover once a fall has begun []. In one study, parkinsonism was most prominent (four of nine patients) when repeated falls were due to tripping or slipping []. fractures are also related to falls [], but they were not particularly excessive among the patients with Parkinson's disease. This apparent anomaly might be explained on the basis of gait speed. Elderly subjects with rapid gait are more likely to fall forward and experience a Colles' fracture, while those with a shuffling gait tend to fall backward or to the side and suffer a hip fracture [9]. The increased risk of proximal humerus, pelvis, and hip fractures is consistent with the notion that patients with parkinsonism experience fractures as a result of falling backward and to the side. Alternatively, the increased risk of fractures might be explained on the basis of excess bone loss due to secondary osteoporosis, related either to the parkinsonism or to one of the other conditions commonly present []. However, we are unaware of any data concerning bone mass in patients with parkinsonism; and vertebral fractures, which are particularly closely associated with osteoporosis [, ], were not more common in cases than in the age- and sexmatched controls. Rib fractures were increased and have also been related to osteoporosis, but they are common among those predisposed to falls as well []. The association between Parkinson's disease and hip fracture might have been even stronger were it not for the increased mortality. As have others [], we found reduced survival among patients with parkinsonism. Because hip fracture rates increase exponentially with age [], this shortened survival reduces the number of fractures that would otherwise be observed. Interventions that might lower the risk of fractures still further are not well developed. Recent calculations by Hayes and coworkers demonstrate that falls directly on the hip generate forces capable of fracturing most femurs []. Unless practical hip protectors can be developed that solve this problem biomechanically more emphasis will have to be placed on reducing falls among those with Parkinson's disease. Some approaches to this problem have been proposed [], but no data on their efficacy have been reported. This is a subject worthy of further exploration by those responsible for the care of patients with this condition. Acknowledgements The authors would like to thank Mrs Judy Bruen and Mrs Janet Deaner for their help in data collection, Mrs Ruth H. Cha for computer programming and Mrs Yvonne Weeldreyer for assistance in preparing the manuscript. This project was supported in part by grants AG- 7 and AR-OS from the National Institutes of Health, United States Public Health Service. References. Livesley B, Atkinson L. Repeated falls in the elderly. Mod Geriatr 97;:-7.. Klawans HL, Topel JL. Parkinsonism as a falling sickness. JAMA 97;:l -7.. Aita JF. Why patients with Parkinson's disease fa\\. JAMA 9;7:-.. Alffram P-A. An epidemiologic study of cervical and trochanteric fractures of the femur in an urban population: analysis of cases with special reference to etiologic factors. Ada Orthop Scand 9;(suppl):l-9.. Alhava EM. Diseases contributing to fragility of Downloaded from at Pennsylvania State University on September 7,

7 O. JOHNELL ET AL. bone in patients with hip fractures. Ann Clin Res 97;:-.. Christodoulou NA, Dretakis EK. Significance of muscular disturbances in the localization of fractures of the proximal femur. Clin Orthop 9;7: Coughlin L, Templeton J. Hip fractures in patients with Parkinson's disease. Clin Orthop 9;:9-.. Eventov I, Moreno M, Geller E, Tardiman R, Salama R. Hip fractures in patients with Parkinson's syndrome. J Trauma 9;: Staeheli JW, Frassica FJ, Sim FH. Prosthetic replacement of the femoral head for fracture of the femoral neck in patients who have Parkinson disease. J Bone Joint Surg [Am] 9;7:-.. Turcotte R, Godin C, Duchesne R, Jodoin A. Hip fractures and Parkinson's disease: a clinical review of 9 fractures treated surgically. Clin Orthop 99;:-.. Kurtzke JF, Kurland LT. The epidemiology of neurologic disease. In: Baker AB, ed. Clinical neurology. Philadelphia: Harper & Row, 9;-.. Melton LJ III. Epidemiology of fractures. In: Riggs BL, Melton LJ III, eds. Osteoporosis: etiology, diagnosis, and management. New York: Raven Press, 9; -.. Melton LJ III, Riggs BL. Clinical spectrum. In: Riggs BL, Melton LJ III, eds. Osteoporosis: etiology, diagnosis, and management. New York: Raven Press, 9;-79.. Grisso JA, Kelsey JL, Strom BL, et al. Risk factors for falls as a cause of hip fracture in women. N EnglJ Med 99 ;:-.. Johnell O, Sernbo I. Health and social status in patients with hip fractures and controls. Age Ageing 9; :-9.. Kurland LT, Molgaard CA. The patient record in epidemiology. Sci Am 9;:-. 7. Rajput AH, Offord KP, Beard CM, Kurland LT. Epidemiology of parkinsonism: incidence, classification, and mortality. Ann Neurol 9;:7-.. Cox DR. Some simple approximate tests for Poisson variates. Biometrika 9;O:-. 9. Melton LJ III, O'Fallon WM, Riggs BL. Secular trends in the incidence of hip fractures. Calcif Tissue Int 97;:7-.. Owen RA, Melton LJ, Johnson KA, Ilstrup DM, Riggs BL. Incidence of Colles' fracture in a North American community. Am J Public Health 9;7:O-7.. Rose SH, Melton LJ III, Morrey BF, Ilstrup DM, Riggs BL. Epidemiologic features of humeral fractures. Clin Orthop 9; :-.. Melton LJ III, Sampson JM, Morrey BF, Ilstrup DM. Epidemiologic features of pelvic fractures. Clin Orthop 9,:-7.. Cooper C, Atkinson EJ, O'Fallon WM, Melton LJ. The epidemiology of vertebral fractures: a population-based study in Rochester, Minnesota, J Bone Min Res (In press).. Kaplan EL, Meier P. Non-parametric estimation from incomplete observations. J Am Stat ^«ocl9;:7-.. Kalbfleisch JD, Prentice RL. The statistical analysis of failure time data. New York: J Wiley &Sons, 9;l-.. Cox DR. Regression models and life-tables (with discussion). J R Stat Soc [B] 97;: Cummings SR, Nevitt MC. Epidemiology of hip fractures and falls. In: Kleerekoper M, Krane SM, eds. Clinical disorders of bone and mineral metabolism. New York: Mary Ann Liebert, Inc., 99;-.. Nevitt MC, Cummings SR, Kidd S, Black D. Risk factors for recurrent nonsyncopal falls: a prospective study. JAMA 99;:-. 9. Cummings SR, Nevitt MC. A hypothesis: the causes of hip fractures. J Gerontol 99;:M7-ll.. Melton LJ III, Kan SH, Frye MA, Wahner HW, O'Fallon WM, Riggs BL. Epidemiology of vertebral fractures in women. Am J Epidemiol 99;9:-.. Gardsell P, Johnell O, Nilsson BE. Predicting fractures in women by using forearm bone densitometry. Calcif Tissue Int 99;:^.. Keso L, Kivisaari A, Salaspuro M. Fractures on chest radiographs in detection of alcoholism. Alcohol Alcohol 9;:-.. Marttila RJ. Epidemiology. In: Koller WC, ed. Handbook of Parkinson's disease. Basel/New York: Marcel Dekker, Inc. 97;-.. Hayes WC, Piazza SJ, Zysset PK. Biomechanics of fracture risk prediction of the hip and spine by quantitative computed tomography. Radiol Clin North Am 99;9:-.. Tinetti ME, Speechley M. Prevention of falls among the elderly. N Engl J Med 99;:-9. Authors' address Department of Health Sciences Research, Mayo Clinic and Foundation, First Street S.W., Rochester, MN 9, USA Address correspondence to Dr L. J. Melton. Present address: O. Johnell, Department of Orthopaedic Surgery, The General Hospital, S- Malmo, Sweden Received June 99 Downloaded from at Pennsylvania State University on September 7,