This longitudinal observational study was conducted at the Departments of Paediatrics and Radiodiagnosis, at a Tertiary Care Hospital. The study was conducted for the period of 16 months (November 2013- March 2015). The study was approved by the Institutional Ethics Committee (IEC) (sanction No- 1-40/64/2013/IEC/Thesis/PGIMER.RMLH/10278, date Nov 16^th 2013, Chairperson- Dr K Satyanarayana ICMR).

Sample size was calculated, taking the study by Trapani S et al., as reference, in which 20 patients of juvenile systemic lupus erythematosus receiving steroids, revealed a LS mean BMD of 0.978 gm/square cm±0.165 Standard Deviation (SD) at baseline and 0.947 gm/square cm±0.184 SD a year later [6]. The mean difference in BMD at two point interval was 0.031. However, the data on SD of the mean difference between the two values was not provided. Hence, to detect a mean difference of 0.031 with an assumed SD of the mean difference of 0.05 for the paired data with an alpha error of 0.05, beta error of 2, a sample size of 21 patients was calculated. Thirty patients in paediatric age group who were to receive systemic glucocorticoids for at least 6-12 weeks were enrolled for the study. Thirty age and sex matched healthy children of nurses and paramedic staff of the hospital were enrolled as controls.

Informed consent was obtained from parents or guardians or care giver of all children enrolled in study and assent was taken where ever necessary. Children with chronic malabsorption, malnutrition, rickets, chronic renal, liver and endocrinal disease or those who received vitamin D and/or calcium supplementation in last six months or receiving glucocorticoid as replacement were excluded from study.

The following data were recorded for each patient at the time of study: Age, sex, diagnosis, Body Mass Index (BMI), Sexual Maturity Rating (SMR) stage [25,26] date of start of steroid, dose of steroid, date of stoppage of steroid, total duration of steroid received, cumulative dose of steroid.

At baseline, for both cases and controls, daily calcium and vitamin D intake in diet, daily sunlight exposure and frequency of weight bearing physical activity per week, prior to illness were calculated as per the method described by Dey S et al., [7]. Thereafter, the patients were advised to take adequate calcium and Vitamin D in diet or supplemented if deficient and also to have adequate sunlight exposure and do weight bearing physical activity of one hour per day at home as far as possible in fully ambulatory patients. Adequacy of intake of calcium (800-1000 mg/day) and vitamin D (600 IU/day), sunlight exposure and frequency of weight bearing exercise per week were assessed every week. Cases and controls, who had deficient serum vitamin D levels were given 6 lacs IU of vitamin D2 once by intramuscular route. NS patients received prednisolone as per the standard protocol [27]. In patients who received systemic steroids other than prednisolone, the prednisolone equivalent of those steroids was calculated for determining the cumulative dose.

Serum calcium, phosphorus and alkaline phosphatase were estimated by an automated analyser- Vitros Chemistry 350. Serum 25(OH) Vitamin D3 and parathyroid hormone levels were also measured by an automated machine by ELISA chemiluminescence. All the biochemical parameters were estimated at baseline, at the end of steroid therapy or three months (whichever was earlier) i.e., first follow-up and at the end of six months of study i.e., second follow-up. A HOLOGIC (Discovery QDR series S/N 84571, Hologic, USA) bone densitometer was used to perform the DEXA scan on patients and controls and APEX System Software Version 3.1 was used for data acquisition and derivation of areal BMD (aBMD). DEXA scans were performed at following three sites on each of the patients and controls- WB, Postero-anterior LS, Non-dominant DR. The densitometric measurement of the TBLH was also derived as per International Society of Clinical Densitometric Official Position statement [4]. Subject positioning was done as per manufacturer guidelines.

BMC and aBMD values were obtained as a machine generated printed report for each of the above skeletal sites. The unit used for expressing BMC was gram (gm) and that for aBMD was gram/cm². Z-score of the BMD values at each of the above four sites, was calculated for patients as well as controls using the following formula [7,18].

Z score of BMD (at specific skeletal site)=(Measured BMD-Mean BMD of control population)/Standard Deviation (SD) of BMD of control population.

In patients, DEXA scans were performed thrice viz., start of study (baseline), first (third month) and second (end of sixth month) follow-up. In order to detect asymptomatic vertebral fractures, X-rays of whole spine of patients, antero-posterior and lateral views were taken along with DEXA scan. In controls clinical, laboratory assessment and DEXA scans were performed at baseline only.

For continuous variables, mean with SD and for categorical variables, frequency with proportions were used. Scale variables between cases and controls were compared using Student’s t-test. Chi-square/Fisher’s-exact test was used for determining statistical significance between qualitative variables. Since, different physiological and pathological factors are known to have a relation with bone densitometric measurements, hence a multiple linear regression analysis was done at baseline for assessing the effect of age, sex, weight, height and BMI on BMD. Thereafter, mixed method repeated measures ANOVA was used to see the overall effect of predictor variables like age, sex and disease on different biochemical and bone densitometric parameters over time.

Within subject, pair-wise comparison over the follow-up period for different parameters was done using Bonferroni correction to avoid confounding of alpha error. The p-value <0.05 were considered as statistically significant. The statistical software IBM PASW (version 22.0) was used for entire analysis. The Pearson’s and Spearman’s correlation coefficient were applied to find correlation between bone densitometric data and cumulative dose and duration of steroid.

Of the 30 patients enrolled, seven were suffering from NS, 12 Systemic Onset Juvenile Idiopathic Arthritis (SOJIA) and 11 Tuberculous Meningitis (TBM). The mean age of the patients and controls was 9.20±3.90 years and 8.80±2.92 years, respectively (p-0.655). The cases and controls comprised of 17 and 19 males, respectively. The male to female ratio was 1.3:1 in cases and 1.7: 1 in controls, difference not being significant (p-0.278). NS patients received prednisolone at the dose of 2 mg/kg/day for six weeks (maximum 60 mg/day) followed by same dose alternate day for next six weeks as per the standard protocol. SOJIA patients received prednisolone at the dose of 2 mg/kg/day for 2-3 weeks followed by a taper over next 2-4 weeks to a minimum dose of 2.5 mg alternate day. In these patients, the duration of prednisolone treatment was calculated from the start of prednisolone till the time patients started receiving prednisolone 2.5 mg alternate day (i.e., lower than the physiological dose) [2]. Patients having TBM received prednisolone at the dose of 2 mg/kg/day initially for 2-3 weeks followed by gradual tapering over next 3-4 weeks and stopped.

Cases and controls were comparable with respect to age, and BMI [Table/Fig-1]. Amongst cases 56.66%, 16.66%, 20%, and 6.6% belonged to Tanner pubertal stage I, II, III and IV respectively whereas 63.33%, 13.33%, 20% and 3.3% of controls were in stage I, II, III, IV, respectively. Serum calcium, alkaline phosphatase and PTH level were significantly reduced in cases [Table/Fig-1]. The daily dietary calcium and vitamin D intake and mean weekly frequency of weight bearing activity was significantly greater in controls as compared to cases [Table/Fig-2]. At baseline, seven cases and nine controls had deficient 25(OH) Vit D3 level (below 20 nmol/L). They were treated with Vitamin D3 to avoid any confounding influence. At each follow-up visit, mean serum level of all the biochemical parameters were within normal range.

On multiple regression analysis and mixed method repeated measures of ANOVA, it was observed that age and sex had an association with BMD parameters over time, hence subsequent analysis was done with age and sex adjusted values of bone densitometric measurements.

At baseline, no significant differences was found between the cases and controls in the age and sex adjusted values of BMC, BMD and Z scores at all the four skeletal sites i.e., WB, TBLH, LS, and DR. Moreover, the Z scores were within the normal range of greater than-1 SD [Table/Fig-3].

Mean cumulative dose and mean duration of treatment of systemic glucocorticoids received by cases were 3.95±1.7 gm (range 1.2-7.2 gm) and 2.7±0.5 months (range 1.5-3 months), respectively.

Though BMC and BMD declined progressively at each follow-up at all the four skeletal sites, a significant reduction of BMC values of LS and DR and of BMD values of WB, LS and DR were observed at the second follow-up as compared to baseline and first follow-up [Table/Fig-4a,b]. The mean BMD of WB, LS and DR at second follow-up decreased by 3.76%, 6.94% and 6.17%, respectively [Table/Fig-4a].

At the first follow-up a significant negative correlation was observed between duration of steroids with Z score of TBLH and LS and BMD LS and between cumulative dose of steroid with BMC DR and Z score TBLH. At the second follow-up, a significant negative correlation was observed between cumulative dose of steroid and Z score TBLH and between duration of treatment and Z score TBLH [Table/Fig-5a,b].

Analysis of the disease subgroups revealed similar decrease in bone densitometric parameters at all sites over time. In patients of NS after an initial increase in BMC and BMD at WB, TBLH and LS at first follow-up, there was a decline at second follow-up. At second follow-up, at DR, BMC and BMD were significantly decreased as compared to baseline in patients with NS and at DR and LS in patients of SOJIA and TBM [Table/Fig-6a,b and c].

X-rays of whole spine (antero-posterior and lateral views) done for all cases at baseline and follow-up, did not show any vertebral fractures.

The present longitudinal study was, designed to evaluate the effect of systemic steroids of 6-12 weeks duration on BMD at four sites viz., WB, TBLH, LS and DR, using adjusted values of BMC, BMD, Z score, after accounting for all the possible factors influencing bone mass accrual (age, sex, weight, height, BMI, sexual maturity, calcium and vitamin D intake, sunlight exposure, weight bearing physical activity and disease; which can be taken as strength of the study). The study revealed few significant findings: i) bone densitometric parameters showed a gradual decline at three and six months after start of steroid therapy at all the skeletal sites, though a significant decline was observed only in BMC of DR at three months and in BMD of WB, LS and DR at six months, meaning thereby that the decline in BMD continued even after stoppage of steroids; ii) at six months, though a negative correlation was observed between the bone densitometric parameters and duration of treatment and cumulative dose of steroids, the negative correlation reached a significant level between the Z score TBLH and the duration and cumulative doe of steroid; iii) no vertebral fractures were observed during the study period. No similar study was available for comparison.

Many cross-sectional and few longitudinal studies have reported negative effect of three months or longer duration of steroids on bone health [6-23], and most of these studied the effect on LS only [10-15,18,19,21-24]. Two studies have evaluated BMD of total body, LS and DR in children [8,9] who received steroids for more than six months; one revealing lower bone mass gains at WB and DR [8] and the other showing decreased bone mass accrual at LS [9]. A decline in BMD up to six months at all four sites was observed in the present study. In contrast Phan V et al., reported an initial decline of LS BMD at three months followed by an increase at six months [14].

On sub-group analysis, it was found that in cases of NS, BMC and BMD values of WB, TBLH and LS were increased initially at first follow-up and then declined at second follow-up, whereas BMD of DR showed a consistent decline and reached significant levels at the second follow-up as compared to baseline. A logical explanation of this discordance of gain of BMD at three sites and loss at one site initially and then decline later at all sites, could not be found. Tsampalieros A et al., in contrast reported that the glucocorticoid increased cortical BMD and decreased trabecular BMD and attributed it to greater muscle mass and suppressed cortical modeling [28]. A significant decrease in spine BMD in patients of NS on steroids in the present study was in concordance to other studies [10,12-14,20-22]. Studies with larger samples would be required to clarify this discordance.

Among the cases of SOJIA, though there was decline in BMC and BMD at all four sites over time but a statistically significant decline was seen in BMC and BMD of LS and DR. The results of the present study were similar to prior studies which showed decreased BMD LS in patients with SOJIA who received steroids [7,9,11,15,17,18].

In the absence of any published study, the results of decreased bone densitometric parameters in TBM patients in the present study could not be compared. A statistically significant negative correlation was observed between cumulative dose and duration of systemic glucocorticoids and bone densitometric measurements in the present study. The negative correlation of cumulative dose of steroid and BMD was in accordance with many studies [6-8,15,17,19-21,23] but contrary to some [11,16,24]. No study could be found in literature which delineated any association of duration of steroid use and bone densitometric parameters.

Since muscle mass and muscle traction forces have also been reported to influence bone mass accrual [27], these could have been confounding variables. Authors could not completely negate these variables in statistical adjustments in the study. This could be considered as a limitation of this study.

To conclude this study was the first one to show that steroids used for a period for 6-12 weeks adversely affect bone mass accrual in both cortical and trabecular bones as revealed by decreased BMD at WB, LS, DR and TBLH, effect being more pronounced three months after stopping steroids but not to the extent of causing vertebral fractures. A negative correlation was observed between cumulative dose and duration of treatment with Z score TBLH.

More follow-up studies would be required to determine the duration up to which negative effect of steroid on bone health persists or whether duration of less than six weeks of steroid use could adversely affect bone health.