Main

Increasing numbers of children surviving malignancy have led to an increasing emphasis being placed on their long-term quality of life(1). A significant morbidity is associated with fractures due to osteoporosis in adult life, which patients may be predisposed to by osteopenia and failure to reach an adequate peak bone mass in early adult life. Children who have received therapy for ALL are at particular risk of osteopenia due to a number of interrelating factors. GHD as a result of hypothalamic pituitary damage after cranial irradiation has been demonstrated to have detrimental effects on bone mineralization, which is improved after therapy with GH(2). Corticosteroids, through direct inhibitory effects on osteoblast activity, reduction in intestinal absorption of calcium, or adverse effects on GH action, may lead to a reduction in bone mineralization during therapy(3). Retrospective analysis of radiographs in children with ALL revealed 24% of subjects to be osteopenic at diagnosis(4), suggesting that the disease process itself may predispose to reduced BMC. Reduced exercise capacity or limitations on physical activity may also predispose to osteopenia because development of skeletal mass in healthy children has been associated with activity patterns(5).

Previous research has suggested that osteopenia after treatment for ALL is a consequence of previous cranial irradiation(6). Gilsanz et al.(6) demonstrated a 10% reduction in bone mineralization at the lumbar spine in children treated for ALL, all of which was accounted for by those who had received cranial irradiation. Therefore, the aims of this retrospective cross-sectional study were to measure the BMC in a group of long-term survivors of ALL who had received chemotherapy and cranial irradiation and to compare results with those from a group of survivors of other malignancies who had received chemotherapy but no cranial radiotherapy, and with their healthy sibling controls. The effects on BMC of the various chemotherapeutic agents, the age at diagnosis, the length of therapy, and the time since discontinuation of therapy were also assessed. Associations between BMC and previously published data on exercise capacity and physical activity in the same cohort were also explored(7,8).

METHODS

Subjects. Measurements were performed on a total of 86 children. Fifty-five were long-term survivors of childhood malignancy identified from the oncology database held at the regional oncology referral center for South Wales in Llandough Hospital and Community NHS Trust, Cardiff. Entry criteria for the study of these children included: 1) continuous first remission; 2) at least 18 mo since completion of therapy because the risk of relapse was considered to be reduced; and 3) between the ages of 7 and 19 y. The lower age limit was chosen because of the need for the child's cooperation with measurements, and the upper limit to allow assessments to be made throughout adolescence. Children were excluded if they were known to have any other chronic illness, including endocrine deficiencies, or were receiving treatment likely to have an influence on bone mineralization, such as GH therapy. The study was approved by the ethics committee of the South Glamorgan Health Authority, and written parental consent was obtained for each child.

ALL group. Survivors of ALL had been treated on standard Medical Research Council protocols during the period 1979 to 1990, which included trials UKALL VI (n = 2), VIII (n = 8)(9), and X (n = 24)(10). Each treatment regimen consisted of a brief induction block to induce remission, followed by a CNS prophylaxis phase consisting of cranial irradiation [18 Gy (n = 30) or 24 Gy (n = 5)] and intrathecal methotrexate, followed by maintenance chemotherapy for 2 or 3 y. The later protocols included further intensification blocks of more intensive chemotherapy. Induction chemotherapy included the use of prednisolone, vincristine, and asparaginase with the introduction of daunorubicin in UKALL VIII and X. Maintenance chemotherapy consisted of monthly mini-reinduction blocks of vincristine and prednisolone with continuous oral methotrexate and 6-mercaptopurine. One child included in this group had received treatment for NHL because the treatment regimen used was similar to that for ALL, including cranial irradiation (17.2 Gy). The mean (range) cumulative dose of prednisolone used was 5.9 (5.9) g/m2 for UKALL VI, 7.5 (6.0-8.7) g/m2 for UKALL VIII, 6.2 (5.3-10.8) g/m2 for UKALL X, and 3.7 g/m2 for NHL.

Other malignancies group. This treated comparison group comprised of children treated for a variety of other malignancies including acute myeloid leukemia (n = 6), NHL (n = 2), Wilms' tumor (n = 6), neuroblastoma (n = 2), yolk sac tumor (n = 2), rhabdomyosarcoma (n = 1), and Hodgkin's lymphoma (n = 1) between 1982 and 1991. This group as a whole had been treated with cytotoxic chemotherapy but no cranial or other radiotherapy. Chemotherapy for this group was given in pulsed blocks, allowing bone marrow recovery between blocks, as opposed to the more continuous nature of therapy for ALL. A variety of chemotherapeutic agents had been used, but apart from the three children treated for NHL or Hodgkin's lymphoma, this group had not received corticosteroid therapy.

Sibling control group. For comparison with the two treatment groups, measurements were also taken on a healthy group of their siblings. For cases in which there was a choice of more than one sibling within the appropriate age range, the one closest in age to the index case was chosen.

The major differences in therapy between the two treated groups included cranial irradiation and the more prolonged continuous nature of therapy in the ALL group compared with no radiotherapy and the shorter and more pulsed nature of chemotherapy for children treated for the other malignancies. The cumulative dose (mg/m2) of the individual chemotherapeutic agents was recorded from retrospective case note analysis. For intrathecal methotrexate, the number of injections was recorded. The dosage for maintenance therapy of oral methotrexate and 6-mercaptopurine in ALL was altered weekly according to the degree of bone marrow suppression. Case note analysis did not allow accurate calculation of cumulative doses for these agents to be obtained, and hence the total number of weeks of therapy was noted.

Anthropometry. Height (to the nearest 0.1 cm) and weight (to the nearest 0.1 kg) were measured by one trained observer using a wall mounted Harpenden stadiometer (Holtain Ltd., Crymych, Dyfed, UK) and Avery beam balance (Avery Ltd., Birmingham, UK), respectively. SDS for each measurement were calculated from the 1990 British reference standards(11). Puberty staging of pubic hair was assessed according to Tanner's classification(12).

Bone mineral content. BMC (g) and projected BA (cm2) for the whole body, lumbar spine (L1-L4), and left hip were measured by DXA using the Hologic QDR 1000/W densitometer (Hologic Inc., Waltham, MA)(13). The analysis is based on the differential attenuation of collimated x-rays [the mean energy of which is rapidly alternated between 70 and 140 kV (peak)] by tissues of differing chemical content. The attenuation data are converted to BMC by comparison with known bone mineral and soft tissue equivalent standards mounted on a calibration wheel through which the radiation passes(13). Hologic software version V5.67 was used for the whole body analysis and V4.66P for the lumbar spine and left hip. The operator was blinded to the clinical state of the patient undergoing measurement.

Each child wore light clothing and removed any objects containing metal before scanning. Whole body scanning was performed with the child lying supine along the long axis of the DXA couch. Lumbar vertebrae (L1-L4) were measured with the legs flexed at the hip by resting them on a box supplied by the manufacturer. Measurement of the hip was achieved by first externally and then internally rotating the left leg at the hip and strapping the foot at an angle of between 25° and 30° in a footholder supplied by the manufacturer. The Hologic densitometer allows regional analysis of BMC within the hip, which can be subdivided into femoral neck and trochanteric and intertrochanteric regions. These regions have been selected to represent critical points in the proximal femur where fractures occur. The location of each region is based on anatomical markers that are defined by the system software to ensure that the same location on the femur in each child is evaluated(13).

Interpretation of bone mineral data in growing children is complex because it is influenced by age, body size, sex, and pubertal stage. DXA expresses bone mineralization data as an areal BMD (BMD = BMC/BA) and not a true volumetric density (g/cm3). Despite this correction, areal BMD remains correlated with body size(14), age, and pubertal stage(15). Recently we have reported a new technique for the interpretation of BMC which allows comparison to be made between individuals of different ages, stages of puberty, and body sizes(16). By multiple regression analysis we derived predictive equations for BMC, using measurements made in the healthy sibling controls, with BMC as the dependent variable and BA, age, height, weight, pubertal stage, and sex as the independent variables. Independent variables were selected in a backward stepwise procedure such that all remaining variables in the final predictive formula for BMC were statistically significant (p < 0.05). All continuous variables were converted to natural logarithms because their relationships with BMC were curvilinear. The measured BMC was expressed as a percentage of the predicted value (%BMC). This technique adjusts for the influence of age, sex, body size, and stage of pubertal development allowing comparison to be made between individuals across the pediatric age range. The predictive formulae for BMC have been derived from the healthy sibling controls, hence the mean %BMC at each site for this group is close to 100%.

Exercise testing and levels of physical activity. Measurement of exercise capacity in response to intense physical activity (peak VO2) was measured while the patient ran on a motorized treadmill using an incremental discontinuous protocol until the point of physical exhaustion was reached(7). Peak VO2 provides a measure of exercise capacity and physical fitness. Levels of habitual daily physical activity were derived from the ratio of TDEE and BMR. TDEE was measured from minute by minute heart rate recording throughout the day. The calculation of energy expenditure requires previous calibration of heart rate with energy expenditure using a treadmill-based exercise test in individual subjects as described elsewhere(8). BMR was measured by ventilated hood calorimetry (Deltatrac II Metabolic Monitor; Datex Instrumentation Corporation, Helsinki, Finland).

Statistical analyses. One-way ANOVA was used to compare group means. Where the F ratio was significant (p < 0.05), the least significant difference post hoc multiple comparison test was applied to look for individual group differences. Pearson correlation coefficients were used to examine relationships between variables. The relationship between %BMC and time since discontinuation of therapy, age at diagnosis, or length of therapy in the two treatment groups were modeled by multiple regression, with %BMC as the dependent variable. Group differences were modeled using a dummy variable (1/0) identifying the ALL group, allowing group interactions with the independent variables to be explored. The influence of previous therapy on %BMC was examined using stepwise multiple regression analysis after converting the cumulative dose, number of injections, and total number of weeks of therapy to natural logarithms, so as to normalize the data. All analyses were performed using the Statistical Package for Social Science version 6.0.

RESULTS

Anthropometric data are shown in Table 1. There were no significant differences in age, height, or weight between groups. There were no significant differences in the mean (range) age at diagnosis, 3.2 (0.9-5.9) versus 3.7 (0.9-14.5) y and the mean (range) number of years off therapy 6.6 (2.6-12.8) versus 6.6 (1.5-11.8) y between the ALL group and other malignancies, respectively. The duration of therapy was significantly longer for those treated for ALL compared with the other malignancies [mean (range) 2.3 (1.7-3.9) and 0.5 (0.1-1.9) y, respectively, p < 0.001].

Table 1 Mean (SD) anthropometric measures for the three groups
Full size table

Despite no significant reduction in %BMC for the whole body in children treated for ALL (Table 2 and Fig. 1a), the %BMC was significantly reduced at the lumbar spine and left hip compared with controls and also compared with the other malignancies at the hip (p < 0.05), (Table 2 and Fig. 1,b and c). The subanalysis of the hip into femoral neck and trochanteric and intertrochanteric areas revealed significant reductions in %BMC for the ALL group at all regions compared with controls (Table 2 and Fig. 1,d-f). This was most marked at the trochanteric and intertrochanteric sites (Table 2 and Fig. 1,e and f). At the lumbar spine, five (14%) of the ALL children had a %BMC less than 2 SDs below the mean compared with two (10%) children from the other malignancies group and none of the controls. For the left hip, 10 (29%) of the ALL children had a %BMC less than 2 SDs below the mean compared with three (15%) children from the other malignancies group and one (3%) of the controls. There were no cases of overt clinical osteonecrosis in any group, and the incidence of previous fractures at the time of the study was not assessed.

Table 2 Mean (SD) % BMC at each anatomical site for the three groups
Full size table
Figure 1
figure 1

Boxplots of %BMC for (a) whole body, (b) lumbar spine, (c) left hip, (d) femoral neck, (e) trochanteric region and (f) intertrochanteric region for the three groups. ANOVA *p < 0.005, **p = 0.06. Other maligs. = other malignancies; Box = inter quartile range; thick central line within box = median; whiskers = full range; O = outliers; ----------- = mean %BMC; - - - - - - = ± 2 SD.

Full size image

There was a significant correlation between %BMC and length of time off therapy at all the scanned sites except for the whole body. Figure 2 shows the correlation for the (a) lumbar spine and (b) left hip. At the lumbar spine, the slope of the regression lines for the ALL and other malignancies was significantly different, suggesting that the length of time off therapy had a greater impact on the %BMC for the other malignancies than for the ALL groups, and any recovery after therapy in %BMC at the spine was slower in children treated for ALL. For the total hip, the intercept of the regression lines for the ALL survivors was significantly lower when compared with other malignancies, indicating that the %BMC for ALL survivors is significantly lower at any given number of years off therapy compared with the other malignancies at the hip. There were no significant correlations between %BMC and age at diagnosis or length of therapy.

Figure 2
figure 2

Relationship between %BMC at (a) the lumbar spine and (b) left hip with time off therapy. O = ALL; * = other malignancies. Regression line for ALL = ----------- (r = 0.43, p < 0.01 for the lumbar spine and r = 0.38, p < 0.05 for the hip); other malignancies = - - - - - - (r = 0.36, p = 0.1 for the lumbar spine and r = 0.39, p = 0.09 for the hip).

Full size image

Comparison with previously published data from the same cohort(7,8) showed correlations between %BMC at the spine and the hip with peak VO2 [r = 0.23, p < 0.05 (Fig. 3a) and r = 0.47, p < 0.001 (Fig. 3b), respectively] and levels of physical activity (TDEE/BMR) [r = 0.20, p = 0.06 (Fig. 3c) and r = 0.29, p < 0.01 (Fig. 3d), respectively].

Figure 3
figure 3

Relationship between peak VO2 (a and b) and levels of physical activity (c and d) with %BMC at the lumbar spine and left hip. O = ALL; * = other malignancies; â–ª = controls; ---------- = regression line for all groups combined.

Full size image

With multiple regression analysis, different chemotherapeutic agents were found to adversely affect the %BMC at the spine and hip. For the spine, the agents that led to a reduction in %BMC were the number of weeks of oral methotrexate and the cumulative dose of ifosfamide and bleomycin [coefficients (SE) = -1.9 (0.6)%, p < 0.01, -5.7 (2.4)%, p < 0.05, and -4.2 (1.6)%, p < 0.05, respectively, constant = 100.7 (2.2)%, p < 0.0001, r2 = 0.26, p < 0.01]. For the hip, the agents that led to a reduction in %BMC were the number of weeks of oral 6-mercaptopurine and the cumulative dose of cisplatin [coefficients (SE) = -2.8 (0.7)%, p < 0.001 and -2.3 (1.0)%, p < 0.05, respectively, constant = 101.0 (2.8)%, p < 0.0001, r2 = 0.23, p < 0.01]. There was no correlation between %BMC and cumulative corticosteroid dosage at any site.

DISCUSSION

The treatment of children for malignancy with intensive chemotherapy and radiotherapy has led to improvements in long-term, disease-free survival. This improved survival, however, is associated with many deleterious side effects of therapy(17,18). Skeletal abnormalities associated with previous therapy for ALL have been described, which include vertebral compression fractures(19), metabolic bone mineralization defects(20), osteopenia(2), and osteoporosis(6,21). No clear mechanism that accounts for these observations has yet been identified. The current study confirms previously published reports that children previously treated for ALL are osteopenic(2,6), with significant numbers having a %BMC of less than 2 SDs below the mean, putting them at increased risk of osteoporosis and fractures.

Measurement of bone mineralization in children by DXA scanning is a relatively new technique. Conventionally the data are expressed as an areal BMD and not as a true volumetric density. True volumetric BMD (g/cm3) can only be measured by quantitative computerized tomography, a technique which exposes the child to significant radiation exposure, e.g. 300 µSv for the lumbar spine(22), and was therefore not used in this study for ethical reasons. In comparison, the radiation dose from DXA is small; 2.5 µSv for each of the lumbar spine and hip and 5 µSv for the whole body giving a total radiation exposure in the current study of only 10 µSv. Given that this exposure is equivalent to that experienced from normal background radiation in 1 d, the technique was thought to be ethically acceptable for use in this research study(22). Given the problems of expressing bone density in the form of areal BMD (see "Methods"), we have used a statistical method of expressing BMC of children of differing ages, adjusted for height, weight, and stages of puberty(16). The use of this method suggests that the reduction in %BMC in subjects treated for ALL is a consequence of relative osteopenia, rather than an artefact due to the effects of differing size or puberty of individual subjects. Comparison between the %BMC of the healthy siblings from the current study with our previously published healthy cohort revealed no significant differences, suggesting that the sibling group was representative of the normal population.

Several hypotheses for the skeletal abnormalities observed in children treated for ALL have been postulated. Osteopenia present at diagnosis may result from direct involvement of bone by disease or an indirect effect from a product of malignant cells(4,23). Examination of markers of bone turnover at diagnosis before the administration of chemotherapy has revealed low levels of osteocalcin and 1,25-dihydroxyvitamin D3(23). This suggests that children with ALL have abnormalities of bone metabolism before treatment, implying that the leukemic process itself is implicated as a mechanism for defective mineralization leading to decreased bone mass in long-term survivors. After completion of therapy, plasma levels of magnesium, alkaline phosphatase, and osteocalcin have been demonstrated to rise compared with levels during therapy(20). This suggests that during therapy osteoblast activity may be reduced leading to a limitation in bone mineralization throughout the treatment period. Recently there has been much interest in the role of circulating cytokines and growth factors in the control of bone mineralization. Tumor necrosis factor and IL-6 have been shown to be powerful inhibitors of bone formation(24,25), whereas insulin-like growth factors, transforming growth factors, and bone morphogenetic proteins have been shown to influence osteoblast proliferation and differentiation(26). Modification or inhibition of such complex molecular mechanisms as a result of chemotherapy or the disease process itself may lead to abnormalities of bone mineral accretion. This area requires further research.

In the current study, there was a significant increase in %BMC with increasing length of time off therapy, suggesting that "catch-up" may occur after cessation of therapy. This would lend support to a hypothesis that bone mineralization is suppressed either before or during treatment, with "catch up" occurring after completion of therapy. However, results in the current study should be interpreted with caution given its cross-sectional design. Although, a recent longitudinal study examining bone mineralization during treatment for ALL showed that suppressed bone mineralization was present in only a small number of children at diagnosis, but as treatment progressed the reduction in BMC became more prevalent(27). Biochemical markers of bone turnover (collagen peptides and bone alkaline phosphatase) have been shown to be suppressed during induction and intensification phases of chemotherapy for treatment of ALL with "catch-up" in the intervening phase(28). These findings together suggest that chemotherapy plays a greater role in the development of osteopenia than the disease itself. Further longitudinal studies after cessation of therapy are required to confirm whether complete "catch-up" occurs after discontinuation of therapy.

Gilsanz et al.(6) studied the BMD of the spine using quantitative computerized tomography in 42 survivors of ALL. Twenty-nine had received chemotherapy and prophylactic cranial irradiation (18-25.2 Gy), and 13 had received chemotherapy alone. Overall there was a 10% reduction in BMD in children treated for ALL compared with a group of age- and sex-matched controls. This reduction was almost entirely accounted for those who had received cranial irradiation, with the nonirradiated group being no different from controls. The reduction of approximately 10% in %BMC at each scanned site (Table 2) for children treated for ALL in our study complies with that of Gilsanz et al.(6). A small reduction in the mean %BMC was also observed within the nonirradiated group compared with controls in our study, but this did not reach statistical significance.

The mechanism by which cranial irradiation leads to a reduction in BMD is unknown but may be a consequence of GHD. GHD as a result of hypothalamic pituitary dysfunction after cranial irradiation during childhood has been demonstrated to lead to osteopenia in adults who have survived malignancy(2). Adults who had been treated for ALL as children and were GHD as a result of cranial irradiation have been found to be osteopenic compared with both a group of adult ALL survivors who had received the same treatment but were not GHD and with a group who had been treated with GH for GHD(2). Because all three groups had received similar treatment and the only difference was the presence or absence of GHD, this evidence would suggest that bone mass can be restored by GH administration in GHD children. In the present study, biochemical measures of GH were not undertaken. However, the children were under regular endocrine review at the time of the study and were demonstrating normal growth patterns. This suggests that significant GHD at the time of assessment was unlikely. Subtle abnormalities in the periodicity of GH release after low-dose cranial irradiation have been reported(29), and it may be postulated that, despite normal growth, these may lead to a slow decline in bone mineral mass. If this were to be the case, bone mass would be expected to decrease the longer follow-up occurred. The present study, therefore, does not support this hypothesis because "catch-up" in %BMC appeared to occur with longer time off therapy. Again this requires confirmation from longitudinal studies.

Peak bone mass is attained several years after the completion of puberty(30), yet bone mineral accretion rates are greatest during the pubertal-growth period(31). Children treated for ALL have been shown to have an early and truncated pubertal growth period possibly as a result of cranial irradiation(32). In adults who were identified with GHD before the onset of puberty, BMD was significantly reduced compared with both age- and sex-matched controls and with GHD children who had received GH therapy during puberty(33). Documentation within the case notes of the patients participating in the current study was not complete enough to accurately pinpoint the timing of onset or the duration of puberty in those children who were postpubertal. However, none of the children were considered to have suffered true precocious puberty (onset at an age earlier than 2 SD below the mean age of pubertal onset for normal boys and girls). In addition, the formulae derived from multiple regression analysis and used to calculate the predicted BMC took into account the pubertal status of the child, hence controlling for any early onset.

Significantly increased BMD had been reported in children who are more physically active(5,34). In the current cohort of children, we have demonstrated that children treated for ALL have a reduced peak VO2 during intense exercise(7) and decreased levels of daily physical activity(8). Correlations exist between the measures of exercise and physical activity with %BMC at the spine and hip (Fig. 3), suggesting that relative inactivity predisposes to the development of osteopenia in such children. Because higher levels of physical activity are associated with improved BMC, it may be possible to prevent osteopenia in children treated for ALL by instigating physical education programs throughout and after treatment. This possibility requires further research.

Identification of other individual factors used in the treatment of childhood malignancy, which may account for the reductions in %BMC, is difficult because of interactions between these variables and the fact that most therapy is rarely administered in isolation. For example, children treated for ALL who received prophylactic cranial irradiation also received corticosteroid therapy, both of which have been shown to have detrimental effects on bone mineralization(3,6). For this reason, multiple regression analysis was used to explore possible interactions between variables in an attempt to find individual agents that may have an influence on %BMC. Methotrexate and 6-mercaptopurine are both DNA synthesis inhibitors, the former inhibiting dihydrofolate reductase, leading to a block in pyrimidine synthesis, and the latter blocking purine synthesis. Administration of oral methotrexate has been implicated previously to be associated with osteoporosis(35,36), although the putative role of 6-mercaptopurine is a new finding. Ifosfamide and cisplatin are both nephrotoxic and may lead to abnormalities in renal calcium and vitamin D metabolism. Ifosfamide therapy has previously been implicated as an agent associated with osteopenia in children treated for soft tissue sarcomas(37). Previous bleomycin therapy is associated with pulmonary fibrosis, which may result in reduced BMD through limitation on exercise capacity (see above). It is not clear why different agents were associated with a reduction in %BMC at the spine and the hip. DXA does not distinguish between trabecular and cortical bone. However, the spine represents an area predominately comprised of trabecular bone [66% versus 34% for cortical bone(38)], which because of its more rapid turnover rate compared with cortical bone, possibly makes it more susceptible to cytotoxic therapy. The various areas of the hip differ in their relative proportion of trabecular and cortical bone [e.g. for the femoral neck 25% and 75%, respectively and for the trochanteric region 50% and 50%, respectively(38)]. This may explain the greater reduction in %BMC observed at the trochanteric region compared with the femoral neck in children treated for ALL. When the %BMC of the body as a whole was analyzed there were no significant differences between groups. Whole body BMC is influenced by a much larger proportion of cortical bone [80% versus 20% for trabecular bone(38)] which, because of its slower rate of turnover may have been less adversely affected by chemotherapy. Surprisingly, corticosteroid dosage was not associated with a reduction in %BMC either alone or when other agents were taken into account. This may reflect the limited variability in the dose of corticosteroid administration in the UKALL protocols. The statistical analysis used to identify agents that may lead to a reduction in BMC is not ideal and further in vitro research into the effect of the various chemotherapeutic agents on osteoblast and osteoclast activity is required to resolve this complex area.

In conclusion, children treated for ALL are osteopenic compared either with healthy sibling controls or with children treated for a variety of other tumors. After therapy there is a period of "catch-up" in %BMC, but further longitudinal studies are required to see whether attainment of a normal adult peak bone mass is achieved. The mechanism for the onset of this osteopenia is complex and probably multifactorial but probably includes cranial irradiation, various chemotherapeutic agents, and physical inactivity.