Osteoporosis in Patients on Long-Term Home Parenteral Nutrition: A Longitudinal Study


  • The authors have no conflict of interest.


The prevalence of osteoporosis was investigated in 88 patients with intestinal failure (IF). Osteoporosis was found in 67%, dependent of body mass index and age when IF occurred. In 56 patients on HPN, followed prospectively, changes in bone density were dependent on the duration of HPN; older patients had a higher increase.

Introduction: It has been suggested that low bone mass and negative bone balance may occur in adult patients receiving home parenteral nutrition (HPN). The aim of this study was to assess prospectively the prevalence of osteoporosis in intestinal failure (IF) patients and the changes in bone mineral density in those on long-term HPN and to analyze the factors that may influence the occurrence and evolution of osteoporosis.

Materials and Methods: Bone mineral density was measured at the lumbar spine and femoral neck in 88 IF patients.

Results: At the first bone mineral density determination (baseline), the prevalence of osteoporosis was 67% in this population (median age, 52 years). Ten percent of the patients with osteoporosis experienced fragility fractures. Osteoporosis was independent of age and gender but occurred earlier in patients who had received corticosteroids. At baseline, the lumbar Z-score was positively correlated mainly to body mass index and age when IF occurred; these two parameters explained 34% of the Z-score. Repeated measurements were performed in 56 patients during long-term HPN (mean duration, 5.5 ± 1.2 years). The changes in Z-score at the lumbar spine were dependent on the age when IF occurred and on the duration of HPN, with a synergistic effect between them. The older the patients, the higher the increase in Z-score during HPN.

Conclusion: HPN had no deleterious effect on cortical bone and actually improved trabecular bone in patients whose intestinal disease started after the age of 21 years.


Home parenteral nutrition (HPN) is used to provide patients with severe malabsorption related to intestinal failure (IF) with the nutrients they require. Several reports in the 1980s mentioned the occurrence of bone diseases associated with HPN as signs of various toxicities or deficiencies. Osteomalacia was initially reported, and this was later linked to the toxicity of the aluminum contained in the parenteral nutrition fluids. Osteomalacia has almost disappeared as a result of the removal of casein hydrolysates from parenteral solutions.(1) Subsequently, low bone mass, negative calcium balance, variable levels of parathyroid hormone, and vitamin D toxicity have been reported in patients receiving HPN, because other toxicities or deficiencies could play a role in bone diseases.(2) Patients receiving HPN for IF therefore have several risk factors that could contribute to the onset of osteoporosis.(3) However, the occurrence of osteoporosis remains to be assessed in this population because previous reports included only a small number of patients whose bone mineral density (BMD) was measured using X-ray absorptiometry at trabecular and cortical sites. In addition, the changes in BMD during long-term HPN remain to be evaluated. We hypothesized that HPN is not deleterious for bone and that bone density might be improved with nutritional assistance.

In this study, we assessed the prevalence of osteoporosis in 88 consecutive patients with intestinal failure. In a subgroup of 53 adult patients with chronic IF, we looked at the change in BMD during long-term HPN and analyzed the factors that could be influencing this pattern.



All patients attending our approved HPN center between June 1995 and June 2000 with a diagnosis of benign nonmalignant IF requiring HPN were included in this study. Patients were included in the analysis regardless of age, underlying causes of benign IF, follow-up time of the disease, and duration of HPN. The cohort consisted of 88 patients, 42 women and 46 men, who were between 18 and 81 years of age (first quartile, 41 years; median, 51 years; third quartile, 65 years) at their last BMD measurement. The age of each patient was subdivided into three periods.

•Age when IF was diagnosed: this ranged from 2 to 79 years old (quartiles: 23–40–55 years). Because bone loss occurs between 50 and 60 years in the general population, we subdivided patients into those below and those above 55 years of age, and this cut-off corresponded to the upper quartile.

•The median of the delay between the diagnosis of IF and the first HPN was 14 months: this ranged from a few days to 30 years. The frequency distribution was as follows: <1 year = 47%, between 1 and 10 years = 33%, between 10 and 20 years = 16%, and >20 years = 4%. Patients were divided into three groups (≤1 year, from 1 to 10 years, >10 years). Patients whose IF lasted for >1 year before the start of HPN were arbitrarily classified as chronic IF; those that lasted <1 year were classified as acute IF.

•The duration of HPN was determined from the start of HPN to the last BMD measurement. It was terminated by the end of follow-up or by death and was <1 year in 11 patients and between 1 and 20 years in 77 patients (quartiles: 3.6–5.5–11 years). Eleven women and six men died during follow-up. HPN was not interrupted, except in eight patients who received HPN for two periods, with a non-HPN interval of 1 month to 11.5 years (mean: 3.6 years).

During follow-up, 217 BMD measurements were performed in 8 years (range, 0–8 years): one measurement in 24 patients, two measurements in 33, three or four measurements in 22, and five to eight measurements in 9 patients. At the time of the first BMD measurement, 30% of the patients had been on HPN for less than 3 months, and the other 70% had a median HPN duration of 3.5 years (range, 3–240 months).

Activity and bed rest period were not evaluated because they were not regular throughout the disease time. However, all patients were ambulant during the follow-up at home.

HPN was tailored to the needs of each patient and was intended to help them approach ideal body weight, taking into account the net absorption of the remnant bowel. HPN input was designed to complement the net absorption of food and to provide a total energy intake equivalent to 1.5 times the resting energy expenditure for each patient. By the end of follow-up, patients were receiving a median of five (range, two to seven) cyclic nocturnal HPN infusions per week; each all-in-one infusion in an ethyl vinyl acetate bag contained 9 (6–14) g of nitrogen, 250 (100–400) g of glucose, and 30 (0–50) g of a lipid emulsion, and was equivalent to 0.9 (0.3–1.2) of the resting energy expenditure of each patient. All-in-one infusion bags also contained 8 (4–12) mmol of calcium, 12 (7.5–15) mmol of phosphate, 14 (8–20) mmol of magnesium, and 220 IU of vitamin D, the latter being supplied by one vial of intravenous polyvitamin solution (Cernevit; Baxter, Maurepas, France). When the number of bags was at most three bags per week, two vials of polyvitamins were given per bag to comply with the American Medical Association recommendations for intravenous vitamins. In addition, all patients received daily oral mineral and vitamin supplements with calcium 2 (1–3) g, magnesium 9 (6–20) mmol, and 1-alpha hydroxycholecalciferol (0.25–1) μg.


Areal BMD (g/cm2) was measured at the femoral neck (FN) and lumbar spine (LS, L2-L4) using DXA (DPX-L densitometer; Lunar, Madison, WI, USA). Because the normal values of adult BMD are higher in men than in women and because BMD decreases with age, the BMD values are expressed as the Z-score ([individual BMD of the patient − mean BMD of referent population of same age and sex]/SD of referent population) and the T-score ([individual BMD of the patient − mean BMD of referent population of same sex but aged 20–25 years]/SD of referent population). The T-score was used to diagnose osteoporosis at each bone site (FN or LS) according to the WHO definition (T-score < −2.5 SD). Patients were defined as osteoporotic if the T-score was below 2.5 SD at either LS or FN. The referent population was a multicenter healthy French population of the same gender and age, referenced by the manufacturer.

Anthropometric parameters and those affecting bone metabolism were recorded: weight, height, and body mass index (BMI; kg/m2), duration of malabsorption (age at the discovery of the disease), and time between diagnosis and start of HPN. The presence of peripheral and/or vertebral fractures and of clinical data influencing BMD were also recorded: whether there had been any previous corticosteroid therapy, the duration of the menopause, and whether menopausal women had received hormone replacement therapy.


Biochemical assessment was performed in a subset of 37 patients (37/88; 42%).

25(OH)D3 was measured by radiocompetition after a step of purification. In brief, after addition of tracer amounts of tritiated 25(OH)D3 to estimate recovery, serum was extracted by acetonitrile and the supernatant was purified on a C18 silica reversed-phase cartridge. 25(OH)D3 was eluated in acetonitrile, evaporated, and redissolved in n-hexane-isopropanol and assayed by radiocompetition using as a specific binder the binding protein of rat kidney, and as a tracer, tritiated 25(OH)D3 with high specific activity (Amersham, Paris, France). Vitamin D deficiency in a normal population living in Paris has been estimated at a level below 8 ng/ml.

1,25(OH)2D3 was measured by radiocompetition after a step of purification and extraction as described above and assayed by radiocompetition using as a tracer tritiated 1,25(OH)2D3 with high specific activity (Amersham). The normal range for our normal population was 20–80 pg/ml.

Plasma parathyroid hormone (PTH)(1–84) was measured by sandwich radioimmunometry with the Intact PTH kit (Nichols, San Juan, CA, USA); the range for normal subjects was 10–65 pg/ml.

Statistical analysis

The effect of HPN on BMD for each patient was assessed from repeated measurements of the Z-score performed during the course of HPN. At any time, the current Z-score level can be viewed as the sum of the initial BMD at the time of first course of HPN and its subsequent change during HPN. The analysis consisted of two steps: analysis of the initial bone measurements and the change during follow-up.

First, the initial BMD was used to define the prevalence of osteoporosis. Assuming that the unknown age of occurrence of osteoporosis was in fact close to one of the first BMD measurements, we estimated the age-specific occurrence of osteoporosis using the non-parametric Kaplan-Meier method.(4) The curves were compared between genders, for acute IF and chronic IF, and for the presence or absence of corticosteroid treatment using a log-rank test.

Second, the initial Z-score was analyzed quantitatively using multiple regression analysis. The initial Z-score was studied using four potential explicative factors: (1) the age at diagnosis of IF disease, (2) the delay between the onset of IF and the start of HPN, (3) BMI (adjusted for gender), and (4) corticosteroid therapy (present, absent). The interactions between each couple (1–2 and 3–4) of variables were also included in the model.

Third, we determined the change in the individual Z-score from the slope of the individual linear regression of the Z-score taking into account the interval between the start of HPN and the Z-score measurements. These slopes define two independent random effects: a within-subject effect (the deviation of each measurement around the individual regression line) and a between-subject effect (the deviation of each individual slope around the mean change). Our aim was to estimate this change and to test which factors influenced this estimate. We based the analysis on the linear mixed-effect model,(5) which differs from multiple linear regression by taking into consideration the fixed and random effects and the correlation between successive data in the same patient. The assumptions underlying this model are largely discussed elsewhere.(6) We included the same explicative variables as for the analysis of the initial Z-score. In addition, we stratified the change in Z-score according to the presence or absence of osteoporosis. In the multiple regression and the mixed linear model, we used the lack-of-fit technique to select the best-fitting and most parsimonious subsets of the factors. Hypothesis testing was assessed by comparing the maximum likelihood observed in two nested models, using the likelihood ratio test. Diagnostic tests to detect non-normality and outliers were performed.

Fisher's test was applied to compare the qualitative data. Results are expressed as the mean and SD. The level of significance was set at 5%. Statistical analysis was performed using the S-PLUS 2000 statistics package for Windows.


Patient characteristics

Table 1 shows the demographical and clinical characteristics of the patients. At the baseline of the study, the women were older, and their IF had occurred later than in the men. The mean BMI was at the lower limit of the normal range in both sexes. For both women and men, the mean bone density of the whole population showed osteopenia at trabecular (lumbar spine) and cortical (femoral neck) sites.

Table Table 1. Underlying Intestinal Pathology by Decreasing Frequency
original image

Table 2 shows the causes of IF that underlined the characteristics of patients. Sixty-five patients (76%) were less than 55 years of age at diagnosis. Among them, 28 had diseases that are usually diagnosed in young patients: congenital volvulus, intestinal pseudo-obstruction, and Crohn's disease. No single class of diagnosis was predominant among the older people. All the Crohn's patients (n = 16) had received corticosteroid treatment.

Table Table 2. Anthropometric Characteristics of the Population Under HPN
original image

The youngest patients had the longest delay before the start of HPN (Spearman correlation, −0.24; p = 0.025) and the longest duration of HPN (Spearman's correlation, −0.32; p = 0.002). The patients under 55 years of age at diagnosis had a median duration of HPN of 5.5 years, whereas the older patients had a median duration of HPN of 3.7 years. There was no significant correlation between the duration of disease before starting HPN and the duration of HPN.

Mean serum levels of 25(OH)D in a subset of 37 patients were 15 ± 1.6 ng/ml. In our patients, four patients had a level below the threshold of 8 ng/ml. In contrast, the 1,25(OH)2D levels were 37 ± 15 pg/ml; all were within the normal range because they were receiving a 1-α hydroxycholecalciferol supplement. PTH levels were 24 ± 17 pg/ml, indicating the absence of hyperparathyroidism.

Prevalence of osteoporosis at baseline

The prevalence of osteoporosis was assessed by T-score according to the WHO definition. At baseline, which was the time of the first BMD, 59 patients (67% of all patients) had a T-score below 2.5 SD, defining osteoporosis, and no patient became osteoporotic during follow-up. Among the osteoporotic patients, six experienced fractures during HPN (two vertebrae, one ankle, two femoral neck, one humerus). No fracture occurred among the patients without osteoporosis.

The prevalence of osteoporosis was 71% and 52% (p = 0.12) for patients below and above 55 years of age, 58% and 76% (p = 0.069) for men and women, 56% and 75% (p = 0.067) for acute IF and chronic IF, and 63% and 68% (p = 0.70) for patients with and without corticosteroid treatment, respectively. To take into account the age at diagnosis of osteoporosis (the first BMD measurement), we compared the cumulative probability distribution of osteoporosis between the genders (p = 0.65), acute IF and chronic IF (p = 0.13), and patients with or without corticosteroid therapy (p < 0.001). Figure 1 shows the deleterious effect of previous corticosteroid treatment, which contributed to a reduction of BMD at a much younger age than expected.

Figure FIG. 1..

Occurrence of osteoporosis among patients with or without corticosteroids. The occurrence of osteoporosis was investigated in 88 patients at the time of the initial bone mineral measurement using a T-score < −2.5 SD as a criteria. The cumulative probability of osteoporosis was calculated using the Kaplan-Meier method. Two groups of patients were analyzed: 16 patients who had received corticosteroids and 62 patients who had never received corticosteroids. Osteoporosis occurred earlier in patients who had received corticosteroids.

Determinants of baseline BMD

To evaluate the determinants of baseline BMD and the changes during HPN, we chose to follow the Z-score because of age and gender heterogeneity of the patient. Moreover, T-score was influenced by age, which has an additive effect on the disease that we studied. Choosing the Z-score allowed assessment of the effect of the disease regardless of the effect of age. Considering BMD at the lumbar spine, 34% (R2) of the variance of initial Z-score could be explained by the age at IF (regression coefficient: 0.039 ± 0.008, p < 0.001) and the BMI (0.16 ± 0.05, p < 0.001): the earlier IF had occurred and the lower the BMI, the lower the Z-score. These two effects were additive; therefore, the Z-score was related to both age at diagnosis and to BMI, which is itself dependent on age. The BMI increased with age by approximately 0.7 kg/m2 every 10 years (regression coefficient = 0.069 ± 0.019, p < 0.001). The initial Z-score did not depend on either of the characteristics of the IF disease (acute or chronic, corticosteroids or no corticosteroids) or on the delay between the diagnosis of IF and the start of HPN. Among the 16 Crohn's patients, 10 had osteoporosis at the first BMD measurement.

Similar results were observed at the femoral neck. The coefficient of determination (R2) was 38%, and regression coefficients were 0.021 ± 0.007 (p < 0.001) for age at IF diagnosis and 0.16 ± 0.04 (p < 0.001) for BMI.

Change in BMD during HPN

During HPN, BMI increased by 0.9 kg/m2 over 10 years (0.088 ± 0.020, p < 0.001). To analyze the individual change in Z-score during HPN, we selected patients with at least two BMD measurements. This subset was composed of 56 subjects with IF diagnosed at various ages (12 when they were under 20 years old and 12 when they were over 55 years old) and with a median duration of HPN of 4 years (range, 0–20 years). The follow-up for BMD measurements was 2 (0–8) years with three (two to eight) measurements per patient (n = 163 DXA).

At the lumbar spine, the evolution of Z-score was dependent on the age when IF was diagnosed (p < 0.01) and on the duration of HPN, with synergism between these two effects. The older the patients at IF diagnosis, the greater the increase of the Z-score during HPN (Fig. 2). The change in the Z-score was not correlated to BMI, to the duration of IF before HPN was started, to whether the patient had received corticosteroid treatment, or to whether they presented with osteoporosis at the first BMD measurement. At the femoral neck, there was no significant change in the Z-score during long-term HPN.

Figure FIG. 2..

Mean change of lumbar spine Z-score during long-term HPN among 56 patients with or without osteoporosis according to the age at IF onset. To illustrate the results, we chose three ages equally spaced within the age range of our patients, and we calculated the evolution using the regression equation. As the duration of treatment decreased linearly with age, we did not extrapolate the evolution above the duration of follow-up. The negative evolution under HPN in young patients became positive with aging, and the change was reversed when the patients reached the age of 21. Similar evolutions were observed among patients with osteoporosis. However, their Z-scores were much lower (all values were reduced by 1.1 SD).


The selection criteria for inclusion of patients in this study was the presence of HPN, and our population was based on the enrollment of benign IF patients, recruited through our HPN approved center for the Ile de France region. This made it possible to avoid bias in the recruitment of patients with symptomatic bone diseases and to estimate the prevalence of osteoporosis in patients with benign, that is, nonmalignant, IF, mainly caused by either short bowel syndrome or dysmotility of the bowel with pseudo-obstruction. The patients received 220 IU of vitamin D in the HPN regimen plus daily 1-alpha hydroxycholecalciferol supplementation. Only four patients had 25(OH)D3 levels compatible with vitamin D deficiency, but the levels of 1,25(OH)2D3 were within the normal range for all of them. Moreover, PTH levels were normal, indicating the absence of chronic vitamin D deficiency in our patients.

Very few previous studies measured BMD using absorptiometry in HPN patients(1,7); indeed, most previous studies were cross-sectional studies including a small number of patients. In these studies, decreased bone density was observed at the lumbar spine, which is trabecular bone, but not at the wrist, which is mainly cortical bone.(1,2) Very recently, BMD was evaluated in 165 patients from nine centers in a retrospective and cross-sectional manner.(8) In their population, the prevalence of osteoporosis was 41%. Our results show that osteoporosis is quite prevalent (67%) in patients with nonmalignant IF who require HPN and that 10% of them have had bone fractures. It is noteworthy that the high prevalence of osteoporosis observed in this overall population was of the same order of magnitude as that in our Crohn's disease patients who had received corticosteroids; therefore, we did not observe a significant effect of the type of benign disease or of the type of IF (acute versus chronic) on the occurrence of osteoporosis, but ultimately all the affected patients suffered from severe malabsorption of variable duration. The lack of any correlation between the estimated duration of IF before HPN and the Z-score may also indicate the negative role of acute IF on BMD in patients who had suffered from an abdominal catastrophe with a long stay in intensive care before being referred to us. Indeed, at time of their inclusion in our HPN program, most patients were recovering from malnutrition, and as expected, BMI had a positive influence on the Z-score value at start of HPN.

Earlier onset of osteoporosis appeared in Crohn's patients that was not linked only to corticosteroid treatment as reported in the literature,(9–11) but also to the younger age at which the disease occurs in these patients. Indeed we observed a significant positive association between the age at which IF was diagnosed and Z-score BMD. Our hypothesis to account for this observation is that the negative impact of IF on bone could be less important in patients who had already reached their peak bone mass before the disease occurred. Conversely, the later in life that the IF occurs, the less the BMD was reduced. These data are in agreement with the recent observations of Pironi et al., who found that Z-score was correlated to the age of patients at which they start HPN.(8)

The specific impact of HPN on bone is a matter of debate. In young male rats without malabsorption, total parenteral nutrition induced reduced bone formation and low bone mass, suggesting that parenteral nutrition per se, without any other contributing factors, might induce bone loss.(12,13) In humans, two small studies in which bone biopsy specimens were performed in patients who had been on HPN for between 6 and 50 months HPN(14) showed a trend toward low-remodeling bone disease, with decreased bone formation.(15) Osteomalacia was not documented in any of these studies.

Our aim was to find out whether HPN decreases or increases BMD and therefore whether it aggravates or improves osteoporosis in these patients. Two studies were performed using photon absorptiometry. BMD of 14 patients receiving HPN was lower at the lumbar spine but not at the wrist compared with controls(7); follow-up of these patients for 28 months did not reveal any significant change at either of these sites. Verhage and Harry studied nine patients over 38 years of age who were receiving HPN.(16) The present study is the largest one designed in a prospective manner with a long-term follow-up. Z-score BMD was low at both lumbar spine and femoral neck when the study began. After 4 years of follow-up, Z-score had increased at the lumbar spine but remained unchanged at the femoral neck. In our patients, the change during HPN depended on age. In all patients, HPN increased BMD in the patients whose malabsorption disease had occurred after 21 years. However, HPN did not allow younger patients to reach normal BMD peak values. An early publication focused on the possible role of vitamin D to induce bone disease in parenteral nutrition.(17) In three patients receiving vitamin D fractures, bone pain and urinary calcium loss occurred, which were reversed by vitamin D withdrawal. Evidence for a causal relationship had not been provided, and these observations have not been confirmed. In our patients, vitamin D supplementation did not promote bone loss and therefore should be proposed to avoid vitamin D deficiency.

Our data from a large series of patients show that low bone density is predominant in patients whose intestinal disease started early in life. This might be caused by low bone formation, which would account for a failure to reach peak bone mass, and also to the corticosteroid treatment they received. In our population, long-term HPN had no deleterious effect on cortical bone and actually improved trabecular bone in patients whose intestinal disease started after the attainment of peak bone mass.