Height loss starting in middle age predicts increased mortality in the elderly



The purpose of this study was to determine the mortality risk among Japanese men and women with height loss starting in middle age, taking into account lifestyle and physical factors.

A total of 2498 subjects (755 men and 1743 women) aged 47 to 91 years old underwent physical examinations during the period 1994 to 1995. Those individuals were followed for mortality status through 2003. Mortality risk was estimated using an age-stratified Cox proportional hazards model. In addition to sex, adjustment factors such as radiation dose, lifestyle, and physical factors measured at the baseline—including smoking status, alcohol intake, total cholesterol, blood pressure, and diagnosed diseases—were used for analysis of total mortality and mortality from each cause of death.

There were a total of 302 all-cause deaths, 46 coronary heart disease and stroke deaths, 58 respiratory deaths including 45 pneumonia deaths, and 132 cancer deaths during the follow-up period. Participants were followed for 20,787 person-years after baseline. Prior history of vertebral deformity and hip fracture were not associated with mortality risk. However, more than 2 cm of height loss starting in middle age showed a significant association with all-cause mortality among the study participants (HR = 1.76, 95% CI 1.31 to 2.38, p = 0.0002), after adjustment was made for sex, attained age, atomic-bomb radiation exposure, and lifestyle and physical factors. Such height loss also was significantly associated with death due to coronary heart disease or stroke (HR = 3.35, 95% CI 1.63 to 6.86, p = 0.0010), as well as respiratory-disease death (HR = 2.52, 95% CI 1.25 to 5.22, p = 0.0130), but not cancer death. Continuous HL also was associated with all-cause mortality and CHD- or stroke-caused mortality. Association between height loss and mortality was still significant, even after excluding persons with vertebral deformity.

Height loss of more than 2 cm starting in middle age was an independent risk factor for cardiovascular and respiratory-disease mortality among the elderly, even after adjusting for potential risk factors. © 2012 American Society for Bone and Mineral Research


Many studies have shown increased fracture risk1–3 and mortality4–8 after clinical vertebral fracture. Even subjects with no clinical fracture and little pain but with vertebral deformity detected by X-ray showed slightly increased mortality.9 Other studies, however, showed no evidence of increased mortality among elderly with vertebral fracture.10 Increased mortality after hip fracture was observed in several studies.7, 11, 12

Kyphosis and height loss are thought to result mainly from underlying vertebral fractures, but have not yet gained much clinical interest other than as markers for osteoporosis.13–18 Height loss, however, not only could be caused by vertebral fracture, but also to some extent by intervertebral disk degeneration that decreases disk height; osteoarthritic conditions of the spine, hip, or knee, various inflammatory and structural/congenital spinal deformities; and weakness of the back muscles.19, 20 Our previous report showed that height loss and vertebral deformity significantly and independently affected quality of life (QOL) in the elderly, and height loss aggravated QOL more significantly than did vertebral deformity in all domains, even with different effect patterns between height loss and vertebral deformity.21 The mechanism behind such decreased height loss–associated QOL remains uncertain. Recent reports have suggested that hyperkyphotic posture or marked height loss might predict future fracture risk22 and mortality.23–25

In the present study, we assessed whether height loss starting in middle age affects all-cause and specific-cause mortality, after taking into account vertebral deformity and hip fracture in Japanese men and women.

Materials and Methods

Data source

Study participants comprised cohort members of the Adult Health Study (AHS), which was established to investigate late health effects of radiation exposure among atomic-bomb survivors in Hiroshima and Nagasaki. The original AHS cohort was comprised of about 20,000 atomic-bomb survivors and their controls selected from residents of Hiroshima and Nagasaki, based on the 1950 national census. Since 1958, the AHS cohort members have been followed through biennial health examinations, including physical examinations; measurements of height, body weight, and blood pressure; and chest X-rays. The health study participants were interviewed by nurses to obtain disease histories and lifestyle information, such as smoking status and alcohol intake. Participation rates in the study were around 70% to 80% throughout the follow-up period. Further information about the cohort and details of the health examinations are available elsewhere.26–28

Subjects of this study numbered a total of 2498 individuals (755 men and 1743 women) aged 47 to 91 years old, undergoing physical examinations in Hiroshima during the health study's 1994 to 1995 examination cycle (Fig. 1). Measurements of height, using a stadiometer, were available for all subjects at each examination since 1962. Participants were measured without shoes, with their heels, buttocks, and back against an upright board. The participants with hyperkyphosis were instructed to stand straight and stretch the muscles in their backs as much as possible. We defined height loss starting in middle age (HL) as the difference between a participant's average height in their 40s and height measured in 1994 to 1995. We calculated average height based on from two to five measurements at ages in the 40s for each participant. If a participant did not have data on average height in the 40s, we then defined HL as the difference between his or her average height in the 50s and height measured in 1994 to 1995 (those for whom height in their 50s was used: 12.5%). We also defined marked HL as a difference of more than 2 cm based on results from receiver operating characteristic (ROC) analysis for mortality.

Figure 1.

Timeline of the study.

The subjects underwent bone mineral density (BMD) measurements at the spine (L1-4, anteroposterior direction) and the total hip using dual X-ray absorptiometry (DXA, QDR-2000 [Hologic Inc, Waltham, MA, USA]) at the time of the examinations in 1994 1995. Morphometric vertebral deformity was diagnosed by lateral and posterior-anterior chest and spinal X-ray examinations. An experienced radiologist diagnosed vertebral deformity using semi-quantitative procedures.29, 30 We defined “prevalent vertebral deformity” as vertebral deformity at thoracic and lumbar vertebrae diagnosed during the 1994 to 1995 examination cycle, that is, prevalent cases in 1994 to 1995. Diagnosis of hip fracture was based on history-taking by a physician. Pathologic fractures or fractures due to traffic accidents or falls from heights were excluded.

The study follow-up of all participants began in the 1994 to 1995 examination cycle. The accumulation of each participant's person-years of risk ended at the date of death, or the date of the last examination before December 2003. Mortality follow-up was conducted through checks of the vital status of cohort members using the Japanese family registration system. We were thus able to completely follow the mortality status of the cohort members.

Statistical Methods

The rates of many diseases increase as some power of age, so a simple linear adjustment factor would undercontrol for age effects. To avoid this bias, we used an age-stratified Cox proportional hazard analysis, whereby people are assigned to an age stratum reflecting their age at baseline according to five-year age intervals. After confirming the assumption that hazard ratios were proportional, we used an age-stratified Cox proportional hazards model to assess the multivariate-adjusted hazard ratio (HR) for mortality. Fitted as categorical variables in the adjustment were assessments obtained at the 1994 to 1995 baseline: prevalent vertebral deformity (yes/no), prevalent hip fracture (yes/no), smoking status (never, current, former smoker, and unknown), alcohol intake (never, current occasional, current often, former drinker, and unknown), preexisting hypertension (yes/no), preexisting hyperlipidemia (yes/no), preexisting diabetes (yes/no), preexisting cardiovascular disease (yes/no), preexisting cancer (yes/no), marked HL (HL ≥ 2 cm/HL < 2 cm). Weight, height, body mass index (BMI: calculated as weight in kilograms divided by height in meters squared), systolic blood pressure (SBP), diastolic blood pressure (DBP), total cholesterol, BMD at baseline, radiation dose, and HL were fitted as continuous variables. For each risk factor, we first evaluated all-cause mortality using an univariate model. We then conducted evaluation with multivariate model, including variables found to be significantly associated with all-cause mortality. We obtained a final model after removing non-significant terms. As a result, we included such variables as sex, preexisting cancer, preexisting cardiovascular disease (CVD), preexisting diabetes, radiation dose, marked HL, smoking status, and alcohol intake in the model. We also evaluated mortalities caused by coronary heart disease (CHD) or stroke, respiratory disease, pneumonia, and cancer. In the same procedure, we analyzed participants excluding 191 participants with prevalent vertebral deformities. We used individual radiation dose estimates on the Radiation Effects Research Foundation's Dosimetry System 2002 (DS02).31

For the mortality analysis, we used the PHREG procedure in SAS program (SAS version 9.1, SAS Institute Inc, Cary, NC, USA), with stratification by 5-year intervals of baseline age, for estimation of the parameters and testing. With consideration for parameter distributions, we tested differences between the alive group and the death group using Student's t-tests for continuous variables and χ2 tests for categorical variables. A value of p < 0.05 was used for determination of statistical significance.

Ethical considerations

The present study was carried out in accordance with such national regulations as the Ethical Guidelines Concerning Epidemiological Studies (Ministry of Education, Culture, Sports, Science and Technology [MEXT], and Ministry of Health, Labour and Welfare [MHLW]). The study was approved by the Research Protocol Committee and the Human Investigation Committee at the Radiation Effects Research Foundation. At the time of the health examinations, informed consent was obtained from the participants. All participants provided written consent for all aspects of the examinations.


Characteristics of the participants taken at baseline are shown in Table 1. In men, mean ages ± 1 standard deviation (SD) in the 1994 to 1995 examination period for the alive group were 61.2 ± 8.9 years, and 70.3 ± 9.1 years for the death group, ranging from 47 to 91 years. In women, mean ages were 64.7 ± 9.1 years and 73.5 ± 8.9 years, respectively, ranging from 47 to 91 years. Mean age of the “death” group was significantly higher than that of the “alive” group. Mean height loss starting in middle age was 0.83 cm for men and 1.85 cm for women. Figure 2 shows HL distribution by sex. We used ≥ 2 cm as the cut-off value through the sensitivity analysis, and compared the death group with the alive group. Twenty-one men and 170 women had prevalent vertebral fracture, and 12 men and 44 women had prior history of hip fracture in the 1994 to 1995 examination period. Prevalence of diseases at baseline is presented in Table 1. The proportion of individuals with cancer and CVD appeared to be higher in the death group than in the alive group in both men and women. The proportion of individuals with hypertension appeared to be higher in the death group than in the alive group in women. Approximately 90% of women were postmenopausal with an average age at menopause of 47.7 years.

Table 1. Baseline (1994–1995) Characteristics of Study Population by Sex and Vital Status in 1994 to 1995
  • HL, historical height loss starting in middle age; BMI, body mass index; BMD, bone mineral density; SBP, systolic blood pressure; DBP, diastolic blood pressure; CVD, cardiovascular disease.

  • Mean (SD).

  • With consideration for parameter distributions, we tested difference between death or alive using t-test for height, weight, BMI, height at 40s or 50s, HL, marked HL, BMD, SBP, DBP, total cholesterol, radiation dose, using a Wilcoxon test for age, and using χ2-test for prevalence of hip fracture, prevalence of vertebral deformity, alcohol intake, smoking status, and diagnosed diseases.

  • a

    Longitudinal data of height are available for all study participants of the cohort since 1962. We defined height loss starting in middle age (HL) as the difference between a participant's average height in his or her 40s and height measured in 1994 to 1995.

  • *

    p < 0.05.

  • **

    p < 0.01.

Number of subjects6271281569174
Age (years)61.2 (8.9)70.3 (9.1)** 64.7 (9.1)73.5 (8.9)**
Height (cm)163.9 (6.0)161.5 (6.3)** 150.7 (5.7)147.6 (6.4)**
Weight (kg)61.4 (8.8)58.2 (9.2)** 52.8 (8.7)48.6 (9.3)**
BMI (kg/m2)22.8 (2.9)22.3 (3.0)23.2 (3.6)22.3 (3.9)**
height at 40s or 50s (cm)a 164.5 (5.8)162.9 (5.8)** 152.3 (5.2)150.9 (5.4)**
HL (cm)0.69 (1.01)1.50 (1.46)** 1.69 (1.94)3.34(2.76)**
marked HL (%)67 (10.7)42 (32.8)** 556 (35.4)127 (73.0)**
BMD (g/cm2)
 Spine (L1-4)0.960 (0.155)0.972 (0.164)0.796 (0.154)0.739 (0.148)**
 Total hip0.739 (0.115)0.709 (0.109)** 0.626 (0.107)0.571 (0.093)**
Prevalent hip fracture7 (1.1%)5 (3.9%)* 34 (2.2%)10 (5.8%)**
Prevalent vertebral deformity15 (2.4%)6 (4.7%)138 (8.8%)32 (18.4%)**
SBP131.8 (20.3)136.3 (22.1)* 130.7 (21.1)136.4 (21.4)**
DBP80.8 (11.4)77.3 (15.2)** 77.3 (11.4)76.4 (12.5)
Total cholesterol203.2 (34.0)202.0 (36.7)221.3 (34.6)211.1 (42.6)**
Diagnosed disease
 Hypertension185 (32.9%)37 (39.0%)390 (27.7%)50 (40.7%)**
 Hyperlipidemia44 (7.8%)6 (6.3%)194 (13.8%)15 (12.2%)
 Diabetes96 (15.3%)28 (21.9%)162 (10.3%)23 (13.2%)
 CVD288 (45.9%)78 (60.9%)** 660 (42.1%)113 (64.9%)**
 Cancer40 (6.4%)18 (14.1%)* 153 (9.8%)33 (19.0%)**
Alcohol intake
 Never105 (16.7%)27 (21.1%)769 (49.0%)102 (58.6%)*
 Current occasional107 (17.1%)29 (22.7%)256(16.3%)31 (17.8%)
 Current often262 (41.8%)31 (24.2%)** 113 (7.2%)9 (5.2%)
 Former14 (2.2%)8 (6.2%)* 13 (0.8%)5 (2.9%)*
 Unknown139 (22.2%)33 (25.8%)418 (26.7%)27 (15.5%)**
Smoking status
 Never88 (15.6%)9 (11.7%)920 (64.5%)67 (58.6%)
 Current210 (33.5%)42 (32.8%)104 (6.6%)11 (6.3%)
 Former167 (26.6%)35 (27.4%)47 (3.0%)6 (3.5%)
 Unknown152 (24.3%)36 (28.1%)406 (25.9%)55 (31.6%)*
Radiation dose (Gy)0.382 (0.634)0.432 (0.608)0.297 (0.514)0.407 (0.568)
Figure 2.

Percentage of those with height loss starting in middle age, for men and women.

Through December 2003, there were 302 all-cause deaths, 46 CHD and stroke deaths, 58 respiratory-disease deaths including 45 pneumonia deaths, and 132 cancer deaths. Mean follow-up was 8.3 years (Table 2). Participants were followed for 20,787 person-years after baseline. The death rate was 14.5 per 1000 person-years.

Table 2. Deaths Observed Between Baseline Examinations in 1994 to 1995 and December 2003
Number of individuals75517432498
Number of all-cause deaths128174302
Mean follow-up period (years)
Death rate (per 1000 person-years)20.711.914.5
Number of deaths by cause
 Coronary heart disease and stroke212546
 Respiratory disease273158

Multivariate adjustments were made for variables including physical and lifestyle factors, as described in “Methods,” which were further adjusted for estimation of mortality risk (Table 3). After these adjustments, mortality hazard ratio for the marked HL was 1.76 (95% CI, 1.31to 2.38), p = 0.0002.

Table 3. Hazard Ratios (HRs) Using Age-Stratified Cox Regression Analysis for All-Cause Mortalitya
Baseline factor in 1994–1995Hazard ratio95% CI
  • CI, confidence interval; HL, height loss starting in middle age; CVD, cardio vascular disease.

  • a

    The analysis included all variables in the table simultaneously.

  • *

    p < 0.05;

  • **

    p < 0.01.

Marked HLYes/No1.761.31–2.38**
Preexisting cancerYes/No1.551.12–2.15**
Preexisting CVDYes/No1.321.03–1.71*
Preexisting DMYes/No1.481.07–2.05*
Radiation dose1 Gy increment1.221.01–1.48*
Alcohol habitCurrent occasional/Never1.140.82–1.57
Current often/Never0.550.36–0.84**

Mortality risk also was analyzed for specific causes of death. Adjusted mortality risk results are presented in Table 4. When causes of death were classified, increased mortality risk for marked HL was observed in CHD- or stroke-caused death (HR = 3.35, 95% CI 1.63 to 6.86, p = 0.0010) and respiratory disease-caused death (HR = 2.52, 95% CI 1.25 to 5.22, p = 0.0130), but not cancer-caused death (p = 0.3143). No significant increase in mortality from cancer was observed. With significance, continuous HL also was associated with all-cause mortality (HR = 1.08 per 1 cm HL increase, 95% CI 1.03 to 1.14, p = 0.0034) and CHD- or stroke-caused death (HR = 1.11, per 1 cm HL increase, 95% CI 1.00 to 1.23, p = 0.0465). Previous history of vertebral deformity and hip fracture were not associated with all-cause mortality risk (Table 4).

Table 4. Hazard Ratios (HRs) Using Age-Stratified Cox Regression Analysis by Continuous HL, Marked HL, Vertebral Fracture, and Hip Fracture for Mortality
DeathContinuous HLMarked HLPrevalent Vertebral DeformityPrevalent Hip Fracture
  1. HL, height loss starting in middle age; CHD, coronary heart disease.

  2. Adjusted for sex, radiation dose, preexisting diabetes, preexisting cardiovascular disease, preexisting cancer, smoking status, and alcohol intake.

All-cause death
 95% CI1.03–1.141.31–2.380.78–1.640.72–2.18
 p value0.00340.00020.52670.4183
CHD- or Stroke-caused death
 95% CI1.00–1.231.63–6.860.86–4.160.67–5.82
 p value0.04650.00100.11230.2186
Respiratory disease–caused death
 95% CI0.99–1.231.25–5.220.63–2.890.17–2.95
 p value0.06840.01300.43780.6316
Cancer-caused death
 95% CI0.96–1.150.80–1.990.48–1.760.47–2.92
 p value0.26340.31430.79440.7367

The hazard ratios for marked HL were reduced only slightly when the 191 prevalent cases of vertebral deformity were excluded (eg, HR of 1.65, rather than 1.76 for all-cause mortality) (analyses not shown).


HL and mortality

This is the first study to show that HL of more than 2 cm increased the risk of all-cause death, CHD- or stroke- and respiratory disease-caused death, but not cancer death, with vertebral fracture assessed simultaneously. Furthermore, the present study showed that HL treated as a continuous variable was associated with significantly increased risk of all-cause mortality and CHD- or stroke-caused mortality.

Our previous report21 showed that height loss and vertebral deformity affected QOL significantly and independently in the elderly. Even after excluding individuals with vertebral deformity, height loss was associated with decreased QOL. Furthermore, it is observed that factors other than vertebral deformity, such as intervertebral disk degeneration and osteoarthritic conditions, also caused height loss. In the present study, we observed association between mortality and height loss starting in middle age, but not prevalent vertebral deformity. The presence of certain adverse health conditions, for example peer muscle strength, possibly causing height loss may be implicated.

Wannamethee et al. followed 4213 men measured for height at ages 40 to 59 and again 20 years later, observing 760 deaths occurring after six more years. In the aforementioned study, Wannamethee et al. described how osteoporotic disease complicated by vertebral fractures was not likely to explain increased mortality risk associated with height loss. Poor muscular strength and low skeletal muscle mass have been linked to bone loss and poor bone structure in men, which could result in height loss.32 The increased risk of CHD and all-cause mortality associated with height loss may thus reflect poor muscular strength and skeletal muscle mass loss from aging (sarcopenia), both of which have been shown to be predictors of mortality.33–35 Wannamethee et al. also discussed the idea that height loss might serve as a marker for sarcopenia and frailty.24 Hyperkyphosis, commonly used as a marker of aging, is frequently observed in the elderly. It is known that hyperkyphosis is associated with restrictive pulmonary disease36 and poor physical function.37–39 These findings suggest that hyperkyphosis also might be associated with occurrence of other states of poor health. Some studies have suggested association between kyphosis and mortality.22, 23, 25 Recently, Kado et al.25 conducted a prospective cohort study of 610 older white women who were diagnosed with kyphosis, and assessed mortality rates over an average follow-up of 13.5 years. They concluded that hyperkyphosis predicted increased risk of death independent of prevalent vertebral fractures. In addition, Kado et al.23 followed 1353 men and women over a period of 4.2 years, with mortality and cause of death confirmed by review of death certificates. They observed that older men and women with hyperkyphotic posture had higher mortality rates.

For CHD mortality, our results are consistent in principle with the results of the two previous studies. Additionally, we observed association between respiratory disease mortality and height loss starting in middle age in both men and women. Furthermore, height loss was associated with mortality even after individuals with vertebral deformity were excluded. The mechanism regarding how height loss might be associated with subsequent mortality is not currently well understood. Resulting height loss could affect normal functioning of the respiratory and gastrointestinal systems,13 which in turn might lead to early satiety, poor nutritional status, and weight loss.13 Height loss also appears to be related to sarcopenia,32 which is defined as the loss of skeletal muscle mass and strength with aging and is associated with weight loss40–43 and increased mortality.33–35

We found increased mortality associated with marked HL due to CHD or stroke and respiratory diseases, but no increased cancer mortality. Kado et al. reported that hyperkyphotic posture was specifically associated with increased rate of death due to atherosclerosis.23 Browner et al. reported that low bone mass was significantly associated with death from CVD and specifically stroke.44 Some evidence indicated similar pathophysiological mechanisms underlying both osteoporosis and cardiovascular disease.45, 46 Risk factors such as age, diabetes, hypertension, inflammation, dislipidemia, homocystienemia, and estrogen deficiency are prevalent in both disorders.44, 47

Osteoporotic fracture and mortality

Bliuc et al.48 reported that excess mortality was highest immediately after almost all fragility fracture events and then declined. The researchers observed that 30% of all post–hip-fracture deaths occurred in the first six months and 21% in the next 18 months. Other studies reported that increased mortality after hip and vertebral fractures was consistent over the initial five-year period.4, 6, 8, 11

In the present study, prevalent morphometric vertebral deformity and prevalence of hip fracture were not associated with increased mortality. Inconsistency between our report and many previous studies can be explained by differences between incidence and prevalence of fracture, because prevalent vertebral deformity and hip fracture in our study included those cases that had developed many years in the past. In addition, in the follow-up period, such differences as whether or not to include morphometric vertebral fracture and adjustment of potential confounders might have resulted in the inconsistency.

Strengths and limitations

One strength of this study is that the investigation was based on measured height using consistent methods throughout biennial health examinations conducted since 1962, thus reducing measurement errors. Since mean height in most age groups has increased recently in many regions around the world, including Japan, height loss would be overestimated in cross-sectional studies, and bias would be significant if recalled height were used.49 Our study was carried out using measured height at ages 40 to 49 and again some years later in a population-based study of men and women. Second, mortality follow-up has been carried out through checks of the vital status of cohort members using the Japanese family registration system. We were thus able to completely follow mortality of the cohort members.

There are some limitations to our findings. First, baseline data for physical activity and lung function were not available. Second, diagnosis of hip fracture was based on history taking by a physician, not X-ray examination. Furthermore, participants were atomic bomb survivors and thus not representative of the general Japanese population, although we adjusted for radiation, and there are no indications from earlier studies of this cohort that radiation affected BMD and fracture frequency.38, 48, 50


In conclusion, height loss starting in middle age is considered to be a factor associated with CVD and respiratory-disease mortality, independent of vertebral deformity, in Japanese elderly men and women. Further studies will be needed to elucidate the mechanisms behind such findings. Although the mechanisms are unknown, height loss, regardless of its causes, is a clinically important finding.


All the authors state that they have no conflicts of interest.


The Radiation Effects Research Foundation (RERF), Hiroshima and Nagasaki, Japan is a private, nonprofit foundation funded by the Japanese Ministry of Health, Labour and Welfare and the U.S. Department of Energy, the latter in part through the National Academy of Sciences. This publication was supported by RERF Research Protocol RP 3-89. This study also was supported by the research of “Effects of Vertebral Deformity and Body Height Loss on Activity of Daily Living and Its Prevention Among the Elderly” (16100201) by the Japanese Ministry of Health, Labour and Welfare.

The respective roles of the authors were the following: N Masunari and S Fujiwara made substantial contributions to conception and design, acquisition of data, and analysis and interpretation of data. F Kasagi, I Takahashi, and M Yamada contributed to acquisition of data. T Nakamura made contributions to interpretation of data.