Introduction

Fragility fractures are common, with approximately 4.3 million new fractures in the European Union (EU) annually [1]. Fragility fractures are associated with severe morbidity, increased mortality, high socioeconomic costs, and an increased risk of new imminent fractures [1, 2]. The risk of imminent fractures could be effectively reduced with readily available preventative treatments [3,4,5].

Denosumab is an antibody that effectively inhibits osteoclasts, resulting in increased bone mineral density and a reduced risk of fragility fractures [4, 6]. Hypocalcemia is a known adverse event [7,8,9,10], especially in patients with calcium-associated diseases, such as in secondary hyperparathyroidism, where the body is dependent on bone as a source of calcium to maintain adequate blood levels [11].

The prevalence of hypocalcemia after denosumab injection was low in the original clinical trial (Fracture REduction Evaluation of Denosumab in Osteoporosis Every 6 Months, FREEDOM trial), at < 1% [4, 6]; however, reports on real-world data indicate a higher prevalence of 4–26% [7,8,9,10]. Methods to assess post-injection hypocalcemia vary, and in previous large-scale studies, calcium was not routinely analyzed directly after injections. Instead, study protocols included calcium analysis every 6 months, i.e., before injection and at predefined intervals (3, 6, 12, 24, and 36 months), or included any non-routine calcium value taken at predefined intervals [4, 6,7,8, 10, 12]. These methods can miss the early transient post-injection hypocalcemia phase [9, 13, 14]. Furthermore, non-routine calcium values might be biased in how the patients were selected for calcium analysis. These study designs could include a risk of both under- and overestimation of the actual prevalence.

In the clinical context, from 2011 until the introduction of the new guidelines in 2021, our local guidelines recommended to analyze ionized calcium (free bioactive calcium) before and 2 weeks after each denosumab injection for all patients, independent of known risk factors. In the present study, we analyzed the long-term real-world data of this unselected osteoporosis cohort.

The purpose of the study was to assess: 1) the prevalence of hypocalcemia after denosumab in a cohort routinely followed up with calcium measurements, 2) the risk factors for developing hypocalcemia, and 3) the risk factors for developing severe hypocalcemia.

Methods

Study design

We performed a retrospective observational study in which all patients administered at least one injection of denosumab at the osteoporosis unit at Linköping University Hospital, Sweden, between 2011 and 2020 were considered. From this cohort, we included all individual denosumab injections fulfilling the inclusion criteria: 1) ionized calcium analyzed pre-injection, and 2) ionized calcium analyzed a maximum of 31 days post-injection. The patients at the osteoporosis unit were mostly primary health care patients; the osteoporosis unit only supported the primary health care centers with injection administration, including peri-injection laboratory testing. The local guidelines during the study period recommended analyzing ionized calcium before and 2 weeks after each denosumab injection. In addition, creatinine estimated glomerular filtration rate (GFR) measurement was recommended at every second injection, i.e., once a year. Other non-routine laboratory tests were performed, most commonly at the first injection time point. Before patients were sent from primary health care, they were supposed to have undergone minor laboratory testing to exclude common secondary causes of osteoporosis.

Method

Patients treated with denosumab at the osteoporosis unit were identified using the local database connected to the Cosmic intelligence software case record system (Cambio, Linköping, Sweden). The study period was 2011 to 2020, and patients were identified using three criteria: 1) appointment coded “denosumab” in the digital case record planning system; 2) appointment subclassified “osteoporosis unit” (i.e., one of three units at the clinic) in the booking schedule; and 3) patient’s visit was coded with the procedure code DT021 (i.e., “drug administration subcutaneously”) in the case record. For the included patients, all separate injection time points were considered for inclusion.

Descriptive data, such as age and sex, as well as laboratory values were extracted from the digital case record system. A limit of 31 days post-injection was set for the calcium value to be considered as being related to the treatment. Furthermore, the GFR (pre-injection), phosphate (pre-injection), parathyroid hormone (PTH) (pre-injection), magnesium (pre-injection), and 25-hydroxy (OH) vitamin D (peri-injection) were recorded when available. The estimated GFR was based on MDRD (Modification of Diet in Renal Disease Study) formula and was calculated at the laboratory. PTH values were not included if the patient was diagnosed with primary hyperparathyroidism.

Blood samples were taken at any laboratory facility in the region, including central laboratories and primary health care units. The analysis was centralized, and transportations followed procedures used in routine care. The normal range of ionized calcium at our laboratory is defined as 1.15–1.33 mmol/L. Analysis of ionized calcium was performed with Biossays 240plus/E6, and the normal range was obtained from the literature [15]. The coefficient of variation (CV%) for the analysis was 2%. In accordance with clinical reference values, an ionized calcium level < 1.15 mmol/L was considered hypocalcemic. Furthermore, in some analyses, we subcategorized hypocalcemia into groups: 1.10–1.14 mmol/L (mild), 1.05–1.09 mmol/L (moderate), and < 1.05 mmol/L (severe).

Statistical analysis

IBM SPSS 28 for Windows software (IBM Statistics, New York, USA) was used for the statistical analysis. Descriptive data are presented as the mean ± standard deviation (SD), median and interquartile range (IQR), or percentages. For continuous variables, the Student’s t-test was used to compare groups. Furthermore, Pearson correlation and univariate and multiple logistic regression analyses were performed. A p value ≤ 0.05 was considered statistically significant.

Results

Descriptive data

In total, 242 unique individuals were included, with a total of 1096 injections administered (1–15 injections per patient) (Fig. 1). The mean age for the first injection was 74 ± 10 years (median: 75 [68–81]), and 88% were female. Pre-injection hypocalcemia (< 1.15 mmol/L) and hypercalcemia (> 1.33 mmol/L) was observed in 1.6% and 7.5% of injections, respectively. The drop-out rate, i.e., injections not included due to missing ionized calcium (before and/or after injection) was 29% (n = 459). There was no gender difference between included and not included injections (female 88% vs 89%, p = 0.452). In those cases where the pre-injection calcium data was available (n = 438), the level did not differ from the pre-injection calcium in those included (1.26 ± 0.05 vs 1.25 ± 0.06 mmol/L, p = 0.279), nor did the GFR (n = 246) (67.3 ± 25.6 vs 65.6 ± 26.2 [mL/min]/1.73 m2, p = 0.881).

Fig. 1
figure 1

Denosumab injection disposition. GFR = glomerular filtration rate, PTH = parathyroid hormone

Full size image

Laboratory data, in addition to ionized calcium, were available for 60% of injections regarding creatinine GFR, 14% regarding magnesium, 6% regarding 25-OH vitamin D, 5% regarding phosphate, and 4% regarding PTH. The mean creatinine GFR was 64 ± 26 [mL/min]/1.73 m2 (median: 63 [44–80]). When grouping patients according to the chronic kidney disease (CKD) criteria [16], 12% were classified as stage 1 (≥ 90 [mL/min]/1.73 m2), 44% as stage 2 (60–89 [mL/min]/1.73 m2), 18% as stage 3a (45–59 [mL/min]/1.73 m2), 20% as stage 3b (30–44 [mL/min]/1.73 m2), and 6% as stage 4/5 (< 30 [mL/min]/1.73 m2). The mean magnesium was 0.82 ± 0.11 mmol/L (median: 0.83 [0.77–0.88]), and 8% had a magnesium value < 0.7 mmol/L, i.e., lower than the reference value. The mean PTH was 6.5 ± 3.5 pmol/L (median: 5.5 [3.9–8.0]), and 25% had a PTH level > 6.9 pmol/L. The mean 25-OH vitamin D was 72 ± 28 nmol/L (median: 72 [55–88]), and 17% had a value ≤ 50 nmol/L and 3% ≤ 25 nmol/L. The mean phosphate value was 1.03 ± 0.23 mmol/L (median: 1.1 [0.9–1.2]), and 13% had low phosphate value, i.e., < 0.8 mmol/L for women and < 0.75 mmol/L for men.

Frequency of post-injection hypocalcemia

Post-injection hypocalcemia was observed at least once in 20% (n = 48) of the unique patients (14% mild, 3% moderate, and 4% severe hypocalcemia). Most of those patients (7 out of 10) were given repeated denosumab injections without the reoccurrence of hypocalcemia.

When analyzing the individual injections separately (n = 1096), hypocalcemia occurred in 6.3% of all injections (4.6% mild, 0.6% moderate, and 1.1% severe). Excluding patients with a high calcium value before injection gave marginally higher results, i.e., 6.7% post-injection hypocalcemia (4.8% mild, 0.7% moderate, and 1.2% severe). The mean duration between the injection and measurement of calcium was 15.1 ± 4.1 days (median: 14.0 [13,14,15,16], range: 1–31), with hypocalcemia detected in all phases (Supplement 1). As shown in Fig. 2, hypocalcemia occurred independent of the number of injections when given repeatedly to the same patient (rate of hypocalcemia varied from 3–8%).

Fig. 2
figure 2

Hypocalcemia frequencies at different injection time points. Hypocalcemia was subdivided into three subgroups: mild (ionized calcium 1.10–1.14 mmol/L); moderate (ionized calcium 1.05–1.09 mmol/L); and severe (ionized calcium ≤ 1.04 mmol/L). Inj. = Injection number. n = is the total number of patients given those numbers of injections

Full size image

Risk factors for hypocalcemia

The pre-injection calcium level correlated positively with the post-injection calcium level (r = 0.558, p < 0.001) (Fig. 3). The mean change in ionized calcium post-injection was − 0.02 ± 0.05 mmol/L (range: –0.30–0.16). Of all patients with post-injection hypocalcemia, only 9% had a low pre-injection calcium value as well (9% sensitivity).

Fig. 3
figure 3

Relation between calcium values taken before and after denosumab injection. In the lower squares are patients with post-injection hypocalcemia. The bottom right shows patients with normal or high pre-injection calcium but low post-injection values. Hypocalcemia was defined as an ionized calcium level < 1.15 mmol/L, which is the reference value at the study sites’ routine laboratory. Green area includes patients with pre-injection calcium in the normal range (1.15–1.33 mmol/L)

Full size image

As shown in Table 1, post-injection hypocalcemia was significantly more common in men than in women, as well as in patients with pre-injection hypocalcemia, hypomagnesemia, hypophosphatemia, and those in severe renal failure (CKD stage 4–5). Stage 3a CKD (GFR 45–59 [mL/min]/1.73 m2) did not increase the hypercalcemia risk significantly. Stage 3b (GFR 30–44 [mL/min]/1.73 m2) showed borderline significance (p = 0.064). The higher frequency in males compared to females remained after adjusting for renal failure and age.

Table 1 Risk factors for post-injection hypocalcemia
Full size table

The highest prevalence of post-injection hypocalcemia was seen in patients with pre-injection hypophosphatemia, hypomagnesemia, or hypocalcemia, where 57%, 38%, and 33% developed post-injection hypocalcemia, respectively. However, the numbers of observations in the groups were low, and the results should be interpreted cautiously. There may be bias regarding which patients these tests were performed in. Regarding phosphate, there was a negative association of phosphate values with pre-injection calcium (–0.266, p = 0.049), PTH (–0.437, p = 0.01), and creatinine GFR (–0.334, p = 0.017), suggesting that these risk individuals are derived from the same cohort. Magnesium, however, did not correlate with other measured laboratory variables.

Severe hypocalcemia (Table 1) was associated with male sex, renal failure (CKD stage 4–5), low vitamin D, and hypophosphatemia. No patient with severe hypocalcemia had magnesium insufficiency.

Discussion

In the present study, using long-term real-world data on denosumab-treated osteoporosis patients followed up routinely regarding peri-injection calcium, we show that one in five patients had experienced post-injection hypocalcemia at least once during the treatment period. Most commonly, however, hypocalcemia was mild and not recurrent. Severe post-injection hypocalcemia, defined as an ionized calcium < 1.05 mmol/L, was seen after 1% of injections, of which a few patients needed inpatient care or an acute open care visit (n = 5). The risk factors for post-injection hypocalcemia were renal failure (CKD stage 4/5), pre-injection hypocalcemia, male sex, hypophosphatemia, hypomagnesemia, and vitamin D insufficiency.

The prevalence of hypocalcemia was higher in our study (6%) than in the original clinical trial (FREEDOM), which reported a prevalence of 0–0.1% [4, 6]. A low prevalence of hypocalcemia (< 1% hypocalcemia) was also reported in the FREEDOM sub-analysis of patients with renal failure and in a post-marketing real-world prospective observational trial [12, 17]. Our findings are in the same range as other large-scale studies on real-world cohorts, which reported the frequency of denosumab-induced hypocalcemia to range from 4 to 8% [7, 8, 10].

Previous hypocalcemia data in large cohorts, including clinical trials, was mainly based on calcium analysis before consecutive injections (e.g., every 6 months) or other predefined intervals not consequently taken shortly after injection [6,7,8, 12]. As post-injection hypocalcemia is mostly seen in the weeks following injection [9, 13, 14], there is a risk of underestimation of the prevalence of hypocalcemia if longer intervals between injection and calcium analysis are allowed. Another explanation for the difference is that patients are more systematically controlled in clinical trials, and possible insufficiencies of calcium and vitamin D will either exclude the patient from the study or delay inclusion until adequately supplemented. Furthermore, patients included in a clinical trial are probably more adherent to supplements, e.g., calcium and vitamin D supplements, which might also affect the lower hypocalcemia frequency observed in these studies. Additionally, previous recommendations, also used in the FREEDOM trial, were 1000 mg calcium supplementation daily, which is twice as high as our national recommendations, which might have affected the results in patients with borderline calcium sufficiency. Accordingly, previous studies highlight the importance of adequate calcium supplements to prevent denosumab-induced hypocalcemia [7, 13]. This could raise the question of whether the general recommendation of 500 mg calcium/day is enough. Whether a higher dose should be recommended to patients on denosumab or preferentially to patients with certain risk profiles remains to be determined.

Using non-routine data, i.e., calcium analysis only if doctors ask for it [17], might include selection bias where only high-risk patients are followed up with calcium measurements. This might yield an overestimation of the prevalence of hypocalcemia, or it might risk missing cases of mild hypocalcemia.

Most previous studies analyzed total calcium, which includes not only the biologically active form (ionized calcium) but also albumin bound and other calcium complexes [11]. In contrast, we analyzed ionized calcium directly, resulting in the assessment of biologically active calcium being more precise [11, 18] and the rate of post-injection hypocalcemia more reliable.

Renal failure is a known risk factor for post-injection hypocalcemia [7, 8, 13, 19], which is in line with our results of a clearly elevated risk of hypocalcemia in patients with severe renal failure, i.e., CKD 4 and 5 (GFR < 30 [mL/min]/1.73 m2), and a tendency through non-significant in patients with CKD 3b (GFR 30–44 [mL/min]/1.73 m2). In contrast, CKD 3a (GFR 45–60 [mL/min]/1.73 m2) did not increase risk of hypocalcemia in our study. In renal failure, conversion of 25-OH vitamin D to the active form calcitriol is impaired, rendering a risk of negative calcium balance where calcium uptake from the intestine is insufficient and the body is dependent on the bone for maintaining normal blood calcium levels. Clinically this is seen by secondary hyperparathyroidism with low or normal (compensated) blood calcium values. In this scenario, an effective antiresorptive agent, such as denosumab, blocks the necessary efflux from bone, resulting in hypocalcemia. To decrease this risk, it is important to adjust the calcium balance before treatment, e.g., with calcitriol, as well as to ascertain adequate calcium supplements.

In our study we found a higher risk of post-injection hypocalcemia in men than in women. Similar findings have been reported previously [7, 8]; however, in a post-analysis of previous data, it was concluded that this was more likely secondary to selection bias [20]. Despite correcting for renal failure and age in our study, the gender difference remained. Whether this is an actual gender difference or cofactors that were not measured and adjusted for in our study is unclear.

Pre-injection hypocalcemia has previously been reported as a risk factor for hypocalcemia [7, 8]. However, in our cohort, pre-injection hypocalcemia was not significantly associated with severe cases of hypocalcemia, and most patients of post-injection hypocalcemia had normal pre-injection calcium levels, thus weakening pre-injection hypocalcemia as a single clinical risk marker to identify high-risk patients (sensitivity: 9%).

In our study, we included measurements of magnesium and phosphate; both minerals are tightly connected to the calcium balance [21, 22]. Low pre-injection values of magnesium or phosphate imposed a substantially increased risk of post-injection hypocalcemia. This could reflect signs of malnutritional status in patients with hypocalcemia, where calcium levels are dependent on calcium from bone. Some of the patients had intestinal diseases with a risk of malabsorption as well as individuals with a clinical suspicion of malnutrition. In addition, one patient had a phosphate-associated disease. Both magnesium and phosphate minerals are easy and inexpensive to analyze, and the present study indicates that they might be valuable markers to predict the risk of post-injection hypocalcemia. However, in the present study, these minerals were not routinely evaluated in all patients, and, especially regarding phosphate, there were few observations (5% of injections) available. Thus, it would be valuable to confirm the findings in a larger cohort.

Post-injection hypocalcemia not only occurred after the first injection, the prevalence was similar throughout the treatment period (3–8%). This is in concordance with the study by Tsvetov et al. [8]. The finding of unchanged prevalence rates seems feasible as risk factors, e.g., renal failure progression and malnutrition, might develop over time as patients age. On the other hand, most patients with post-injection hypocalcemia did not have recurrent hypocalcemia, which was possibly caused by a more alert preventive approach on the following injections, e.g., use of extra calcium and vitamin D as well as other preventive treatments such as nutrition optimization.

Severe hypocalcemia was rare, i.e., following 1% of all injections. In five cases, patients were admitted to the hospital or required an acute outpatient visit after the injection. These patients had a complicated disease history, including severe kidney disease, intestinal disease, and/or malnutrition.

Measuring post-injection laboratory values are costly, including analysis costs and organizational costs with staff dedicated to organizing testing and follow-up of the results. Furthermore, the extra follow-up causes inconvenience for the patients and might involve relatives/personal to assist as there will be an extra visit to the laboratory and phone contact with nurse. Therefore, although the risk of hypocalcemia needs to be addressed and high-risk individuals, as described above, need to be identified, the vast majority (94%) of injections were not followed by hypocalcemia. Thus, post-injection calcium screening does not seem motivated in general.

Furthermore, the clinical significance of mild and moderate post-injection hypocalcemia is of uncertain clinical significance. Some cases of mild hypocalcemia may even reflect a normal biological variance, i.e., when there is a small decrease in calcium from the normal reference. We do not have a least significant change (LSC) value or reference change value (RCV) for ionized calcium. Assuming that it behaves in a similar manner to total calcium, we might expect an RCV of approximately 6%. When applying this to our data, 21 cases of mild hypocalcemia could possibly reflect normal variation. Excluding these from the analysis would yield a total post-injection hypocalcemia frequency of 4.5% instead of 6.3% (rates of moderate and severe hypocalcemia are not altered). However, this assumes similar behavior in the two calcium analysis methods.

Similar to previous reports, we believe that patients with moderate and severe renal failure need to be followed up more carefully, possibly including routine post-injection calcium measurements. Other risk groups identified are patients with malnutrition/malabsorption. In both risk groups, adequate pretreatment is important to ascertain a balanced calcium system that is not dependent on calcium from bone. Thus, it is necessary to administer adequate calcium and vitamin D3 supplementation as well as in renal failure caltiriol treatment if required. Chronic magnesium insufficiency may blunt the PTH response and thus disable the compensation mechanism in hypocalcemia, which is why adequate magnesium is important. Further studies on the value of including magnesium, phosphate, 25-OH vitamin D, and PTH in the pre-testing would be valuable.

Strengths and limitations

In our study, pre- and post-injection calcium values were available for all injection time points, whereas the other laboratory values were taken less frequently (e.g., creatinine GFR, which was recommended once every 12 months, and magnesium, often before the first injection) or non-routinely (e.g., phosphate, 25-OH-vitamin D). Therefore, there is a risk of both over- and underestimating the impact of these factors. Phosphate was only taken in 5% of cases, including one patient with a known phosphate disturbance, which could affect the result. We chose to include all patients as we believe that this best mirrors a real-world cohort. The low and non-systematically measured values of vitamin D and PTH is a clear limitation and risks reducing the power of the data as well as creating selection bias. Other limitations of this study include the lack of data on symptoms and dosing of calcium and vitamin D supplements at each injection time point. The former was not systematically included in the case records, and the latter could not be extracted retrospectively from the case records.

The strengths of the study are the study design, including a typical real-world osteoporosis cohort, which results in the data being generalizable. In addition, the study design, with routinely analyzed post-injection calcium measurements in a well-defined time-period between the injection and follow-up value (i.e., maximum of 31 days), thus avoiding selection bias, is a strength.

Conclusion

Post-injection hypocalcemia was observed following 6% of denosumab injections, independent of the injection number. Severe post-injection hypocalcemia was rare, with severe renal failure and nutritional status seemingly important predictive factors. Magnesium and phosphate might add value in the pre-injection risk assessment; however, the findings need to be confirmed in a larger cohort.