Associations of cumulative voriconazole dose, treatment duration, and alkaline phosphatase with voriconazole-induced periostitis

This retrospective study was approved by the UCSF institutional review board. A search of radiology reports was performed using a clinical informatics tool (mPower, Nuance Communications Inc, Burlington, MA) with the search terms (voriconazole OR vori OR vfend). One hundred and thirty-five patients with reported voriconazole use were identified. Data collected from the electronic health records included prescribed voriconazole dose, frequency, route of administration, and treatment duration. Age at voriconazole initiation, sex, immune status, alkaline phosphatase (ALP), and voriconazole levels during the duration of treatment were also collected. Patients were defined as immunocompromised if any of the following applied during voriconazole treatment: history of malignancy, solid-organ or bone marrow transplant, HIV infection, genetic immunodeficiency, diabetes mellitus, use of steroids or other immunosuppressant medications.

All radiographs, computed tomography (CT), magnetic resonance imaging (MRI), bone scans, and positron emission tomography (PET) scans during the duration of treatment were reviewed by two radiology trainees (WA, SS), and imaging features of VIP were confirmed by a board-certified musculoskeletal radiologist (TL). VIP lesions were initially diagnosed and best characterized on radiographs and CT studies which were used for morphological analysis of the lesions. Standard radiographs were obtained of various bones and joints in at least two planes. CT studies of the chest were obtained on three different scanners—General Electric (GE) LightSpeed VCT, GE Discovery CT750 HD, and Phillips Brilliance 64—using the acquisition parameters listed in Supplementary Table 1.

The length, location in the body, location in the bone, type, and morphology of VIP lesions were recorded. The lesion type was classified as either continuous or interrupted. Lesions were further subclassified based on morphology as eggshell, lobulated, soap bubbles, solid, single layer, multilayered, spiculated or sunburst for continuous-type lesions and wedge-shaped, Codman triangle, interrupted onion skins, or interrupted spiculae for interrupted-type lesions [14]. Representative lesions are included in Figs. 1 and 2.

Incident VIP was defined as new periostitis on imaging after 28 days or more of voriconazole treatment in the absence of an alternative diagnosis such as bone infection, metabolic disorders, and paraneoplastic disorders. Three patients were excluded from the analysis given that voriconazole dose, frequency, and treatment duration could not be confirmed due to incomplete or missing medical records. An additional patient was excluded from the analysis because although they developed new periostitis and enthesopathy after chronic voriconazole use, it could not be confirmed that their lesions were solely secondary to voriconazole use. Cumulative voriconazole dose was calculated as the summation of the prescribed daily dose from the date of initiation until the date of incident VIP. Voriconazole treatment duration was defined as the difference between the date of prescribed voriconazole initiation and the date of incident VIP. Average ALP was calculated as the mean of ALP measurements for each patient during the duration of treatment.

Differences in the group characteristics between patients with VIP and patients without VIP were assessed using the Mann–Whitney U test for continuous variables and the chi-square test or Fisher’s exact test for categorical variables as appropriate. To reduce small-sample bias in maximum likelihood estimation, univariate Firth’s logistic regression models were performed. These models used incident VIP as the outcome and cumulative voriconazole dose, treatment duration, and average ALP as predictors, with incremental changes set at 31.5 g, 63 days, and 50 IU/L, respectively. Incremental changes for cumulative voriconazole dose and treatment duration were selected based on treatment guidelines for invasive aspergillosis [15], a common indication for voriconazole use. Incremental change in ALP was selected as 50 IU/L as this may increase the ALP above the upper limit of normal and signal an alteration in bone metabolism warranting clinical attention. Predictors were selected a priori based on the literature review of VIP [7, 8, 16, 17]. Furthermore, we performed a sensitivity analysis to assess whether the results differed after adjusting for age in multivariate Firth’s logistic regression models (two predictors, one outcome). We conducted additional sensitivity analyses to assess for potential selection bias arising from the exclusion of the previously noted four patients. In these analyses, the patient presenting with lesions potentially attributable to voriconazole use was categorized alternately as both affected and unaffected by VIP in distinct sensitivity analyses. All predictors and covariates were analyzed as linear variables. Statistical analysis was performed using Stata software (Version 15, College Station, TX, USA; StataCorp LP). P-values < 0.05 were considered as statistically significant.

Patient characteristics are summarized in Table 1. There were nine patients with VIP and 122 patients without VIP. Compared to patients without VIP, patients with VIP had higher average ALP (276.0 vs 129.5 IU/L, P = 0.002). There was insufficient evidence to conclude that the distributions of age, sex, immunocompromised status, voriconazole level, and treatment duration were significantly different between groups defined by VIP status. Additionally, there was a trend that patients with VIP had a higher cumulative voriconazole dose than those without VIP.

The imaging characteristics of VIP lesions are described in Table 2. All lesions were diagnosed on standard radiographs and/or CT studies. The total number of lesions was 116. The median number of lesions per patient was 10 (interquartile range of 6–16) ranging from 3 to 36 lesions per patient. The mean length was 26.5 (11.5) mm ranging from 2.8 to 112.0 mm. The most common lesion locations in the body were the ribs (37%), hands (23%), and legs (16%). The most common lesion locations in the bone were the diaphysis (42%) and body of the ribs (37%). Most of the lesions demonstrated continuous-type periosteal reactions (87%). The most common lesion morphology was solid (44%) followed by lobulated (29%) and wedge-shaped (13%). When comparing imaging modalities, there were 59 lesions diagnosed by radiographs and 57 lesions diagnosed by CT studies. The distributions for lesion type and morphology by modality are described in Supplementary Table 2. Imaging characteristics of VIP lesions by patient are included in Supplementary Table 3.

As shown in Table 3, a 31.5-g increase in cumulative voriconazole dose was associated with higher odds of incident VIP (odds ratio 1.08 [95% CI 1.02, 1.14], P = 0.009). Increased treatment duration (63 days) was associated with higher odds of incident VIP (odds ratio 1.07 [95% CI 1.02, 1.12], P = 0.005). Increased average ALP (50 IU/L) was associated with higher odds of incident VIP (odds ratio 1.34 [95% CI 1.11, 1.61], P = 0.002).

In a sensitivity analysis adjusting for age, the associations of cumulative voriconazole dose, treatment duration, and average ALP with incident VIP remained statistically significant with estimates that were consistent in direction and similar in magnitude (Table 3). These findings remained consistent in additional sensitivity analyses evaluating for potential selection bias arising from the excluded four patients (data not shown).