A Predictive Equation to Guide Vitamin D Dose in Patients
A Predictive Equation to Guide Vitamin D Dose in Patients
The testing volume for 25-hydroxyvitamin D increased from <300 to >12,000/year in 2007 to 2012, without meaningful change in the average serum 25- hydroxyvitamin D concentrations (Figure 1 and Table 1 ). The proportions of patients in each of the subgroups with serum 25-hydroxyvitamin D concentrations of <30, <20, <12, and >150 ng/mL were not different to any clinically meaningful extent, although there was a statistically significant (P < .05) decline in mean serum 25-hydroxyvitamin D concentrations and an increase in the proportion of patients with serum 25-hydroxyvitamin D concentrations of <12, <20, and <30 ng/mL.
(Enlarge Image)
Figure 1.
Annual volume of 25-hydroxyvitamin D testing and mean serum concentrations of 25-hydroxyvitamin D in each year. The year-to-year changes in the mean concentrations (inset) of 25-hydroxyvitamin D are statistically significantly different (P < .05). The testing volume increased from <300 to >12,000 per year without any improvement in the outcome of average serum concentrations of 25-hydroxyvitamin D, despite the providers' prescription being in keeping with recommended doses of vitamin D, suggesting that the recommended doses were inadequate.
Of the 2485 patients reviewed, 1327 (943 women, 384 men) had at least 2 serum 25-hydroxyvitamin D concentrations with documentation of treatment after the first test. We excluded 1158 patients (46.6%) from further analysis because they either had only one determination of serum concentration of 25-hydroxyvitamin D or had multiple concentrations documented but there was no evidence of a prescription for replacement vitamin D or there was documentation of a lack of compliance with treatment. A valid episode of treatment required 2 serum 25-hydroxyvitamin D concentrations with documentation of treatment between the 2 measurements. From the prescribed dose and the interval between 2 laboratory measurements of serum 25-hydroxyvitamin D, the average daily dose of vitamin D was calculated.
There were 3885 episodes of 2 vitamin D measurements with documented treatment between the 2 readings. There were an average of 2 valid episodes for each ambulatory patient and 8 for nursing home patients. Among 1552 episodes, vitamin D treatment was associated with a decrease in serum 25-hydroxyvitamin D concentration or no change. The average and median daily doses of vitamin D in this group were 1907 and 1000 IU/day, respectively. In 2333 episodes of treatments there was an increase (any increase) in serum 25-hydroxyvitamin D concentrations, and the average and median doses of vitamin D were 4707 and 4000 IU/day, respectively. An increase of ≥10 ng/mL was seen in 1236 observations; average and median daily doses of vitamin D were 5682 and 4800 IU/day, respectively. The corresponding values for ambulatory and nursing home patients are given in Table 2 .
In 68.5% episodes the serum concentration of 25-hydroxyvitamin D was <30 ng/mL before treatment. This included patients who were treated, and some had multiple cycles of treatment. After treatment, the proportion of patients with a serum 25-hydroxyvitamin D concentration <30 ng/mL was 55.3%, a drop of only 13.2 percentage points. On average, there was an increase of only 5.3 ng/mL in concentrations of serum 25-hydroxyvitamin D after treatment. The responses of the various subgroups of patients to the average daily doses are given in Table 3 .
A multiple linear regression analysis was performed to identify the best model for predicting the change from baseline to post-treatment serum concentrations of 25-hydroxyvitamin D in the 3885 valid encounters. Table 4 displays regression coefficients for the full model and reduced model, which includes only statistically significant (P < .05) predictors of change. The full model (R = 0.424; P < .001) and reduced model (R = 0.423) explained about 42% of the variability in change in serum 25-hydroxyvitamin D concentrations. The equation for predicting change in serum 25-hydroxyvitamin D concentrations (derived from the reduced model) is:
?dctmLink chronic_id='0901c7918078fc90' object_id='0901c7918078fc90' edit_widget_type=graphic??dctmEditor selectedObject='0901c7918078fc90'?
The dose required for a given desired change in serum 25-hydroxyvitamin D concentration is:
?dctmLink chronic_id='0901c7918078fc91' object_id='0901c7918078fc91' edit_widget_type=graphic??dctmEditor selectedObject='0901c7918078fc91'?
For each additional IU of vitamin D administered we anticipate a 0.002-ng/mL increase in serum 25-hydroxyvitamin D. For every additional 1.0 ng/mL of 25-hydroxyvitamin D before treatment, there will be a decrease of 0.62 ng/mL in the concentration after treatment. For every 1.0-unit increase in BMI there will be a reduction in 25-hydroxyvitamin D of 0.20 ng/mL. For every 1.0-g/dL increase in albumin we expect a 25-hydroxyvitamin D increase of 1.74 ng/mL. For every additional year of life (age) there is a 25-hydroxyvitamin D increase of 0.07 ng/mL, or, more correctly, the need for a replacement dose of vitamin D is lower with increasing serum albumin concentration and age. The same explanations apply to all of the regression analyses.
A similar multiple linear regression analysis was performed to identify the best predictive model for serum 25-hydroxyvitamin D after treatment (end). Table 5 displays the regression coefficients for the full and reduced models. Both models explained about 24% of the variability in post-treatment serum 25-hydroxyvitamin D concentrations (R = 0.236; P < .001). The equation for predicting serum 25-hydroxyvitamin D concentration after treatment is:
?dctmLink chronic_id='0901c7918078fc92' object_id='0901c7918078fc92' edit_widget_type=graphic??dctmEditor selectedObject='0901c7918078fc92'?
Bivariate comparisons were performed to determine whether the key factors used to predict serum 25-hydroxyvitamin D outcomes differed between nursing home and ambulatory patients. χ test was used to compare the sexes. Independent sample t tests were used for the continuous variables. When the comparisons between ambulatory and nursing home patients failed to meet the assumption of equality of variances, a Mann-Whitney U test was used. Table 6 displays descriptive statistics and P values for these bivariate comparisons. For each variable there was a statistically significant difference between nursing home and ambulatory encounters (P < .05 for both). Therefore, we separated the nursing home and ambulatory encounters and performed regression analyses for each subgroup.
Nursing Home Patients. The regression coefficients for the full and reduced models for predictors of change in serum 25-hydroxyvitamin D concentrations were essentially identical and explained about 60% of the variability (R = 0.595; P < .001). The equation for predicting change in serum 25-hydroxyvitamin D concentrations (derived from the reduced model) is:
?dctmLink chronic_id='0901c7918078fc93' object_id='0901c7918078fc93' edit_widget_type=graphic??dctmEditor selectedObject='0901c7918078fc93'?
We performed a multiple linear regression analysis for predicting the end concentration of serum 25-hydroxyvitamin D among the 1122 nursing home encounters. Coefficients for the full and reduced models for predictors of the end serum 25-hydroxyvitamin D concentration were virtually identical and explained about 16% of the variability (R = 0.160; P < .001).
?dctmLink chronic_id='0901c7918078fc94' object_id='0901c7918078fc94' edit_widget_type=graphic??dctmEditor selectedObject='0901c7918078fc94'?
Ambulatory Patients. A multiple linear regression analysis was performed to predict the change from baseline to post-treatment serum concentrations of 25-hydroxyvitamin D among the 2763 ambulatory encounters. The regression coefficients for the full model and the reduced model, which includes only statistically significant predictors of change in serum 25-hydroxyvitamin D concentration, were again nearly identical and explained about 36% of the variability (R = 0.364; P < .001). The equation for predicting change in serum 25-hydroxyvitamin D concentration is:
?dctmLink chronic_id='0901c7918078fc95' object_id='0901c7918078fc95' edit_widget_type=graphic??dctmEditor selectedObject='0901c7918078fc95'?
The coefficient for the dose of vitamin D is 0.003 for ambulatory patients compared with 0.002 for nursing home patients. This is in keeping with the higher doses needed for nursing home patients.
The regression coefficients for the full and reduced models of ambulatory patients were essentially similar and explained about 25% of the variability in end serum 25-hydroxyvitamin D concentration (R = 0.247; P < .001).
?dctmLink chronic_id='0901c7918078fc96' object_id='0901c7918078fc96' edit_widget_type=graphic??dctmEditor selectedObject='0901c7918078fc96'?
When multiple observations of a given patient were removed from regression analyses and only the last observation in the set was kept, the results for ambulatory patients were not meaningfully different from those presented above. For nursing home patients the small number of observations did not allow for meaningful analysis.
The comparative findings among the 3 racial groups are presented in Table 7 . The average doses resulting in (1) no increase or decease in serum concentrations of 25-hydroxyvitamin D, (2) any increase, and (3) increase of ≥10 ng/mL were not different among the 3 races to any clinically meaningful extent. The baseline serum concentrations of 25-hydroxyvitamin D before each episode of treatment also were not meaningfully different among the 3 groups.
Unstructured observations included the following:
Results
The testing volume for 25-hydroxyvitamin D increased from <300 to >12,000/year in 2007 to 2012, without meaningful change in the average serum 25- hydroxyvitamin D concentrations (Figure 1 and Table 1 ). The proportions of patients in each of the subgroups with serum 25-hydroxyvitamin D concentrations of <30, <20, <12, and >150 ng/mL were not different to any clinically meaningful extent, although there was a statistically significant (P < .05) decline in mean serum 25-hydroxyvitamin D concentrations and an increase in the proportion of patients with serum 25-hydroxyvitamin D concentrations of <12, <20, and <30 ng/mL.
(Enlarge Image)
Figure 1.
Annual volume of 25-hydroxyvitamin D testing and mean serum concentrations of 25-hydroxyvitamin D in each year. The year-to-year changes in the mean concentrations (inset) of 25-hydroxyvitamin D are statistically significantly different (P < .05). The testing volume increased from <300 to >12,000 per year without any improvement in the outcome of average serum concentrations of 25-hydroxyvitamin D, despite the providers' prescription being in keeping with recommended doses of vitamin D, suggesting that the recommended doses were inadequate.
Of the 2485 patients reviewed, 1327 (943 women, 384 men) had at least 2 serum 25-hydroxyvitamin D concentrations with documentation of treatment after the first test. We excluded 1158 patients (46.6%) from further analysis because they either had only one determination of serum concentration of 25-hydroxyvitamin D or had multiple concentrations documented but there was no evidence of a prescription for replacement vitamin D or there was documentation of a lack of compliance with treatment. A valid episode of treatment required 2 serum 25-hydroxyvitamin D concentrations with documentation of treatment between the 2 measurements. From the prescribed dose and the interval between 2 laboratory measurements of serum 25-hydroxyvitamin D, the average daily dose of vitamin D was calculated.
There were 3885 episodes of 2 vitamin D measurements with documented treatment between the 2 readings. There were an average of 2 valid episodes for each ambulatory patient and 8 for nursing home patients. Among 1552 episodes, vitamin D treatment was associated with a decrease in serum 25-hydroxyvitamin D concentration or no change. The average and median daily doses of vitamin D in this group were 1907 and 1000 IU/day, respectively. In 2333 episodes of treatments there was an increase (any increase) in serum 25-hydroxyvitamin D concentrations, and the average and median doses of vitamin D were 4707 and 4000 IU/day, respectively. An increase of ≥10 ng/mL was seen in 1236 observations; average and median daily doses of vitamin D were 5682 and 4800 IU/day, respectively. The corresponding values for ambulatory and nursing home patients are given in Table 2 .
In 68.5% episodes the serum concentration of 25-hydroxyvitamin D was <30 ng/mL before treatment. This included patients who were treated, and some had multiple cycles of treatment. After treatment, the proportion of patients with a serum 25-hydroxyvitamin D concentration <30 ng/mL was 55.3%, a drop of only 13.2 percentage points. On average, there was an increase of only 5.3 ng/mL in concentrations of serum 25-hydroxyvitamin D after treatment. The responses of the various subgroups of patients to the average daily doses are given in Table 3 .
Predicting Change in Serum Concentrations of 25-Hydroxyvitamin D From Before to After Treatment
A multiple linear regression analysis was performed to identify the best model for predicting the change from baseline to post-treatment serum concentrations of 25-hydroxyvitamin D in the 3885 valid encounters. Table 4 displays regression coefficients for the full model and reduced model, which includes only statistically significant (P < .05) predictors of change. The full model (R = 0.424; P < .001) and reduced model (R = 0.423) explained about 42% of the variability in change in serum 25-hydroxyvitamin D concentrations. The equation for predicting change in serum 25-hydroxyvitamin D concentrations (derived from the reduced model) is:
?dctmLink chronic_id='0901c7918078fc90' object_id='0901c7918078fc90' edit_widget_type=graphic??dctmEditor selectedObject='0901c7918078fc90'?
The dose required for a given desired change in serum 25-hydroxyvitamin D concentration is:
?dctmLink chronic_id='0901c7918078fc91' object_id='0901c7918078fc91' edit_widget_type=graphic??dctmEditor selectedObject='0901c7918078fc91'?
For each additional IU of vitamin D administered we anticipate a 0.002-ng/mL increase in serum 25-hydroxyvitamin D. For every additional 1.0 ng/mL of 25-hydroxyvitamin D before treatment, there will be a decrease of 0.62 ng/mL in the concentration after treatment. For every 1.0-unit increase in BMI there will be a reduction in 25-hydroxyvitamin D of 0.20 ng/mL. For every 1.0-g/dL increase in albumin we expect a 25-hydroxyvitamin D increase of 1.74 ng/mL. For every additional year of life (age) there is a 25-hydroxyvitamin D increase of 0.07 ng/mL, or, more correctly, the need for a replacement dose of vitamin D is lower with increasing serum albumin concentration and age. The same explanations apply to all of the regression analyses.
Predicting End (Post-treatment) Serum 25-Hydroxyvitamin D Concentration
A similar multiple linear regression analysis was performed to identify the best predictive model for serum 25-hydroxyvitamin D after treatment (end). Table 5 displays the regression coefficients for the full and reduced models. Both models explained about 24% of the variability in post-treatment serum 25-hydroxyvitamin D concentrations (R = 0.236; P < .001). The equation for predicting serum 25-hydroxyvitamin D concentration after treatment is:
?dctmLink chronic_id='0901c7918078fc92' object_id='0901c7918078fc92' edit_widget_type=graphic??dctmEditor selectedObject='0901c7918078fc92'?
Nursing Home versus Ambulatory Patients
Bivariate comparisons were performed to determine whether the key factors used to predict serum 25-hydroxyvitamin D outcomes differed between nursing home and ambulatory patients. χ test was used to compare the sexes. Independent sample t tests were used for the continuous variables. When the comparisons between ambulatory and nursing home patients failed to meet the assumption of equality of variances, a Mann-Whitney U test was used. Table 6 displays descriptive statistics and P values for these bivariate comparisons. For each variable there was a statistically significant difference between nursing home and ambulatory encounters (P < .05 for both). Therefore, we separated the nursing home and ambulatory encounters and performed regression analyses for each subgroup.
Nursing Home Patients. The regression coefficients for the full and reduced models for predictors of change in serum 25-hydroxyvitamin D concentrations were essentially identical and explained about 60% of the variability (R = 0.595; P < .001). The equation for predicting change in serum 25-hydroxyvitamin D concentrations (derived from the reduced model) is:
?dctmLink chronic_id='0901c7918078fc93' object_id='0901c7918078fc93' edit_widget_type=graphic??dctmEditor selectedObject='0901c7918078fc93'?
We performed a multiple linear regression analysis for predicting the end concentration of serum 25-hydroxyvitamin D among the 1122 nursing home encounters. Coefficients for the full and reduced models for predictors of the end serum 25-hydroxyvitamin D concentration were virtually identical and explained about 16% of the variability (R = 0.160; P < .001).
?dctmLink chronic_id='0901c7918078fc94' object_id='0901c7918078fc94' edit_widget_type=graphic??dctmEditor selectedObject='0901c7918078fc94'?
Ambulatory Patients. A multiple linear regression analysis was performed to predict the change from baseline to post-treatment serum concentrations of 25-hydroxyvitamin D among the 2763 ambulatory encounters. The regression coefficients for the full model and the reduced model, which includes only statistically significant predictors of change in serum 25-hydroxyvitamin D concentration, were again nearly identical and explained about 36% of the variability (R = 0.364; P < .001). The equation for predicting change in serum 25-hydroxyvitamin D concentration is:
?dctmLink chronic_id='0901c7918078fc95' object_id='0901c7918078fc95' edit_widget_type=graphic??dctmEditor selectedObject='0901c7918078fc95'?
The coefficient for the dose of vitamin D is 0.003 for ambulatory patients compared with 0.002 for nursing home patients. This is in keeping with the higher doses needed for nursing home patients.
The regression coefficients for the full and reduced models of ambulatory patients were essentially similar and explained about 25% of the variability in end serum 25-hydroxyvitamin D concentration (R = 0.247; P < .001).
?dctmLink chronic_id='0901c7918078fc96' object_id='0901c7918078fc96' edit_widget_type=graphic??dctmEditor selectedObject='0901c7918078fc96'?
When multiple observations of a given patient were removed from regression analyses and only the last observation in the set was kept, the results for ambulatory patients were not meaningfully different from those presented above. For nursing home patients the small number of observations did not allow for meaningful analysis.
The comparative findings among the 3 racial groups are presented in Table 7 . The average doses resulting in (1) no increase or decease in serum concentrations of 25-hydroxyvitamin D, (2) any increase, and (3) increase of ≥10 ng/mL were not different among the 3 races to any clinically meaningful extent. The baseline serum concentrations of 25-hydroxyvitamin D before each episode of treatment also were not meaningfully different among the 3 groups.
Unstructured observations included the following:
The recommended dose of 800 IU/day for nursing home residents and ambulatory patients is generally inadequate for maintaining normal serum concentrations of 25-hydroxyvitamin D. An example of such an observation in a nursing home patient is shown in Figure 2.
Acute illnesses tend to deplete serum 25-hydroxyvitamin D concentrations, and despite documented deficiency of serum 25-hydroxyvitamin D and hypocalcaemia, acutely ill patients often did not receive supplemental vitamin D.
Increase in weight tended to reduce serum concentrations of 25-hydroxyvitamin D; the reverse was also true.