Douglas C. Bauer

Randomised trial of effect of alendronate on risk of fracture in women with existing vertebral fractures

BACKGROUND Previous studies have shown that alendronate can increase bone mineral density (BMD) and prevent radiographically defined (morphometric) vertebral fractures. The Fracture Intervention Trial aimed to investigate the effect of alendronate on the risk of morphometric as well as clinically evident fractures in postmenopausal women with low bone mass. METHODS Women aged 55-81 with low femoral-neck BMD were enrolled in two study groups based on presence or absence of an existing vertebral fracture. Results for women with at least one vertebral fracture at baseline are reported here. 2027 women were randomly assigned placebo (1005) or alendronate (1022) and followed up for 36 months. The dose of alendronate (initially 5 mg daily) was increased (to 10 mg daily) at 24 months, with maintenance of the double blind. Lateral spine radiography was done at baseline and at 24 and 36 months. New vertebral fractures, the primary endpoint, were defined by morphometry as a decrease of 20% (and at least 4 mm) in at least one vertebral height between the baseline and latest follow-up radiograph. Non-spine clinical fractures were confirmed by radiographic reports. New symptomatic vertebral fractures were based on self-report and confirmed by radiography. FINDINGS Follow-up radiographs were obtained for 1946 women (98% of surviving participants). 78 (8.0%) of women in the alendronate group had one or more new morphometric vertebral fractures compared with 145 (15.0%) in the placebo group (relative risk 0.53 [95% Cl 0.41-0.68]). For clinically apparent vertebral fractures, the corresponding numbers were 23 (2.3%) alendronate and 50 (5.0%) placebo (relative hazard 0.45 [0.27-0.72]). The risk of any clinical fracture, the main secondary endpoint, was lower in the alendronate than in the placebo group (139 [13.6%] vs 183 [18.2%]; relative hazard 0.72 [0.58-0.90]). The relative hazards for hip fracture and wrist fracture for alendronate versus placebo were 0.49 (0.23-0.99) and 0.52 (0.31-0.87). There was no significant difference between the groups in numbers of adverse experiences, including upper-gastrointestinal disorders. INTERPRETATION We conclude that among women with low bone mass and existing vertebral fractures, alendronate is well tolerated and substantially reduces the frequency of morphometric and clinical vertebral fractures, as well as other clinical fractures.

Elevated Aortic Pulse Wave Velocity, a Marker of Arterial Stiffness, Predicts Cardiovascular Events in Well-Functioning Older Adults

Background—Aging results in vascular stiffening and an increase in the velocity of the pressure wave as it travels down the aorta. Increased aortic pulse wave velocity (aPWV) has been associated with mortality in clinical but not general populations. The objective of this investigation was to determine whether aPWV is associated with total and cardiovascular (CV) mortality and CV events in a community-dwelling sample of older adults. Methods and Results—aPWV was measured at baseline in 2488 participants from the Health, Aging and Body Composition (Health ABC) study. Vital status, cause of death and coronary heart disease (CHD), stroke, and congestive heart failure were determined from medical records. Over 4.6 years, 265 deaths occurred, 111 as a result of cardiovascular causes. There were 341 CHD events, 94 stroke events, and 181 cases of congestive heart failure. Results are presented by quartiles because of a threshold effect between the first and second aPWV quartiles. Higher aPWV was associated with both total mortality (relative risk, 1.5, 1.6, and 1.7 for aPWV quartiles 2, 3, and 4 versus 1; P=0.019) and cardiovascular mortality (relative risk, 2.1, 3.0, and 2.3 for quartiles 2, 3, and 4 versus 1; P=0.004). aPWV quartile was also significantly associated with CHD (P=0.007) and stroke (P=0.001). These associations remained after adjustment for age, gender, race, systolic blood pressure, known CV disease, and other variables related to events. Conclusions—Among generally healthy, community-dwelling older adults, aPWV, a marker of arterial stiffness, is associated with higher CV mortality, CHD, and stroke.

Prognostic Value of Usual Gait Speed in Well-Functioning Older People—Results from the Health, Aging and Body Composition Study

Objectives: To define clinically relevant cutpoints for usual gait speed and to investigate their predictive value for health‐related events in older persons.

Endogenous Hormones and the Risk of Hip and Vertebral Fractures among Older Women

BACKGROUND AND METHODS In postmenopausal women, the serum concentrations of endogenous sex hormones and vitamin D might influence the risk of hip and vertebral fractures. In a study of a cohort of women 65 years of age or older, we compared the serum hormone concentrations at base line in 133 women who subsequently had hip fractures and 138 women who subsequently had vertebral fractures with those in randomly selected control women from the same cohort. Women who were taking estrogen were excluded. The results were adjusted for age and weight. RESULTS The women with undetectable serum estradiol concentrations (<5 pg per milliliter [18 pmol per liter]) had a relative risk of 2.5 for subsequent hip fracture (95 percent confidence interval, 1.4 to 4.6) and subsequent vertebral fracture (95 percent confidence interval, 1.4 to 4.2), as compared with the women with detectable serum estradiol concentrations. Serum concentrations of sex hormone-binding globulin that were 1.0 microg per deciliter (34.7 nmol per liter) or higher were associated with a relative risk of 2.0 for hip fracture (95 percent confidence interval, 1.1 to 3.9) and 2.3 for vertebral fracture (95 percent confidence interval, 1.2 to 4.4). Women with both undetectable serum estradiol concentrations and serum sex hormone-binding globulin concentrations of 1 microg per deciliter or more had a relative risk of 6.9 for hip fracture (95 percent confidence interval, 1.5 to 32.0) and 7.9 for vertebral fracture (95 percent confidence interval, 2.2 to 28.0). For those with low serum 1,25-dihydroxyvitamin D concentrations (< or =23 pg per milliliter [55 pmol per liter]), the risk of hip fracture increased by a factor of 2.1 (95 percent confidence interval, 1.2 to 3.5). CONCLUSIONS Postmenopausal women with undetectable serum estradiol concentrations and high serum concentrations of sex hormone-binding globulin have an increased risk of hip and vertebral fracture.

Factors associated with appendicular bone mass in older women

Fractures are an important cause of disability in elderly women, especially among those with osteopenia. Decreased bone mass increases the risk for most types of fractures [1-4]. In particular, we recently reported [5] that most fractures in elderly women are associated with low bone mass at appendicular sites (the radius and calcaneus). The risk for these fractures, including some fractures not previously linked to osteopenia, such as of the leg and hand, increased 40% to 80% for each standard deviation reduction in appendicular bone mass [5]. We also found that appendicular bone mass predicts the risk for all nonspine fractures as well as for bone mass measured at the hip and spine [6]. Despite the strong relation between appendicular bone mass and risk for fracture, little consensus exists about the many proposed risk factors for decreased bone mass [7, 8], perhaps because most studies have examined a limited number of risk factors in relatively small selected cohorts. To determine what factors might contribute to osteopenia in older women, we examined a wide variety of potential correlates of appendicular bone mass in the Study of Osteoporotic Fractures, a large multicenter, community-based study of elderly women. Methods Patients From September 1986 to October 1988, women who were at least 65 years old were recruited in four areas of the United States: Portland, Oregon; Minneapolis, Minnesota; Baltimore, Maryland; and the Monongahela Valley near Pittsburgh, Pennsylvania. Age-eligible women were identified from membership lists from several sources, as previously reported [1]. We excluded black women because of the reduced incidence of hip fractures in this group [9], and we excluded women who were unable to walk without the assistance of another person or who had bilateral hip replacements. Measurement of Bone Mass Bone mass (in g/cm2) was measured using OsteoAnalyzers (Siemens-Osteon, Wahiawa, Hawaii). We scanned three sites: the distal radius, the midradius, and the calcaneus. The protocol for the bone mass measurements has been described elsewhere [1]. To describe the reproducibility of these measurements, a participant in the study was measured five times on 5 consecutive days at each of the four clinical centers. The average coefficients of variation in these older women (standard deviation/mean for each subject) were 1.5% for the distal radius, 2.0% for the midradius, and 1.3% for the calcaneus. To assess variations between scanners at the four centers, two investigators were measured by all four machines; the mean of their coefficients of variation in these younger subjects was 0.4% for the distal radius, 0.5% for the midradius, and 1.2% for the calcaneus [10]. The correlation coefficients between the measurement sites were distal and midradius, 0.75; calcaneus and distal radius, 0.66; and calcaneus and midradius, 0.63. Predictor Variables Participants completed a self-administered questionnaire and were interviewed and examined at the clinical center. A selected medical history was obtained, including a history of a physician diagnosis of osteoporosis, spine fracture, arthritis, gastric surgery, hyperthyroidism, and stroke. Reproductive history, including age of last menstrual period, genitourinary surgery (hysterectomy, oophorectomy), number of pregnancies, and breast-feeding, was recorded. Previous fractures in subjects and their parents or sisters were noted. We collected detailed records about specific health habits including lifetime smoking history, alcohol use, and caffeine intake. Women who had smoked less than 100 cigarettes in their lifetime were considered nonsmokers. We assumed that a cup of coffee contained 95 mg of caffeine and that tea and cola drinks contained 55 mg and 45 mg, respectively. Participants were asked to bring all prescription and nonprescription medications with them to the interview for verification. The dose and duration of use of sex hormones, diuretics, corticosteroids, thyroid supplements, aluminum-containing antacids, vitamin D, sedative hypnotics, and antiepileptic medications were obtained. The frequency and duration of use of calcium supplements were recorded; use of TUMS (which can be taken as an antacid or calcium supplement) was recorded separately. Responses to the questionnaire were checked by a trained interviewer and were verified against the labels of medications that the subject brought to the interview. Recent dietary history (past 12 months), particularly calcium, phosphorus, and protein intake, was assessed by a checklist-interview method developed from the HANES-II survey [11]. Food models were used to estimate portion sizes, and foods that account for 80% of calcium intake in most adults were included. This instrument has a correlation of 0.76 with calcium intake assessed by a 7-day diet diary, but it tends to underestimate calcium intake by approximately 150 mg/day [12]. The frequency of milk consumption as a child and young adult was also assessed by questionnaire. Physical activity was examined with a modified Paffenbarger survey that has been validated in postmenopausal women [13-16]. The type and duration of weight-bearing recreational activities from the previous 12 months were recorded, and these were converted into weekly caloric expenditure. Other current activities, such as stair climbing, walking, and heavy household chores, were included. Intensity-weighted measures were calculated by designating activities as low (for example, walking or gardening), medium (dancing or tennis), or high intensity (jogging or skiing) and multiplying the reported frequency of the activity by 2.5, 5, and 7.5, respectively. Weight (in light indoor clothes with shoes removed) was recorded with a balance beam scale, and height was measured using a standard held-expiration technique with a wall-mounted Harpenden stadiometer [17]. Maximal right knee extension, triceps (arm extension), and hip abduction torque strength were measured with a hand-held isometric dynamometer (Sparks Instruments and Academics, Coralville, Iowa). Grip strength of the right and left hand was assessed as the average of two attempts with the dynamometer. Waist, hip, and abdominal circumferences were measured using standard methods [18]. Statistical Methods The cohort was randomly divided into equal-sized (n = 4852) training and validation groups. In the training group, we analyzed potential associations with bone mass in univariate analyses, adjusted for age. We then adjusted associations (P < 0.05) for other plausible confounding effects. For example, associations between the use of thiazide diuretics and bone mass were also adjusted for body weight, because those who were taking diuretics tended to be heavier. Potentially nonlinear associations between bone mass and continuous variables, such as calcium intake, were examined by plotting bone mass against the median of each decile of the predictor variable. Associations were tested for statistical significance with simple linear regression, Student t-test, or analysis of variance. Variables that were associated with bone mass in univariate analyses (P < 0.05) were examined by multivariate analyses using PROC GLM (SAS Institute, Inc., Cary, North Carolina). Some categories of predictor variables, such as calcium intake and muscle strength, contained several variables associated with bone mass in univariate analyses. After examining these variables for multicollinearity, separate models for strength, family history of fractures, dietary calcium, and physical activity were analyzed to determine the individual variables that explained most of the variance of bone mass within each category. For example, grip strength explained most of the variance for bone mass related to strength; thus, we selected this variable as the measure of strength for subsequent multivariate analyses. Both current dietary calcium and calcium from milk (ingested between ages 18 to 50 years) contributed to bone mass variance; therefore, both were included in the final multivariate models. A history of any maternal fracture after age 50 accounted for most of the bone mass variance from family history, and intensity-weighted lifetime physical activity accounted for most of the variance from physical activity. Selected variables (see Table 2) were then entered into a single multivariate model. We found that results from multivariate models were very similar in both training and validation groups; thus results are reported for the entire (combined) cohort. Results were generally similar for the three bone mass sites, and for those analyses that were concordant at all three sites, only the results for the distal radius are presented. We presented results at the other two sites only when they differed from the distal radius. Results The average age at enrollment was 71.1 years, and the age distributions were similar in the training and validation groups (Table 1). Thirteen percent of patients reported a previous wrist or hip fracture. Table 1. Baseline Characteristics of Patients Demographic and Anthropometric Data Bone mass was strongly and inversely associated with age, decreasing by approximately 5% with every additional 5 years of age after 65 years (Table 2). Northern European ancestry, hair color, and educational level were not associated with bone mass (Table 3). Table 2. Univariate and Multivariate Correlates of Distal Radius Bone Mass Table 3. Variables Not Associated with Distal Radius Bone Mass, Age-Adjusted Univariate Analyses* Several anthropometric measurements were strongly associated with bone mass (see Table 2). Both weight [in kilograms] and obesity (as Quetelet index, in kilograms per square meter) were associated with increased bone mass: after adjusting for age, bone mass increased 5.0% for every 10-kg increase in weight (Figure 1). Although weight and Quetelet were highly correlated (r = 0.91), weight was more strongly associated with bone mass than

Bisphosphonates and Fractures of the Subtrochanteric or Diaphyseal Femur

BACKGROUND A number of recent case reports and series have identified a subgroup of atypical fractures of the femoral shaft associated with bisphosphonate use. A population-based study did not support this association. Such a relationship has not been examined in randomized trials. METHODS We performed secondary analyses using the results of three large, randomized bisphosphonate trials: the Fracture Intervention Trial (FIT), the FIT Long-Term Extension (FLEX) trial, and the Health Outcomes and Reduced Incidence with Zoledronic Acid Once Yearly (HORIZON) Pivotal Fracture Trial (PFT). We reviewed fracture records and radiographs (when available) from all hip and femur fractures to identify those below the lesser trochanter and above the distal metaphyseal flare (subtrochanteric and diaphyseal femur fractures) and to assess atypical features. We calculated the relative hazards for subtrochanteric and diaphyseal fractures for each study. RESULTS We reviewed 284 records for hip or femur fractures among 14,195 women in these trials. A total of 12 fractures in 10 patients were classified as occurring in the subtrochanteric or diaphyseal femur, a combined rate of 2.3 per 10,000 patient-years. As compared with placebo, the relative hazard was 1.03 (95% confidence interval [CI], 0.06 to 16.46) for alendronate use in the FIT trial, 1.50 (95% CI, 0.25 to 9.00) for zoledronic acid use in the HORIZON-PFT trial, and 1.33 (95% CI, 0.12 to 14.67) for continued alendronate use in the FLEX trial. Although increases in risk were not significant, confidence intervals were wide. CONCLUSIONS The occurrence of fracture of the subtrochanteric or diaphyseal femur was very rare, even among women who had been treated with bisphosphonates for as long as 10 years. There was no significant increase in risk associated with bisphosphonate use, but the study was underpowered for definitive conclusions.

Risk for fracture in women with low serum levels of thyroid-stimulating hormone

Osteoporosis and thyroid dysfunction are both common in older women; 8% to 13% of women older than 50 years of age have biochemical evidence of thyroid dysfunction (1, 2), and 30% are osteoporotic according to bone density criteria (3). Although osteoporotic fractures have long been associated with florid hyperthyroidism (4) and, more recently, with a history of hyperthyroidism in older women (5), the relationship between biochemical evidence of excess thyroid hormone and fracture risk is not known (6, 7). Indirect evidence suggests that excess thyroid hormone due to endogenous disease or exogenous overuse of thyroid hormone may be associated with detrimental effects on bone, even in asymptomatic persons. For example, several biochemical markers of bone turnover are elevated in women with excess thyroid hormone (8, 9). Findings from studies of the relationship between excess thyroid hormone and bone mass are conflicting (10-17). However, factors other than bone mass, such as neuromuscular function and bone quality, contribute to risk for fracture (5) and may be adversely affected by excess thyroid hormone. Results of previous small, retrospective studies of thyroid function and fractures have also been conflicting (18-21). To our knowledge, no large prospective studies have examined the relationship between excess thyroid hormone and subsequent fracture. In light of the conflicting information on bone mass and the paucity of studies with fracture as an end point, several experts have noted the need for longitudinal studies of thyroid function and fracture risk (22-26). To test the hypothesis that low levels of serum thyroid-stimulating hormone (TSH) increase the risk for hip, vertebral, and any nonspine fracture, we performed a prospective study of postmenopausal women enrolled in the Study of Osteoporotic Fractures. Methods Patients The Study of Osteoporotic Fractures is a prospective cohort study of risk factors for fracture among 9704 women (5). White women older than 65 years of age were recruited in 1986 to 1988 from population-based listings at four clinical centers (Portland, Oregon; Minneapolis, Minnesota; Pittsburgh, Pennsylvania; and Baltimore, Maryland). The institutional review boards of all four centers gave approval for this study involving human research subjects. Measurements Baseline Measurements Participants were interviewed and examined during the baseline visit. Detailed information about physician-diagnosed medical conditions and past medication use was collected, and trained interviewers confirmed current medication use by examination of pill bottles. Participants were asked specifically about self-rated health, previous physician diagnoses of hyperthyroidism or Graves disease, and previous use of thyroid hormone. In addition to standardized assessments of height and weight, bone mass of the calcaneus was determined by using single-photon absorptiometry (OsteoAnalyzer, Siemens-Osteon, Wahiawa, Hawaii) and lateral radiographs of the thoracic and lumbar spine were obtained (27, 28). Serum was collected from each participant and stored at 190 C. Approximately 2 years after the baseline visit, bone mineral density of the proximal femur was measured in 82% of the cohort by using dual-energy x-ray absorbtiometry (Hologic QDR 1000, Waltham, Massachusetts) (29). Levels of TSH were measured in archived sera obtained at baseline by using a highly sensitive, third-generation chemiluminescent assay (Endocrine Science, Calabasas, California). The normal range for this assay is 0.5 to 5.5 mU/L; the functional sensitivity (defined as the concentration at which the interassay coefficient of variation is 20%) is approximately 0.05 mIU/L (30). At TSH concentrations of 0.5 mIU/L, the intra-assay coefficient of variation is 4.7% and the interassay coefficient of variation is 6.3%. Thirty randomly selected specimens were blindly submitted for duplicate analysis; the correlation between these two TSH results was high (r = 0.95). Previous studies have shown that TSH levels are highly stable in frozen sera over prolonged periods (31, 32). Other studies have demonstrated that among ambulatory adults, TSH levels of 0.1 mIU/L or less are highly correlated with a diminished response to thyroid-releasing hormone stimulation (33) and are associated with an increased incidence of atrial fibrillation (34). Ascertainment of Incident Fractures After the baseline visit, women were contacted by mail every 4 months about the occurrence of fractures. Hip fractures were confirmed by review of the appropriate radiographs by a radiologist at the coordinating center; other nonspine fractures were confirmed by review of written radiology reports. Fractures resulting from excessive trauma (such as motor vehicle accidents) were excluded. Follow-up for fracture and vital status was more than 99% complete. Lateral spine radiographs were repeated in 7299 women (79% of surviving women) after a mean (SD) follow-up of 3.7 0.4 years, and 7238 pairs of radiographs were judged to be adequate for assessment of incident vertebral fractures. Women without follow-up radiographs were older and reported poorer health at baseline compared with those who had follow-up radiographs (35). Vertebral fractures were identified by using computer-assisted morphometric evaluation (36), and incident vertebral fractures were defined as a 20% or greater and 4 mm or greater reduction in anterior, mid-vertebral, or posterior vertebral height between the baseline and follow-up radiographs (37). The persons who assessed the radiographs had no knowledge of the participants medical history or TSH level. Selection of Case and Control Samples for Fracture Analyses Using an efficient case-cohort approach that maintains statistical power but avoids expensive biochemical measurements in the entire cohort (38-40), we randomly selected baseline serum samples from 148 women with hip fracture and 149 women with incident vertebral fracture after the baseline visit. We randomly selected 398 women from the original cohort, independent of fracture status, to be controls. This random sample, which we refer to as the subsample in this report, included 14 of the 148 women selected as incident hip fracture cases and 15 of the 149 women selected as incident vertebral fracture cases; these women were removed from the subsample and were analyzed as cases of hip and vertebral fracture, respectively. To create a fracture-free control group, we excluded women from the subsample with other nonspine fractures during follow-up (n = 80 for the hip fracture analyses and n = 58 for the vertebral fracture analyses). Ninety women in the subsample had missing or technically inadequate radiographs and could not be analyzed for vertebral fracture outcomes. The analyses of any nonspine fracture were performed in the randomly selected subsample by using standard prospective cohort methods. After 14 women with unconfirmed fracture, 5 women with fracture from extreme trauma, and 6 women with spine fractures were excluded, the analysis of nonspine fracture included 100 women with documented nonspine fracture occurring after study entry and 273 without fracture. Random selection was done by using a computerized random-number generator. Statistical Analysis Continuous variables were plotted, and distributions, means, and standard deviations were examined. Levels of TSH were categorized as low ( 0.1 mIU/L), borderline low (>0.1 but <0.5 mIU/L), normal (0.5 to 5.5 mIU/L), or high (>5.5 mIU/L). Associations with hip fracture were examined by using proportional hazards analyses (Epicure, Hirosoft International, Seattle, Washington) that took into account the case-cohort sampling design. The proportionality assumption was not violated. Results are reported as relative hazards with 95% CIs. Logistic regression was used to analyze incident vertebral fracture; these results are reported as odds ratios with 95% CIs. Cox proportional-hazards models were used to determine associations with nonspine fracture among the randomly selected subsample. Multivariate models were constructed to adjust for potential confounders. Potential confounders were selected on the basis of biologic plausibility (for example, use of thyroid hormone) or a strong univariate association (P 0.1) with TSH level (for example, age) or fracture (for example, estrogen use). We found no association (P>0.1) between TSH level and maternal history of fracture, height, neuromuscular function, or corticosteroid use, which are known to be associated with fracture in this cohort (5). The final multivariate models for each fracture type included TSH level, age, previous hyperthyroidism, self-rated health, and current use of thyroid hormone and estrogen. To determine whether the increased risk for fracture in women with low TSH levels was mediated by reduced bone mass or some other mechanism, we examined the effect of further adjusting the final multivariate models for calcaneal bone mass measured at the baseline visit. The effect of adjustment for bone mineral density at the femoral neck, measured approximately 2 years after the baseline visit, was similar to that observed for calcaneal bone mineral density. Role of the Funding Source The funding source had no role in the collection, analysis, interpretation, or publication of these data. Results During a maximum follow-up of 5.9 years, 332 women had a first hip fracture, 389 had an incident vertebral fracture detected on paired spinal radiographs, and 2520 had nonspine fractures. Women who had incident hip, vertebral, or any nonspine fractures were older and had lower bone mass than controls (Table 1). Women with hip fractures were more likely to report previous hyperthyroidism. Mean TSH levels were similar among women with and without fracture, but the proportion of women with a low TSH level ( 0.1 mIU/L) was significantly greater among those with hip or vertebral fracture. Overall, 11% of participants

Central Nervous System–Active Medications and Risk for Falls in Older Women

OBJECTIVES: To determine whether current use of central nervous system (CNS)‐active medications, including benzodiazepines, antidepressants, anticonvulsants, and narcotics, increases the risk for subsequent falls.

Meta-analysis: Subclinical Thyroid Dysfunction and the Risk for Coronary Heart Disease and Mortality

Context Is subclinical thyroid dysfunction associated with increased risk for coronary heart disease and mortality? Contribution This systematic review of 12 prospective cohort studies found that both subclinical hypothyroidism and hyperthyroidism were possibly associated with a small increased risk for coronary heart disease and mortality. Caution Data were uncertain. Confidence intervals around risk estimates were wide, particularly for those related to subclinical hyperthyroidism. Higher-quality studies showed lower estimates of risk than lower-quality studies. Implication Randomized trials testing the efficacy of thyroxine replacement and antithyroid medications for subclinical hypothyroidism and subclinical hyperthyroidism are needed. The Editors Subclinical thyroid dysfunction refers to patients who have an abnormal thyrotropin (thyroid-stimulating hormone [TSH]) level and a normal free thyroxine (T4) level (1). The prevalence of subclinical hypothyroidism is about 4.3% in adults (0.7% for subclinical hyperthyroidism), and prevalence is higher in older adults and women (25). Controversy persists about whether screening and treating subclinical thyroid dysfunction is warranted (1, 57) because current evidence about the risks is limited (1, 5) and randomized, controlled trials on relevant clinical outcomes have not been done (5, 8). Subclinical hypothyroidism has been associated with elevated cholesterol levels (911) and increased risk for atherosclerosis (12, 13). Yet, data on the relationship between subclinical hypothyroidism and coronary heart disease (CHD) events are conflicting (12, 1417). In a previous meta-analysis (18), we found that subclinical hypothyroidism was associated with a 1.65-fold increased risk (CI, 1.28 to 2.12) for CHD. However, that meta-analysis included several cross-sectional and casecontrol studies and only 5 small prospective studies. Recently, 3 large prospective studies on this issue have been published (14, 16, 17), with somewhat inconsistent results. Because these new data include many additional CHD events, data are now sufficient to do a meta-analysis that includes only prospective studies, which provide greater validity. Data on the association between subclinical hypothyroidism and mortality are also conflicting (14, 17, 19, 20). The consequences of subclinical hyperthyroidism have been less frequently studied than those of subclinical hypothyroidism. Subclinical hyperthyroidism has been associated with cardiovascular and total mortality (15), but with conflicting data (14, 17). Two of the 3 recent, large prospective studies (14, 17) also examined CHD and mortality in subclinical hyperthyroid participants. To summarize prospective evidence about the relationship between subclinical thyroid dysfunction and CHD and mortality, we did a systematic review of prospective cohort studies. Methods We followed a standardized protocol and conducted and reported this analysis according to the guidelines of the Meta-analysis Of Observational Studies in Epidemiology group (21). Data Sources and Searches We conducted a systematic literature search of MEDLINE for articles in any language on the association between subclinical thyroid dysfunction (both subclinical hypothyroidism and hyperthyroidism) and CHD or mortality (cardiovascular and total) published from 1950 to January 2008. To avoid missing any relevant study, we also searched the bibliographies of key articles in the field and those included in this review. We did our search on an Ovid (MEDLINE) server by using broadly defined Medical Subject Headings, such as thyroid diseases, hypothyroidism, hyperthyroidism, thyroid hormones, thyrotropin, mortality, myocardial ischemia, survival, and cardiovascular diseases and such keywords as subclinical hypothyroidism, subclinical hyperthyroidism, subclinical dysthyroidism, and subclinical thyroid, combined with the filter designed by knowledge information specialists from BMJ to select prospective studies (MEDLINE cohort-study filter) (22) but without their year limitation. Study Selection Two reviewers independently screened the abstracts and titles of the search results and eliminated articles only if they did not clearly study the association between subclinical thyroid dysfunction and CHD or mortality (cardiovascular or total) in a prospective design. The same 2 reviewers independently evaluated the remaining full-text articles for eligibility on the basis of a predefined set of eligibility criteria. Disagreements were resolved by consensus. We included only full-text, published, longitudinal cohort studies that measured thyroid function and followed persons prospectively, assessed CHD or mortality, and provided risk estimates or sufficient data to calculate risk estimates associated with subclinical thyroid dysfunction compared with normal thyroid function. Subclinical hypothyroidism was defined as elevated TSH levels and normal T4 levels (1). Several reviews suggest a TSH upper limit cutoff of 4.5 to 5.0 mU/L (1, 5), but others suggest decreasing the upper limit of the TSH range to 2.5 to 3.0 mU/L (23, 24). In the absence of consensus, we did not prespecify a TSH cutoff value to define subclinical hypothyroidism and did a sensitivity analysis by limiting the analysis to studies with a TSH cutoff of 4.5 mU/L or greater (5). Because most adults with elevated TSH levels have subclinical and not overt hypothyroidism (2), we included 2 studies with participants who had elevated TSH levels without a T4 measurement report (25, 26) and did a sensitivity analysis excluding those studies. For subclinical hyperthyroidism, we did not specify a TSH cutoff value (in the absence of consensus), but all studies had a cutoff value close to 0.3 to 0.5 mU/L. We included 1 study with participants who had low TSH levels without a reported T4 measurement (26) and did a sensitivity analysis excluding this study. For CHD, we considered myocardial infarction, angina, the acute coronary syndrome, revascularization (coronary artery surgery, percutaneous transluminal coronary angioplasty), and significant coronary stenosis (defined as 50%) (27). We also considered death due to CHD or cardiovascular disease and did a sensitivity analysis excluding studies that only included the latter. We assessed methods and criteria used for adjudication of those outcomes. The agreement between the 2 reviewers was 99.5% for the first screen (titles and abstracts; = 0.79) and 100% for the full-text screen (= 1.00). Data Extraction and Quality Assessment Two reviewers independently abstracted data on participant characteristics, criteria used to define subclinical thyroid dysfunction, CHD and mortality data, and study results with adjustment factors by using a standardized data collection form. Discrepancies in data extraction between reviewers were resolved by consensus. We systematically assessed key indicators of study quality (28): methods of outcome adjudication and ascertainment that account for confounders and completeness of follow-up ascertainment. Similar to our previous meta-analysis (18), study populations were considered either a convenience or a population-based sample (defined as a random sample of the general population) (29). Methods of outcome adjudication were categorized as use of formal adjudication procedures and adjudication without knowledge of thyroid status. A formal adjudication procedure was defined as having clear criteria for the outcomes that were reviewed by experts for each potential case (29) (for example, specific electrocardiogram or cardiac enzyme modifications for CHD). We did not consider CHD adjudication based only on death certificates as a formal adjudication procedure. If an article did not clearly mention 1 of these criteria, we considered that it had not been done. We contacted the authors of 7 studies (12, 14, 15, 19, 26, 30, 31) that met inclusion criteria but did not provide specific data on the associations between subclinical thyroid dysfunction and CHD or mortality. We obtained risk estimates and CIs for cardiovascular and total mortality from cohort studies in the United Kingdom (15) and the Netherlands (19), as well as specific data for CHD mortality from a cohort of cardiac patients in Italy (31). Authors of a cohort study that published data about the relationship between autoimmune thyroid disease and CHD (30) provided us with data specific to subclinical hypothyroidism that were available for a subgroup of the participants. Three studies provided us with specific numbers of outcomes in each thyroid group (12, 14, 26). We used the most adjusted risk estimates available (the model containing the greatest number of covariates), unless a separate model further adjusted for thyroid antibodies, because thyroid autoimmunity has been hypothesized to be a mediator in the association between subclinical hypothyroidism and CHD (20). We did a sensitivity analysis without the studies that adjusted for cholesterol because high cholesterol might be on the causal pathway. When risk estimates and CIs were not provided but raw data were available (25, 30, 32), we calculated relative risks (RRs) and CIs by using the Woolf method (3335). Data Synthesis and Analysis We first qualitatively synthesized data, paying particular attention to which definitions of subclinical thyroid dysfunction were used and which outcomes were measured. To calculate summary estimates and CIs of the risk for subclinical thyroid dysfunction, we pooled both RRs and hazard ratios (HRs) by using random-effects models based on the variance model developed by DerSimonian and Laird (36). Analyses were repeated by using fixed-effects models for comparison. The presence of heterogeneity across studies was evaluated by using the Q statistic with a conservative P value of 0.10 (37). We also calculated the I 2 statistic, which describes the total variation across studies attributable to heterogeneity rather than chance; an I

Circulating Levels of Inflammatory Markers and Cancer Risk in the Health Aging and Body Composition Cohort

Background: Chronic inflammation is associated with processes that contribute to the onset or progression of cancer. This study examined the relationships between circulating levels of the inflammatory markers interleukin-6 (IL-6), C-reactive protein (CRP), and tumor necrosis factor-α (TNF-α) and total as well as site-specific cancer incidence. Methods: Study subjects (n = 2,438) were older adults (ages 70-79 years) participating in the Health Aging and Body Composition study, who did not report a previous cancer diagnosis (except for nonmelanoma skin cancer) at baseline. Incident cancer events (n = 296) were ascertained during an average follow-up of 5.5 years. Inflammatory markers were measured in stored baseline fasting blood samples. Results: The adjusted hazard ratios (95% confidence intervals) for incident cancer associated with a 1-unit increase on the natural log-scale were 1.13 (0.94-1.37), 1.25 (1.09-1.43), and 1.28 (0.96-1.70) for IL-6, CRP, and TNF-α, respectively. Markers were more strongly associated with cancer death: hazard ratios were 1.63 (1.19-2.23) for IL-6, 1.64 (1.20-2.24) for CRP, and 1.82 (1.14-2.92) for TNF-α. Although precision was low for site-specific analyses, our results suggest that all three markers were associated with lung cancer, that IL-6 and CRP were associated with colorectal cancer, and that CRP was associated with breast cancer. Prostate cancer was not associated with any of these markers. Conclusions: These findings suggest that (a) the associations between IL-6, CRP, and TNF-α and the risk of cancer may be site specific and (b) increased levels of inflammatory markers are more strongly associated with the risk of cancer death than cancer incidence.