Determination of FGF23 Levels for the Diagnosis of FGF23-Mediated Hypophosphatemia

Abstract

Fibroblast growth factor-23 (FGF23) measurement is a critical tool in the evaluation of patients with disordered phosphate homeostasis. Available laboratory reference ranges for blood FGF23 were developed using samples from normophosphatemic individuals. Reliance on such values can lead to misdiagnosis in patients with FGF23-mediated hypophosphatemia, such as X-linked hypophosphatemia (XLH) and tumor-induced osteomalacia (TIO), in whom pathology-driving FGF23 levels can be in the “normal range.” To determine FGF23 levels that are diagnostic for the identification of patients with FGF23-mediated hypophosphatemic disorders, we studied 149 patients with various disorders of FGF23-mediated and FGF23-independent hypophosphatemia and defined cut-off levels for both intact FGF23 (iFGF23) and C-terminal FGF23 (cFGF23) that can accurately distinguish between FGF23-mediated and FGF23-independent hypophosphatemia. In addition, to demonstrate the relationship between FGF23 and phosphate across the spectrum of human physiology, we assessed blood levels of FGF23 and phosphate in 434 patients with various forms of hypophosphatemia, hyperphosphatemia, and normophosphatemia. An intact FGF23 cut point of 27 pg/mL was 100% sensitive and specific in distinguishing FGF23-mediated from FGF23-independent hypophosphatemia, and a cFGF23 cut point of 90 RU/mL was 100% sensitive and specific in distinguishing specifically TIO from FGF23-independent hypophosphatemia. There was overlap in the cFGF23 range of 45–90 RU/mL between genetic forms of FGF23 excess and FGF23-independent hypophosphatemia, substantiating the superiority of iFGF23 over cFGF23 in making the diagnosis of FGF23-mediated hypophosphatemia. In this cohort, using the laboratory upper limit of normal for cFGF23 (180 RU/mL) would result in a misdiagnosis in more than half of patients with FGF23-mediated hypophosphatemia. In this, the largest study of FGF23 in chronic hypophosphatemia to date, we established iFGF23 and cFGF23 cut-off values to assist in the evaluation and diagnosis of hypophosphatemic conditions. © 2022 American Society for Bone and Mineral Research (ASBMR). This article has been contributed to by US Government employees and their work is in the public domain in the USA.

Graphical Abstract

Introduction

Chronic hypophosphatemia is a disorder of mineral metabolism that results in significant long-term musculoskeletal complications and impaired quality of life.^(1–3) In muscle, hypophosphatemia-associated tissue hypoxia manifests with proximal myopathy;⁽⁴⁾ in bone, chronic hypophosphatemia leads to impaired mineralization that results in osteomalacia in adults and rickets in children. Clinically, adults with chronic hypophosphatemia present with muscle weakness, bone pain, fractures, and height loss.⁽⁵⁾ Onset of hypophosphatemia in childhood affects skeletal development as well, leading to impaired growth, rickets, and deformity.^(1,2,6)

Phosphate homeostasis is a result of a finely tuned balance between urinary phosphate losses and net phosphate absorption from the gastrointestinal tract. It is coordinated by parathyroid hormone (PTH), 1,25-dihydroxyvitamin D (1,25D), and fibroblast growth factor 23 (FGF23).^(1,7) FGF23 is produced by bone cells and is the primary regulator of phosphate homeostasis.⁽⁸⁾ The intact hormone (iFGF23) acts via the fibroblast growth factor receptor 1 (FGFR1) and its coreceptor Klotho to inhibit both the sodium phosphate cotransporters (NaPi-2a and NaPi-2c) in the renal proximal tubule and 25-hydroxyvitamin D 1-α-hydroxylase expression.⁽⁹⁾ These actions lead to increased renal phosphate excretion and decreased serum 1,25D, which is important for intestinal phosphate absorption, together resulting in lower blood phosphate.⁽¹⁰⁾ When this equilibrium fails, hyperphosphatemia or hypophosphatemia occurs. For example, in chronic kidney disease (CKD), the most common cause of elevated FGF23, progressive reduction in phosphate filtration results in a compensatory increase of FGF23 that maintains normophosphatemia up to significantly decreased glomerular filtration rates, when this compensation is surpassed and hyperphosphatemia develops.⁽¹¹⁾

Chronic hypophosphatemia is mediated either by inappropriately high circulating levels of FGF23 (FGF23-mediated hypophosphatemia) or by causes independent of FGF23 action, such as decreased intestinal absorption, hyperparathyroidism, transcellular shifts, or proximal renal tubule dysfunction (FGF23-independent hypophosphatemia).^(1,12) Both FGF23-mediated and FGF23-independent hypophosphatemia may be acquired or inherited. The most common, non-iatrogenic, cause of acquired FGF23-mediated hypophosphatemia is tumor-induced osteomalacia (TIO), which is caused by ectopic FGF23 production by usually benign or rarely malignant tumors.⁽⁵⁾ Important genetic conditions of FGF23 excess result from variants in the genes PHEX (X-linked hypophosphatemia [XLH]), FGF23 (autosomal dominant hypophosphatemic rickets [ADHR]), and DMP1 and ENPP1 (autosomal recessive hypophosphatemic rickets [ARHR]).⁽¹⁾ FGF23-mediated hypophosphatemia is also associated with fibrous dysplasia (FD) of bone,⁽⁸⁾ cutaneous skeletal hypophosphatemia syndrome (CSHS),⁽¹³⁾ neurofibromatosis type 1 (NF1), and some formulations of infused iron supplementation.^(14,15)

FGF23-independent hypophosphatemia can be due to hyperparathyroidism,⁽¹⁶⁾ vitamin D deficiency, malabsorption, transcellular shifts (usually acute and transient), proximal renal tubulopathies usually associated with generalized renal tubulopathies (i.e., Fanconi syndrome), or, rarely, renal phosphate transport defects, such as in hypophosphatemic rickets with hypercalciuria (HHRH).⁽¹⁷⁾ Fanconi syndrome can be due to inherited causes (e.g., nephropathic cystinosis⁽¹⁸⁾ or oculocerebrorenal syndrome of Lowe) or acquired causes (e.g., exposure to heavy metals or drugs such as tenofovir).^(1,19)

Appropriate treatment of the many hypophosphatemic disorders is determined by the underlying etiology.^(20,21) Therefore, distinguishing FGF23-mediated from FGF23-independent chronic hypophophatemia is essential to determine the most appropriate management.

Circulating FGF23 measurement is a valuable tool in diagnosing chronic hypophosphatemia.⁽²²⁾ Determining the FGF23 level that distinguishes between FGF23-dependent and independent causes is crucial for accurate diagnosis. Laboratory reference ranges for blood FGF23 are established from normophosphatemic patients. In the context of hypophosphatemia, in patients with normal intact feedback, FGF23 is expected to be suppressed.⁽¹⁰⁾ Patients with FGF23-mediated diseases may have FGF23 levels that are either overtly high or inappropriately within the normal range for the level of hypophosphatemia.⁽²³⁾ This is analogous to primary hyperparathyroidism in which elevated calcium with “inappropriately normal” or nonsuppressed intact parathyroid hormone (iPTH) can still be consistent with the disease. Several FGF23 immunoassays have been developed, with variable availability around the globe.^(24,25) C-terminal FGF23 assays (cFGF23) measure both active intact hormone and inactive carboxy-terminal FGF23 fragments. Intact FGF23 assays (iFGF23) measure only the full-length, active molecule. Determining whether a FGF23 level is “inappropriately normal” or actually normal is subjective and can result in clinical confusion. In this study, the largest of its kind to date, we define optimal cFGF23 and iFGF23 cut points for the diagnosis of FGF23-mediated hypophosphatemia. In addition, to deepen our understanding of phosphate homeostasis, we explore the relationship between FGF23 and phosphate across a wide spectrum of phosphate-related disorders.

Methods

Subjects

Data from 434 individuals, comprising three cohorts, are included in this study: (i) cut-point determination (n = 142); (ii) iFGF23 assay comparison (n = 34); and (iii) FGF23/phosphate by disease (n = 434). National Institutes of Health (NIH) subjects were enrolled in Institutional Review Board (IRB)-approved protocols from the National Human Genome Research Institute (NHGRI), National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), National Institutes of Child Health and Human Development (NICHD), National Institute of Dental and Craniofacial Research (NIDCR), or the NIH Clinical Center Department of Transfusion Medicine. Pontificia Universidad Católica de Chile subjects were evaluated under a protocol approved by the Ethics Committee of the Faculty of Medicine. Subjects with CKD were evaluated at the Division of Pediatric Nephrology, University of California, Los Angeles (UCLA), in a study approved by the UCLA IRB. All adult patients or parents/guardians of minors provided written informed consent. Normophosphatemic controls also included purchased samples from Lee Biosolutions, Inc. (Maryland Heights, MO) and Innovative Research, Inc. (Novi, MI).

Cut-point determination cohort

One-hundred fourteen patients evaluated at the NIH Clinical Center between January 1999 and July 2020 were identified retrospectively and included if they carried an established diagnosis of hypophosphatemia, determined using published age and gender-specific normal ranges, and had previously obtained concomitant blood phosphate and FGF23 levels or had stored blood samples available for FGF23 measurement. Twenty-eight additional subjects with XLH who were evaluated at the Center for Translational research in Endocrinology of Pontificia Universidad Católica de Chile were also included using recruitment methods as previously published.⁽³⁾ Samples were not included if patients were normophosphatemic at the time of the blood draw or were receiving either infigratinib, a FGFR1 blocker, or burosumab, a FGF23 monoclonal antibody that interferes directly with FGF23 assays.

Intact FGF23 assay comparison cohort

To make a direct comparison of two intact FGF23 assays, we assessed 34 NIH subjects (11 with TIO, 2 with XLH, 5 with FD, 10 with HFTC, and 7 normophosphatemic controls) with previously determined FGF23 levels or stored blood samples.

FGF23/phosphate by disease cohort

A total of 434 patients, including patients in the cut-point and assay comparison analyses, with hypophosphatemic disorders (n = 149), hyperphosphatemic disorders (n = 218), or normophosphatemic controls (n = 59) were included to create diagrams demonstrating the relationship between cFGF23, iFGF23, and phosphate across various disease states (Fig. 1). The hypophosphatemic cohort included an additional five patients with TIO and two with XLH who had FGF23-mediated physiology but had been excluded from cut-point analysis because their phosphate levels were in the low-normal range. Eight previously published nomophosphatemic patients with paraganglioma/pheochromocytoma and somatostatinoma associated with polycythemia syndrome due to gain-of-function mutation of HIF2A were included. This cohort has normal phosphate, normal iFGF23, high cFGF23, and high erythropoietin levels, highlighting a recently established, but still ill-defined, relationship between the hypoxia-inducible factor (HIF)/erythropoietin/iron pathway and FGF23.⁽²⁶⁾

Fig. 1.

Breakdown of patients by diagnosis.

Patients with hyperphosphatemic disorders included 169 patients with CKD with a mean age of 17.6 years, 44 of whom had CKD Stage 4 or 5, and 125 of whom were on hemodialysis. Also included were 17 patients with hyperphosphatemic familial tumoral calcinosis (HFTC) and 32 patients with hypoparathyroidism.

In CKD, the most common cause of elevated FGF23, progressive reduction in renal phosphate filtration, results in compensatory FGF23 elevation that precedes the development of hyperphosphatemia.⁽¹¹⁾ Late-stage CKD patients may have high phosphate and extremely high iFGF23 and cFGF23 levels. HFTC is a rare genetic disease of iFGF23 deficiency that is characterized by low iFGF23, high cFGF23, and resulting hyperphosphatemia.⁽²⁷⁾ Hypoparathyroidism is a disease of low iPTH resulting in low calcium and high phosphate levels.

Biochemical data from noncystinosis moderate to severe CKD patients were obtained from the Division of Pediatric Nephrology, University of California, Los Angeles, as previously published.⁽¹⁸⁾ Patients with HFTC and hypoparathyroidism were enrolled in IRB-approved studies at the NIH. The normophosphatemic controls were either purchased or obtained as part of other IRB-approved protocols at the NIH. There was limited ability to screen for comorbidities in the purchased samples (n = 35), but when history was available, patients with abnormal phosphate, calcium, vitamin D, iPTH, or an estimated glomerular filtration rate of <90 mL/min/1.73 m² were excluded.

FGF23 measurements

Intact FGF23 (iFGF23) concentrations were measured by either an enzyme-linked immunosorbent assay (ELISA) on plasma (iFGF23, Immutopics International, Quidel Corporation, Athens, OH, USA) or a chemiluminescence enzyme immunoassay (CLEIA) on serum (iFGF23-CLEIA, Eagle Biosciences, Amherst, NH, and Mayo Clinic Laboratories, Rochester, MN, USA). The carboxy-terminal FGF23 ELISA, which measures both biologically active intact FGF23 and inactive carboxy-terminal fragments, was performed on plasma (cFGF23, Immutopics International, Quidel Corporation, Athens, OH, USA).

iFGF23, iFGF23-CLEIA, and cFGF23 were determined using test kits according to the manufacturer’s protocol. Alternatively, some iFGF23-CLEIA and cFGF23 results were from the Mayo Clinic Laboratories (Rochester, MN, USA). iFGF23 and iFGF23-CLEIA results that were reportedly below the manufacturer-reported limit of detection of 1.5 and 5 pg/mL, respectively, were replaced with the limit of detection value. If the reported result exceeded the upper limit of detection, the samples were diluted and reanalyzed. The cFGF23/iFGF23 ratio was calculated by dividing the cFGF23 (RU/mL) value by the iFGF23 (mg/dL) value without any unit conversion.

Phosphate assessment

Blood and urine phosphate and creatinine were measured at the NIH Clinical Center (Bethesda, MD, USA), Universidad Católica (UC/Christus), or UCLA laboratory using standard methodology. Any result that was reported as below the lower limit of detection was replaced with the limit of detection value.

Phosphate levels are expressed both as blood concentration and Z-score. Z-scores were calculated assuming a normal distribution of phosphate and published age- and gender-specific reference ranges for a Roche Cobas machine, used because the majority of phosphate levels were obtained using this platform. Phosphate normal ranges were as follows: age 1–4 years: 4.1–6.4 mg/dL; 5–12 years: 4–5.6 mg/dL; 13–15 years: male 3.5–5.8 mg/dL, female 3.1–5.3 mg/dL; 16–18 years 2.9–4.8; >19 2.5–4.5 mg/dL.⁽²⁸⁾ When gender was not known for a sample of a patient between 13 and 15 years, the male normal range was used.

Phosphate Z-scores were determined using the following equation, where ULN and LLN represent the upper limit and lower limits of the normal range, respectively:

$$P_{\text{SD}}=\frac{\text{Blood}\text{phosphate}-\overline{x}}{\left(\frac{\text{ULN}-\overline{x}}{2}\right)},$$

where

$$\overline{x}=\frac{(\text{ULN}+\text{LLN})}{2}.$$

Tubular reabsorption of phosphate (TRP) was calculated using the following formula:

$$\begin{array}{l}\text{TRP}\%=100\\ \times \left(1-\left(\frac{\text{Urine}\text{phosphate}\times \text{Blood}\text{creatinine}}{\text{Blood}\text{phosphate}\times \text{Urine}\text{creatinine}}\right)\right).\end{array}$$

Negative TRP results were recorded as 0.

Additional biochemical assessment

Blood iPTH, alkaline phosphatase (AP), and 1,25D were measured at the NIH Clinical Center (Bethesda, MD, USA) and Universidad Católica (UC/Christus) using standard methodology. Any result that was reported to be below the lower limit of detection was replaced with the limit of detection value.

Statistical analyses

Continuous variables were presented either as mean ± SD or median (interquartile range [IQR]). Categorial variables were presented as n (%). Continuous variables were compared between two groups using a Mann–Whitney U test, with a p value of less than 0.05 considered significant. Continuous variables were compared among more than two groups using a Kruskal–Wallis test and Dunn’s multiple comparisons test. Correlation analysis was performed using two-tailed Spearman correlation analysis. Receiver-operating characteristic (ROC) curves were used to determine cut points. Bland–Altman plot, linear regression analysis, and two-tailed Spearman correlation analysis were used to compare the two iFGF23 assays. Linear regression analysis and two-tailed Spearman correlation analysis was performed on log-transformed data to determine log(FGF23) as a function of phosphate for FGF23-mediated (HFTC, TIO, genetic causes of FGF23-excess) and FGF23-independent diseases (CKD, hypoparathyroidism, and FGF23-independent hypophosphatemia).

Analyses and graphs were prepared using GraphPad Prism 8.

Results

Cut-point determination

Characteristics and biochemical parameters of 142 hypophosphatemic patients are summarized in Table 1. One hundred twenty-three patients had FGF23-mediated hypophosphatemia, including TIO (n = 35), XLH (n = 34), CSHS (n = 5), ENPP1 deficiency (n = 12), FD (n = 36), and NF1 (n = 1). Nineteen patients were diagnosed with FGF23-independent hypophosphatemia, including cystinosis (n = 16), oculocerebrorenal syndrome of Lowe (n = 1), familial Fanconi syndrome (n = 1), and HHRH (n = 1).

Table 1.

Characteristics of Patients with FGF23-mediated and FGF23-lndependent Flypophosphatemia

Parametera		n	Sex	Age	Phosphate		iFGF23	CFGF23	c/iFGF23	1,25D	TRP	25D	Creatinine	Hct
Units/Reference interval			n (%)
			Male	year	mg/dL	Z-score	<54 pg/mL	<180 RU/mLb	RU/pg	19–79 pg/mL	85–95%	>32 ng/mL	mg/dL	%
All patients		142	55 (39)	25 (8–40)	2.1 (1.6–2.6)	3.8 ( 4.8 ( 2.8))	105 (70–183)	163 (87–330)	1.4 (1.1–3.0)	41 (24–59)	79 (57–88)	30 (21–38)	0.58 (0.4–0.8)	39 (37–42)
All FGF23-mediated hypophosphatemia		123	47 (38)	27 (9–47)	2.1 (1.6–2.6)	3.6 ( 4.6 ( 2.8))	123 (84–232)	195 (126–350)	1.9 (1.3–5.9)	38 (20–56)	83 (67–88)	29 (21–38)	0.53 (0.35–0.74)	40 (37–42)
TIO		35	20 (57)	48 (39–59)	1.6 (1.4–1.9)	−3.8 (−4.3–(−3.3))	314 (125–828)	458 (220–792)	1.2 (0.7–3)	19 (14–28)	67 (54–75)	33 (23–41)	0.8 (0.68–0.95)	40 (38–44)
Genetic	XLH	34	11 (32)	27 (9–37)	2 (1.6–2.4)	−4 (−5.3–(−3.1))	103 (73–166)	110 (85–157)	1.1 (0.9–1.3)	28 (27–43)	85 (77–89)	22 (15–29)	0.49 (0.42–0.6)	39 (37–40)
	CSHS	5	1 (20)	16 (13–17)	2.2 (1.2–2.6)	−5.5 (−5.5–(−2.6))	162 (93–259)	124 (100–230)	1.5 (1.5–1.6)	37 (14–54)	90 (90–92)	36 (29–39)	0.47 (0.43–0.74)	40 (39–40.2)
	ENPP1	12	3(25)	10.5 (7–29)	2.3 (2.1–2.5)	−4.5 (−5.1–(−3.7))	92 (77–130)	180 (140–235)	1.9 (1.4–2.6)	43 (37–62)	84 (75–88)	32 (24–44)	0.5 (0.35–0.69)	38 (37–42)
	Deficiency
	FD	36	12(33)	11.5 (7–28)	3 (2.3–3.7)	−2.8 (−3.2–(−2.3))	109 (102–124)	208 (154–315)	2 (1.8–4)	56 (50–63)	88 (85–92)	30 (24–44)	0.34 (0.25–0.52)	40 (37–42)
	NF1	1	0(0)	60	1.3	−4	114	342	3	40	66	45	0.61	−
p-valuec			0.008	<0.0001	<0.0001	0.5	<0.0001	<0.0001	0.5	<0.0001	<0.0001	0.2	<0.0001	0.2
All FGF23-lndependent Hypophosphatemia		19	8(42)	12(4–21)	2.1 (1.6–2.4)	−4.7 (−6– (−2.8))	3 (2–9)	39 (28–61)	9 (3.4–14)	62 (51–74)	20 (0–31)	34 (32–43)	0.95 (0.62–1.57)	38 (35–40)
Cystinosis		16	7(44)	11.5 (4–21)	2.1 (1.6–2.4)	−4.9 (−6.4– (−3.6))	2 (2–9)	38 (29–59)	9 (4.2–14.7)	62 (54–73)	20 (0–29)	34 (32–47)	0.95 (0.55–1.52)	38 (36–40)
Other		3	1 (33)	23 (22–25)	2.1 (1.5–2.1)	−2.8 (−4– (−2.8))	8 (5–16)	61 (40–65)	8 (5.8–10.9)	80 (57–102)	54 (33–75)	29 (25–34)	1.27 (0.98–1.55)	35 (35–35)
p valued			0.8	0.003	0.6	0.03	<0.0001	<0.0001	<0.0001	0.0005	<0.0001	0.2	<0.0001	0.05

Values presented as n (%) or median (IQR).

^aWhen lab results came in below the lower limit of detection, it was denoted as the lower limit of detection for that assay.

^b<230 RU/mL in pediatric patients.

^cTIO and genetic causes of FGF23-mediated hypophosphatemia were compared using a chi Fisher’s test or Mann-Whitney U test.

^dFGF23-independent and FGF23-mediated groups were compared using a chi Fisher's test or Mann-Whitney U test.

Abbreviations: 1,25D = 1,25-dihydroxyvitamin D; 25D = 25-hydroxyvitamin D; c/iFGF23 = cFGF23/iFGF23 ratio; cFGF23 = FGF23 measured using a C-terminal assay that measures both intact FGF23 and inactive C-terminal fragments; CSHS = cutaneous skeletal hypophosphatemia syndrome; FD = fibrous dysplasia; Hct = hematocrit; iFGF23 = intact FGF23; NF1 = neurofibromatosis type 1; TIO = tumor-induced osteomalacia; TRP = tubular reabsorption of phosphate; XLH = x-linked hypophosphatemic rickets.

Overall, 39% of patients were male, with a median age of 25, ranging from 1 to 75 years old. Patients with FGF23-mediated hypophosphatemia were significantly older compared to those with FGF23-independent hypophosphatemia (p = 0.003). Patients with TIO were significantly older than those with genetic causes of FGF23-excess (p < 0.0001). There was no significant correlation between FGF23 level and age in both the FGF23-mediated group (cFGF23: r = 0.14, p = 0.2; iFGF23: r = 0.11, p = 0.3) and the FGF23-independent group (cFGF23: r = 0.32, p = 0.19; iFGF23: r = 0.44, p = 0.13). There were also no significant between-sex differences in FGF23 levels within the FGF23-mediated and FGF23-independent cohorts.

Per the inclusion criteria, all patients had phosphate Z-scores less than −2. Phosphate levels (median [IQR]) were 2.1 (1.6–2.6) mg/dL; Z-scores ranged from −9.4 to −2. Phosphate Z-scores were lower in the FGF23-independent group compared to the FGF23-mediated group (p = 0.03), although the raw phosphate levels were not significantly different, which may be a reflection of the age difference between the two groups. 1,25D levels were significantly lower in FGF23-mediated hypophosphatemia compared to FGF23-independent hypophosphatemia (p = 0.0005), consistent with suppression of 1-alpha hydroxylase by FGF23. TRP was low in both groups but significantly lower in the FGF23-independent group, many of whom had concomitant kidney failure.

Seventy-four percent of patients with FGF23-mediated hypophosphatemia, but only 21% of patients with FGF23-independent hypophosphatemia, were not taking any phosphate or activated vitamin D supplementation at the time of their laboratory evaluation. Other patients were either taking supplemental phosphate or activated vitamin D or we were unable to determine whether they were taking medications (n = 5). No significant differences in phosphate or 1,25D were found on or off therapy within each disease group. iFGF23 and cFGF23 were not significantly different when on or off therapy in the FGF23-independent group. However, both FGF23 levels were higher on therapy compared to off therapy in the FGF23-mediated group (iFGF23 p < 0.0017, cFGF23 p < 0.0001, respectively).

Hematocrit levels were available in 114 of 141 patients with hypophosphatemia. Within the FGF23-independent and FGF23-mediated hypophosphatemia subgroups, there was no correlation between hematocrit level and the iFGF23/cFGF23 ratio (FGF23-independent: r = −0.04, p 0.6; FGF23-mediated: r = 0.003, p = 0.98) or between hematocrit level and delta FGF23 (FGF23-independent: r = 0.1, p = 0.8; FGF23-mediated: r = −0.1, p = 0.3). There was also no correlation between hematocrit level and absolute cFGF23 or iFGF23 level in both cohorts (FGF23-independent: cFGF23 r = 0.2, p = 0.4; iFGF23 r = 0.2, p = 0.6; FGF23-mediated: cFGF23 r = 0.1, p = 0.6, iFGF23 r = 0.1, p = 0.4).

Patients with FGF23-mediated hypophosphatemia had significantly higher FGF23 levels (iFGF23 123 [84–232] pg/mL, cFGF23 195 [126–350] RU/mL) compared to patients with FGF23-independent disease (iFGF23 3 [2–9] pg/mL, cFGF23 39 [28–61]; iFGF23 p < 0.0001; cFGF23 p < 0.0001) (Fig 2A,B). There was wide variation in FGF23 levels in FGF23-mediated disease that was primarily due to patients with TIO who had iFGF23 levels between 70 and 20,464 pg/mL and cFGF23 levels between 93 and 8850 RU/mL. Median iFGF23 and cFGF23 levels were significantly higher in patients with TIO compared to other causes of FGF23-mediated hypophosphatemia (p < 0.0001) (Fig. 2C,D). Only patients with TIO had FGF23 values higher than 800 (RU/mL or pg/mL) using either assay.

Fig. 2.

Intact and C-terminal FGF23 in various FGF23-independent and FGF23-mediated disorders of hypophosphatemia. (A, C) Intact FGF23 (iFGF23) and (B, D) C-terminal FGF23 (cFGF23) levels in FGF23-independent and FGF23-mediated hypophosphatemic diseases. Box-and-whisker plots represent the 25th, 50th, and 75th percentiles and the minimum and maximum values. Colored circles represent individual values. The iFGF23 cut point of 27 pg/mL and the cFGF23 cut point of 90 pg/mL are represented by black dashed lines. The gray horizontal dashed line (B, C) represents the upper limit of normal for cFGF23 in normophosphatemic patients (180 RU/mL). The gray boxes in (B) and (D) show the overlap between FGF23-independent disease and genetic causes of FGF23 excess. Receiver-operating characteristic (ROC) curves for (E) iFGF23 and (F) cFGF23. Blue circles show relevant cut points. The iFGF23 27 pg/mL cut point of 27 had 100% sensitivity and specificity. cFGF23 upper limit of normal for normophosphatemic patients of 180 RU/mL had a sensitivity of only 54% and specificity of 100%. A cut point of 90 RU/mL resulted in a sensitivity of 85% and specificity of 100%. A cut point of 45 RU/mL had a sensitivity of 100% but specificity of only 58%. XLH, X-linked hypophosphatemia; FD, fibrous dysplasia of bone; CSHS, cutaneous skeletal hypophosphatemia syndrome; NF1, neurofibromatosis type 1; TIO, tumor-induced osteomalacia; ENNPP1D, autosomal recessive hypophosphatemic rickets due to ENPP1 deficiency.

The previously published iFGF23 cut point of 30 pg/mL^(22,29) had a 99% sensitivity and 100% specificity for identifying FGF23-mediated disease. Only one patient with FGF23-mediated disease had an iFGF23 just under at 29 pg/mL. This was a 7-year-old patient with ENPP1 deficiency and phosphate of 2.8 mg/dL who was not on phosphate or calcitriol supplementation at the time of evaluation. The highest iFGF23 in the FGF23-independent group was 24 pg/mL. ROC curves show that a cut point of 27 pg/mL would be 100% sensitive and specific for FGF23-mediated disease (Fig. 2E).

The cFGF23 ULN for normophosphatemic adults and children is reported in our assay to be 180 RU/mL and 230 RU/mL, respectively. Using these as cut points, the sensitivity for FGF23-mediated disease was 58% in the adults and 32% in the children in our cohort, although they are 100% specific. A cut point of 90 RU/mL improved the sensitivity to 85% (Fig. 2F) and distinguished TIO from FGF23-independent disease with 100% sensitivity and specificity. There is an overlap between genetic causes of FGF23-excess and FGF23-independent disease between 45 RU/mL and 90 RU/mL. A practical schematic of FGF23 cut points established by this study is provided in Fig. 3.

Fig. 3.

Schematic summarizing cut points for intact FGF23 (iFGF23) and C-terminal FGF23 (cFGF23) in hypophosphatemic patients. The cut point of 27 pg/mL can distinguish FGF23-independent and FGF23-mediated hypophosphatemia with 100% sensitivity and 100% specificity. For cFGF23, the cut point of 90 RU/mL distinguishes between TIO and FGF23-independent hypophosphatemia. Values less than 45 RU/mL are 100% sensitive for FGF23-independent disease. However, a “gray zone” exists between 45 and 90 RU/mL in which genetic FGF23-mediated diseases overlap with FGF23-independent diseases.

Intact FGF23 assay comparison

The mean intact FGF23 level for the 34 studied samples was 445 ± 1150 pg/mL (range 7.5 to 5130 pg/mL) by Quidel ELISA and 320 ± 888 pg/mL (range 5 to 4175 pg/mL) by CLEIA (Fig. 4). One patient was removed prior to analysis because the iFGF23 measured by ELISA was a statistical outlier inconsistent with the diagnosis. In this patient with HFTC, the plasma iFGF23 level by ELISA was very elevated, though it is typically low in this disease and had been lower in this patient at other times. The serum iFGF23-CLEIA run on the same day was low, as expected.

Fig. 4.

Linear regression between intact FGF23 (iFGF23) by ELISA method (Quidel Corp., San Diego, CA) and iFGF23-CLIEA by chemiluminescence enzyme immunoassay (CLEIA) method (EagleBio, Amherst, NH) using (A) all values and (C) values restricted to <100 pg/mL. Bland–Altman analysis using (B) all values showed a bias of −37 pg/mL with 95% limits of agreement of −447 pg/mL to 373 pg/mL. In the restricted analysis (D), bland–Altman analysis found a bias of −3.9 and 95% limits of agreement of −26 pg/mL to 18 pg/mL.

There was a significant positive correlation between the two FGF23 assays (r = 0.9, p < 0.0001). Linear regression of iFGF23 as a function of iFGF23-CLEIA resulted in a slope of 1.1 and a y-intercept of −4.6 (r² = 0.97) (Fig. 4A). Bland–Altman analysis found a bias of −37 pg/mL with 95% limits of agreement of −447 pg/mL to 373 pg/mL (Fig. 4B). Although this suggests large differences between the two assays, the differences were reduced when analyses were restricted to only values less than 100 pg/mL (Fig. 4C,D). The linear regression of iFGF23 as a function of iFGF23-CLEIA resulted in a slope of 1.5 and y-intercept of −1. Bland–Altman analysis found a bias of −3.9 and 95% limits of agreement of −26 to 18 pg/mL. Of note, all 18 patients with hypophosphatemic disorders (11 TIO, two XLH, five FD) had iFGF23 levels above the cut point of 27 pg/mL on both assays.

FGF23 and phosphate across disease states

The relationships between FGF23 and phosphate across a wide spectrum of diseases, representing diverse pathophysiological abnormalities in phosphate regulation, are depicted in Fig. 5. Hypophosphatemic disorders on the left of each graph include FGF23-mediated disorders with high FGF23 and FGF23-independent disorders with suppressed FGF23. On the right, hyperphosphatemic disorders include CKD and hypoparathyroidism with high FGF23 and HFTC, which has a low iFGF23 but high cFGF23. Notably, the group of normophosphatemic controls used in this study had iFGF23 of 57 ± 33 pg/mL and cFGF23 138 ± 180 RU/mL, which are slightly higher than we might expect in healthy individuals. However, controls were not universally screened for comorbidities besides phosphate levels, so other causes of increased FGF23, such as decreased kidney function, may have been present in these patients.

Fig. 5.

Overview of relationship between FGF23 and phosphate across disease states. (A) Intact FGF23 (iFGF23) and (B) C-terminal FGF23 (cFGF23) levels versus phosphate Z-score in various hyperphosphatemic and hypophosphatemic diseases, as well as in normophosphatemic controls. In diseases in which the pathologic state is driven by abnormal renal phosphate handling independent of FGF23, including CKD, hypoparathyroidism, and FGF23-independent hypophosphatemia, iFGF23 and cFGF23 positively correlated with phosphate, as represented by the light gray lines graphed on a log-linear scale (log(iFGF23) = 0.13x + 2, r² = 0.33, p < .0001; log(cFGF23) = 0.12x + 2.3, r² = 0.37, p < .0001). In diseases caused by inappropriate intact FGF23 levels (black lines) including TIO, genetic FGF23 excess, and HFTC, iFGF23 had an inverse correlation with phosphate (log(iFGF23) = −0.09x + 1.8, r² 0.39, p < .0001). Excluding the HFTC cohort, cFGF23 was not significantly correlated with phosphate (log(cFGF23) = −0.03x + 2.3, r² = 0.009, p = .32). Schematic representation of the relationships between (C) iFGF23 or (D) cFGF23 and phosphate in different disease states are also shown. TIO, tumor-induced osteomalacia; Genetic, genetic disorders of FGF23-excess; CKD, chronic kidney disease; HFTC, hyperphosphatemic familial tumoral calcinosis; Hypopara, hypoparathyroidism; Normophos, normophoshatemic patients.

As expected, in diseases with altered renal phosphate handling independent of FGF23, such as CKD, hypoparathyroidism, and FGF23-independent hypophosphatemia, iFGF23 and cFGF23 had a positive correlation with phosphate (r² = 0.33 and r² = 0.37, respectively; p < 0.0001). Conversely, in diseases in which intact FGF23 is the primary pathophysiological driver, including FGF23 excess (TIO and genetic) and FGF23 deficiency (HFTC), iFGF23 had an inverse relationship with phosphate (r² = 0.39, p < 0.0001). When the same analysis was performed on cFGF23, after excluding the FTC because cFGF23 measures primarily inactive fragments in this cohort, cFGF23 was not significantly correlated with phosphate (r² = 0.01, p = 0.3).

The relationship between cFGF23 and iFGF23 varied depending on the underlying condition (Fig. 6). The cFGF23 assay detects both the active intact FGF23 molecule and inactive C-terminal fragments, whereas iFGF23 assays measure only active intact FGF23. As expected, in HFTC, which is caused by excessive posttranslational cleavage of intact FGF23 prior to secretion, the cFGF23/iFGF23 ratio was significantly elevated (56 ± 47, n = 16) compared to most other disease states (TIO 1.8 ± 1.4, n = 21; genetic FGF23-mediated hypophosphatemia 1.6 ± 1.0, n = 51; hypoparathyroidism 1.6 ± 1.1, n = 26; normophosphatemic 3.4 ± 5.3, n = 60; CKD 6.1 ± 19, n = 166). HIF2A patients tended to have higher cFGF23/iFGF23 ratios as well (16 ± 17, n = 8), as expected based on prior reports showing that this cohort has normal iFGF23 but elevated cFGF23 owing to the proposed role of the HIF/EPO pathway in the posttranslational regulation of FGF23 (Roszko, 2020). FGF23-independent hypophosphatemia also had an elevated cFGF23/iFGF23 ratio (10 ± 8, n = 14) compared to the other groups, which likely reflects the low absolute values of both cFGF23 and iFGF23 in this group. Among causes of FGF23-mediated hypophosphatemia, differences in cFGF23/iFGF23 were not significant likely due to wide variability and small sample size (TIO 1.8 ± 1.4, n = 21; XLH 1.2 ± 0.5, n = 32; ENPP1D 2.3 ± 1,2, n = 11; FD 2.7 ± 1.3, n = 5; CSHS 1.5 ± 0.09, n = 2; NF1 1.2, n = 1). However, as in a previous study, the ratio tended to be higher in FD compared to other etiologies of FGF23 excess, perhaps due to suspected greater processing of iFGF23 to cFGF23 in FD.⁽³⁰⁾ iFGF23 strongly correlated with cFGF23 after excluding HFTC and FGF23-independent disease with an r value of 0.8 (p < 0.0001). Linear regression of iFGF23 as a function of cFGF23 resulted in a slope of 2.1 and y-intercept of −294 (r² = 0.92).

Fig. 6.

Relationship between C-terminal (cFGF23) and intact FGF23 (iFGF23) across disease states. cFGF23/iFGF23 ratios (A) as a function of phosphate Z-score and(B) presented independently across various disease states. Box-and-whisker plots represent the 2.5th, 25th, 50th, 75th, 97.5th percentiles, with plotted outliers. The cFGF23/iFGF23 ratio is markedly elevated in HFTC and FGF23-independent hypophosphatemia compared to other diseases. (C) iFGF23 was strongly correlated with cFGF23 after excluding HFTC and FGF23-independent disease (r = 0.86, p < .0001). Linear regression of iFGF23 as a function of cFGF23, excluding HFTC and FGF23-independent cohorts, resulted in a slope of 2.1 and y-intercept of −294 (r² = 0.92). TIO, tumor-induced osteomalacia; Genetic, genetic disorders of FGF23-excess; CKD, chronic kidney disease; HFTC, hyperphosphatemic familial tumoral calcinosis; Hypopara, hypoparathyroidism; Normophos, normophoshatemic patients; HIF2A, paraganglioma/pheochromocytoma-somatostatinoma associated with polycythemia syndrome due to gain-of-function HIF2A mutation.

Discussion

In this relatively large study of patients with phosphate-related diseases, we established clinically useful iFGF23 (27 pg/mL) and cFGF23 (45–90 RU/mL) cut points for distinguishing FGF23-mediated from FGF23-independent hypophosphatemia (Fig. 3). We further explored the relationship between cFGF23, iFGF23, and phosphate in various disease states, demonstrating how various phosphate-related diseases can reveal the underlying physiological interplay of FGF23 and phosphate.

In hypophosphatemic disorders, determining the cause of low phosphate can be challenging. The initial evaluation of patients with chronic hypophosphatemia should include a thorough clinical history and physical exam as well as biochemistries that include both assessment for hyperparathyroidism and the presence or absence of renal phosphate wasting.⁽¹⁾ When renal phosphate excretion is elevated and is suspected to be the cause of hypophosphatemia, FGF23 should be measured to distinguish between FGF23-mediated and FGF23-independent hypophosphatemia. However, once FGF23 is obtained, interpretation of an “inappropriately normal” FGF23 can be subjective and confusing. In our cohort, using the laboratory-provided cFGF23 ULN of 180 RU/mL in adults and 220 RU/mL in children, 42% of adults and 68% of children with FGF23-mediated disease would be misdiagnosed. These findings are consistent with previous reports,⁽³¹⁾ highlighting the limited diagnostic performance of FGF23 levels for differential diagnosis of chronic hypophosphatemia, when normal ranges are used for its interpretation.

Endo et al. were the first to attempt to establish a cut point for iFGF23 in the setting of hypophosphatemia. They arrived at an iFGF23 level of 30 pg/mL as measured by the Kainos ELISA assay. This was established based on a cohort of 76 subjects with hypophosphatemia, including 32 patients with TIO, 28 with XLH, and 16 with FGF23-independent disease.^(22,32) This value was recently confirmed in three studies using the CLEIA platform, including 22 hypophosphatemic patients (18 TIO, three XLH, one vitamin D dependency Type 2), 65 hypophosphatemic patients (38 XLH, five TIO, 18 vitamin D deficiency, three denosumab-induced, and one Fanconi syndrome), and 35 hypophosphatemic patients (eight TIO, eight genetic FGF23-mediated, three drug-induced, 16 other FGF23-independent causes).^(29,33,34) In our study, the largest of its kind to date, which included 142 subjects and a wide spectrum of distinct diseases of chronic hypophosphatemia, we found that the iFGF23 cut point established by Endo et al. is valid when using the currently available Quidel ELISA assay with 95% sensitivity and 100% specificity, despite interassay variation. However, we found the cut point of 27 pg/mL was even better, with a sensitivity and specificity of 100% in this study. Prior to this study, no hypophosphatemia-specific cut point was published for cFGF23 despite this being the only commercially available assay in many countries. This study demonstrates that a cFGF23 cut point of 90 RU/mL is 100% sensitive and specific for distinguishing TIO from FGF23-independent hypophosphatemia. However, other forms of FGF23 excess had lower cFGF23 levels, often overlapping with FGF23-independent disease down to 45 RU/mL. Thus, when cFGF23 falls between 45 and 90 RU/mL, TIO is highly unlikely, but both FGF23-independent disease and other forms of FGF23-mediated hypophosphatemia should be differentiated by family history, genetic testing, dental phenotype, or other syndromic features. iFGF23 level could also be especially helpful in this subset of patients, if it is available. For both iFGF23 and cFGF23, we found that any patient with hypophosphatemia and a FGF23 level greater than 800 had a diagnosis of TIO, establishing that very high FGF23 levels are supportive, though certainly not definitive, of a diagnosis of TIO versus other causes of elevated FGF23.

Various factors, including iron deficiency, inflammation, erythropoiesis, and kidney disease, have recently been shown to influence the production or cleavage of FGF23.^(26,35) The uncoupling of FGF23 transcription from FGF23 cleavage, affected by these and other external factors, may explain the diminished specificity of cFGF23 compared to iFGF23. To investigate the impact of iron deficiency, which can cause isolated increases in cFGF23 and a decrease in the cFGF23/iFGF23 ratio, on FGF23 levels in chronic hypophosphatemic disorders, we used hematocrit as a proxy for iron deficiency. Surprisingly, we found no correlation between hematocrit level and either absolute FGF23 level, iFGF23/cFGF23 ratio, or delta FGF23. This suggests that in these cohorts, anemia did not have a large impact on FGF23 transcription or cleavage. The low phosphate levels in FGF23-independent disease and aberrant FGF23 production in FGF23-mediated disease may have overpowered any effect that anemia might have otherwise exerted on FGF23 regulation. It is important to note that the majority of patients in this cohort had normal or near-normal hemaglobin levels and that our cohort included no patients with autosomal dominant hypophosphatemic rickets (ADHR), which is known to be influenced by iron-deficiency anemia. Further investigation into the impact of other contributing factors on the uncoupling of cFGF23 and iFGF23 in hypophosphatemia is an important area for future research.

When comparing iFGF23 using Quidel ELISA to iFGF23-CLEIA, it was found that in the target range of <100 pg/mL, the assays were highly correlated, but iFGF23 was consistently slightly higher than iFGF23-CLEIA. These results agreed with prior findings observed using the Kainos ELISA assay, in which FGF23 levels were consistently higher than iFGF23-CLEIA.⁽³³⁾ Nevertheless, the cut points in this cohort appear valid across assays.⁽²⁹⁾

The examination of the relationship among cFGF23, iFGF23, and phosphate in a broad spectrum of phosphate-related disorders and normophosphatemic patients is enlightening (Fig. 6). Although there is significant overlap in FGF23 versus phosphate Z-score values for the various disease states, a clear and informative pattern emerges. For the diseases in which altered FGF23 is the primary pathophysiological driver—be it FGF23 excess (TIO and genetic) or FGF23 deficiency (HFTC)—there is an inverse relationship between iFGF23 and phosphate. Conversely, when phosphate, rather than altered FGF23, is the driver (FGF23-independent hypophosphatemia, hypoparathyroidism, and CKD), there is a positive correlation. This interesting observation supports the physiological endocrine regulation of phosphate by FGF23 with feedback regulation of FGF23 by phosphate across a broad spectrum of human pathophysiological conditions.

Examination of the ratio of cFGF23 to iFGF23 is also of interest and informative. In all cohorts, the mean and median ratio were above 1. In part, this is due to the nature of the design and the sensitivity and specificity of the assays at detecting iFGF23. The so-called cFGF23 assays, in which both the capture and detection antibodies recognize a motif that is distal to the furin cleavage site, detect not only the intact FGF23 molecule but also the C-terminal degradation product. It has recently become apparent that iFGF23 processing by furin is an important and regulated process⁽³⁶⁾ whose significance is revealed in an analysis of the cFGF23/iFGF23 ratio.⁽³⁰⁾ The reported ratios will be useful in future investigations of physiology that involves regulated FGF23 processing, as is seen in inflammation-associated elevations in FGF23.

Although the results of this study are informative, the sample size relatively large, and findings likely to stand the test of time, in absolute terms the sample size is small, particularly in the FGF23-independent group due to the rarity of the conditions. Furthermore, the FGF23-independent group consisted of only patients with primary renal phosphate wasting, so cut points may not generalize to other forms of FGF23-independent diseases, such as hyperparathyroidism or malabsorption. Because this was a retrospective study, we did not have complete data on every patient, and assays for phosphate and other biochemical measures, except for FGF23, were performed on different machines with slightly different normal ranges. The FGF23-independent group was also significantly younger than the FGF23 dependent group, which we attempted to account for by using age-specific phosphate Z-scores. Although a percentage of patients was on phosphate supplements at the time samples were obtained, there were no significant differences in phosphate or 1,25D levels between patients who were on or off conventional therapy. However, in the FGF23-mediated group, iFGF23 and cFGF23 were higher in patients who were on therapy when compared to those who were not. It is known that FGF23 levels rise with phosphate therapy in patients with XLH, likely due to an altered phosphate setpoint⁽³⁷⁾; this may be true in other forms of FGF23-mediated hypophosphatemia as well. Additionally, there may be an element of selection bias as more severely affected patients were kept on therapy out of clinical necessity and included in the analyses, for example, patients with TIO with severe disease. It is also possible that patients with milder disease who were on phosphate replacement therapy may have been excluded because phosphate levels were within the normal range. The impact of additional variables that affect FGF23 levels, such as iron, ferritin, and erythropoietin, were not systematically studied in these cohorts, but this would be an interesting avenue for future research.

Based on the observed greater sensitivity of the iFGF23 assay, we recommend measuring iFGF23 over cFGF23, using a cut point greater than 27 pg/mL across intact FGF23 assays, to confirm the diagnosis of FGF23-mediated hypophosphatemia. However, in most cases of FGF23 excess, the cFGF23 correlates moderately well with iFGF23, so use of the cFGF23 assay is a reasonable choice when iFGF23 measurement is not available. When making the diagnosis of TIO, after ruling out other causes of FGF23 excess, a cFGF23 cut point of 90 RU/mL will support the diagnosis with 100% sensitivity and specificity. cFGF23 levels between 45 and 90 RU/mL, however, should prompt repeat testing using an iFGF23 assay or further evaluation for either an acquired tubulopathy or genetic etiology. To mitigate confusion, commercial laboratories could consider including hypophosphatemic-specific references ranges with their results.

In conclusion, FGF23 measurements, including the ratio of cFGF23/iFGF23, are an important tool for informing the physiological relationship between phosphate and FGF23 and in evaluating patients with disorders of phosphate homeostasis. The introduction of cut points that more accurately support the diagnosis of various disorders into reporting on laboratory results will provide useful guidance to clinicians, allowing for more rapid and accurate diagnosis and ensuring appropriate disease-specific therapies.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.