Predictors for development of denosumab-induced hypocalcaemia in cancer patients with bone metastases determined by ordered logistic regression analysis

This retrospective study was undertaken to identify predictors for the development of hypocalcaemia even with prophylactic administration of calcium and vitamin D, and to help guide future strategies to improve the safety, efficacy, and QOL of patients receiving denosumab. Between January 2016 and February 2020, a total of 327 advanced cancer patients at our hospital who were receiving denosumab were enrolled. Variables associated with the development of hypocalcaemia were extracted from the clinical records. The level of hypocalcaemia was evaluated using CTCAE version 5. Multivariate ordered logistic regression analysis was performed to identify predictors for the development of hypocalcaemia. Optimal cut off thresholds were determined using ROC analysis. Values of P < 0.05 (2-tailed) were considered significant. 54 patients have developed hypocalcemia (≥ Grade 1). Significant factors identified included concomitant use of vonoprazan [odds ratio (OR) = 3.74, 95% confidence interval (CI) 1.14–12.26; P = 0.030], dexamethasone (OR = 2.45, 95%CI 1.14–5.42; P = 0.022), pre-treatment levels of serum calcium (OR = 0.27, 95%CI 0.13–0.54; P < 0.001), ALP/100 (OR = 1.04, 95%CI 1.01–1.07; P = 0.003), and haemoglobin (OR = 0.79, 95%CI 0.68–0.93; P = 0.004). ROC curve analysis revealed that the threshold for pre-treatment levels of serum calcium was ≤ 9.3 mg/dL, ALP was ≥ 457 U/L, and haemoglobin was ≤ 10.4 g/dL. In conclusion, concomitant use of vonoprazan or dexamethasone, and pre-treatment levels of serum calcium (low), ALP (high) and haemoglobin (low) were identified as significant predictors for the development of denosumab-induced hypocalcaemia.


Results
Of the 348 patients who received denosumab, 21 were excluded from this study (no preventive medication for hypocalcaemia, n = 1; insufficient data, n = 20). Table 1 presents the clinical characteristics of the 327 enrolled patients, the potential variables related to the development of hypocalcaemia, and the results of univariate analyses. The forward stepwise selection procedure identified five variables (concomitant use of vonoprazan, dexamethasone, calcium level at the start of denosumab, ALP and haemoglobin). Multivariate ordered logistic regression analysis was performed using these variables. Significant factors identified for the development of hypocalcaemia included concomitant use of vonoprazan, dexamethasone, pre-treatment levels of serum calcium, ALP/100, and haemoglobin ( Table 2). The ROC curve analysis of the group likely to develop hypocalcaemia (≥ Grade 1) revealed concomitant use of vonoprazan with a sensitivity of 11.1% and specificity of 97.1% (AUC = 0.54) (Fig. 1A), dexamethasone with a sensitivity of 27.8% and specificity of 87.9% (AUC = 0.58) (Fig. 1B). The ROC curve analysis revealed that the threshold for the calcium level at the start of denosumab was ≤ 9.3 mg/ dL, with 63.0% sensitivity and 76.6% specificity [area under the curve (AUC) = 0.72] (Fig. 1C), ALP threshold Table 1. Patient characteristics, extracted variables, and results of univariate analyses (n = 327). CI, confidence interval; BMI, body mass index; BSA, body surface area. *P < 0.05. **Evaluated by National Cancer Institute Common Terminology Criteria for Adverse Events (NCI-CTCAE) version 5. ***Odds ratio per ALP 100 U/L.  (Fig. 1D), and haemoglobin threshold was ≤ 10.4 g/dL with 38.9% sensitivity and 83.4% specificity (AUC = 0.65) (Fig. 1E).

Discussion
The multivariate ordered logistic regression analysis performed in this study showed that significant predictors for the development of denosumab-induced hypocalcaemia even when receiving prophylactic administration of calcium and vitamin D included concomitant use of vonoprazan and dexamethasone, and pre-treatment levels of serum calcium, ALP and haemoglobin. On ROC curve analysis of the potential factors responsible for the development of hypocalcaemia, the cut-off value for the calcium level at the start of denosumab was ≤ 9.3 mg/ dL, that of ALP was ≥ 457 U/L, and that of haemoglobin was ≤ 10.4 g/dL. For concomitant drugs, vonoprazan and dexamethasone were extracted as significant factors in this study. Previous research has reported that co-administration of PPIs was significantly related to hypocalcaemia [5][6][7] .
Regarding the relationship between combined use of PPIs and the occurrence of hypocalcaemia, an increase in gastric pH due to the inhibitory effect of PPIs on gastric acid secretion has been reported to decrease the absorption of calcium, which becomes soluble under acidic conditions [5][6][7] . In this study, among the PPIs, only vonoprazan was extracted as a significant factor. Conventional PPIs are unstable in the acidic environment of the stomach and do not remain for a long time. On the other hand, vonoprazan is said to be stable even in acidic environments, and so can remain for a long time and exert sustained activity 11,12 . This may be one reason why vonoprazan alone was extracted as significantly associated with hypocalcaemia among the PPIs investigated in this study.
Steroids reduce calcium absorption in the intestine and kidney, and have been associated with osteoporosis as an adverse effect. In this study, among steroids, only dexamethasone was extracted as a significant factor. Dexamethasone is often used for an extended period as an antiemetic 13 and for prevention of edema during chemotherapy 14 , and shows a longer duration of action than prednisolone. These may be the reasons why dexamethasone was extracted as a risk factor. Further verification is needed regarding this issue.
Low pre-treatment levels of serum calcium have been reported as a risk factor for hypocalcaemia 15 . Our result was also in agreement with those of previous studies. ROC analysis identified a cut-off value of 9.3 mg/ dL (≥ grade 1). Clinicians should pay close attention to the onset of hypocalcaemia after administration of denosumab among patients with pre-treatment levels of serum calcium ≤ 9.3 mg/dL and should consider the administration of medications to prevent hypocalcaemia before denosumab administration.
In the current study, pre-treatment level of ALP was identified as a risk factor, and ROC analysis showed a cut-off value of ALP ≥ 457 U/L. Previous study discussed elevated bone-specific alkaline phosphatase levels may indicate potential calcium deposition in osteoid and undermineralised bone matrix, a phenomenon that can persist for weeks or months after osteoclast inhibition 10 . Our result is consistent with previous findings 9, 10,16 . Clinicians thus need to know about the incidence and severity of hypocalcaemia, especially in patients with pre-treatment levels of ALP ≥ 457 U/L.
Pre-treatment level of haemoglobin was also identified as a risk factor. As a result of ROC analysis, the cut-off value for haemoglobin was ≤ 10.4 g/dL. Anaemia is also a risk factor for electrolyte abnormalities 17 . In addition, the low haemoglobin level was also influenced by the nutritional status as a patient background factor, suggesting that hypocalcaemia was also influenced by decreases in dietary calcium intake and absorption 18 . Care regarding the onset of hypocalcaemia is thus warranted in patients with anaemia before denosumab administration.
On the other hand, renal dysfunction before administration of denosumab has been reported as a risk factor for hypocalcaemia 19 , but was not extracted as a risk factor in this study. This may be because the number of patients with creatinine clearance of National Cancer Institute Common Terminology Criteria for Adverse Events (NCI-CTCAE) version 5 20 grade 3 or higher were a clear minority, representing only 9 of the 327 patients. NCI-CTCAE is a standard assessment tool of treatment-related adverse events.
Several limitations to the current study need to be considered. First, the retrospective nature of the study may have decreased the validity of the data obtained. Second, since this study was performed at a single institute, prospective multicentre studies are needed to confirm these results.
In conclusion, concomitant use of vonoprazan or dexamethasone, low calcium level before denosumab administration, and high ALP and low haemoglobin were identified as significant predictors for the development of hypocalcaemia in cancer patients with bone metastasis. However, our findings need to be confirmed in Table 2. Results of multivariate ordered logistic regression analysis for variables extracted by forward selection (n = 327). CI, confidence interval. *P < 0.05. **Odds ratio per ALP 100 U/L.   Statistical analysis. Independent variables were analysed for multicollinearity (correlation coefficient |r| ≥ 0.7), since when correlations exist among the variables, this can lead to unreliable and unstable results of regression analyses. Independent variables were extracted based on correlation strength with the level of the denosumab-induced hypocalcaemia (dependent variable) or clinical significance. First, univariate ordered logistic regression analysis between the outcomes and each potential independent variable was performed. Subsequently, a multivariate ordered logistic regression model was constructed by employing the forward stepwise selection procedure with the potential candidate variables. The model used a variable entry criterion of 0.15 and a variable retention criterion of 0.1. Ordered logistic regression analysis was employed, because the level of hypocalcaemia was evaluated by a graded scale and multiple factors really associated as predictors for the development of denosumab-induced hypocalcaemia had to be analysed concurrently. Optimal cut off thresholds were determined using receiver operating characteristic curve (ROC) analysis 21 . For all statistical analyses, values of P < 0.05 (2-tailed) were considered significant. All analyses were performed using JMP version 14.3.0. (SAS Institute, Cary, NC).