C L I N I C A L R E S E A R C H A R T I C L E Prevalence and Mortality of Individuals With X-Linked Hypophosphatemia: A United Kingdom Real-World Data Analysis Downloaded from https://academic.oup.com/jcem/article/105/3/e871/5626435 by Universitat Autònoma de Barcelona user on 27 September 2021 Samuel Hawley,1 Nick J Shaw,2,3 Antonella Delmestri,1 Daniel Prieto-Alhambra,1,4 Cyrus Cooper,5 Rafael Pinedo-Villanueva,1 and M Kassim Javaid1,5 1 Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford, Oxford OX3 7LD, UK; 2Birmingham Women’s and Children’s Hospital NHS Foundation Trust, Birmingham B4 6NH, UK; 3Institute of Metabolism & Systems Research, University of Birmingham, Birmingham B15 2TT, UK; 4GREMPAL Research Group, Idiap Jordi Gol and CIBERFes, Universitat Autònoma de Barcelona and Instituto de Salud Carlos III, Barcelona 08193, Spain; and 5MRC Lifecourse Epidemiology Unit, University of Southampton, Southampton SO16 7YD, UK ORCiD number: 0000-0002-7034-6168 (S. Hawley), 0000-0001-7985-0048 (M. K. Javaid). Background: X-linked hypophosphatemia (XLH) is a rare multisystemic disease with a prominent musculoskeletal phenotype. We aim here to improve understanding of the prevalence of XLH across the life course and of overall survival among people with XLH. Methods: This was a population-based cohort study using a large primary care database in the United Kingdom (UK) from 1995 to 2016. XLH cases were matched by age, gender, and practice to up to 4 controls. Trends in prevalence over the study period were estimated (stratified by age) and survival among cases was compared with that of controls. Findings: From 522 potential cases, 122 (23.4%) were scored as at least possible XLH, while 62 (11.9%) were classified as highly likely or likely (conservative definition). In main analyses, prevalence (95% CI) increased from 3.1 (1.5–6.7) per million in 1995–1999 to 14.0 (10.8–18.1) per million in 2012–2016. Corresponding estimates using the conservative definition were 3.0 (1.4–6.5) to 8.1 (5.8–11.4). Nine (7.4%) of the possible cases died during follow-up, at median age of 64 years. Fourteen (2.9%) of the controls died at median age of 72.5 years. Mortality was significantly increased in those with possible XLH compared with controls (hazard ratio [HR] 2.93; 95% CI, 1.24–6.91). Likewise, among those with likely or highly likely XLH (HR 6.65; 1.44–30.72). Conclusions: We provide conservative estimates of the prevalence of XLH in children and adults within the UK. There was an unexpected increase in mortality in later life, which may have implications for other fibroblast growth factor 23–related disorders. (J Clin Endocrinol Metab 105: e871–e878, 2020) Key Words: XLH, hypophosphatemia, rickets, prevalence, survival, rate X -linked hypophosphatemic rickets (XLH) is a by renal phosphate wasting due to mutations in the rare multisystemic disease of mineral homeostasis PHEX gene (1). It is the most common form of herit- that has a prominent skeletal phenotype characterized able rickets (2). The key molecular mechanism involves excess fibroblast growth factor 23 (FGF23) production, ISSN Print 0021-972X ISSN Online 1945-7197 Printed in USA Abbreviations: ALP, alkaline phosphatase; CPRD, UK Clinical Practice Research Datalink; © Endocrine Society 2019. FGF23, fibroblast growth factor 23; GP, general practitioner; IQR, interquartile range; This is an Open Access article distributed under the terms of the Creative Commons ONS, Office for National Statistics; XLH, X-linked hypophosphatemia. Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits un- restricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. Received 15 July 2019. Accepted 14 November 2019. First Published Online 15 November 2019. Corrected and Typeset 17 February 2020. doi:10.1210/clinem/dgz203 J Clin Endocrinol Metab, March 2020, 105(3):e871–e878   https://academic.oup.com/jcem   e871 e872  Hawley et al   XLH: Prevalence and Associated Mortality J Clin Endocrinol Metab, March 2020, 105(3):e871–e878 a phosphatonin first identified in autosomal dominant for England and Wales were also obtained from linkage to hypophosphatemic rickets (3) and tumor-induced osteo- the Office for National Statistics (ONS) dataset which is con- sidered the gold standard for mortality data. Date of death malacia (4, 5). XLH usually manifests early in life with was therefore based on ONS data for the 60% of GP prac- shortened height and bowing of the legs, and while tices where this linkage was available and of sufficient quality, these can be improved with pharmacotherapy, they otherwise death date as recorded by CPRD was used (which likely persist into adulthood along with increased risk of has been shown to be comparable with the ONS(14)). Downloaded from https://academic.oup.com/jcem/article/105/3/e871/5626435 by Universitat Autònoma de Barcelona user on 27 September 2021 fractures, arthritis, dental abscesses, and enthesopathy There is no agreed algorithm for identifying cases with XLH (calcification of tendons and ligaments) (2, 6). using real world data. Methods from previous studies(10–12) were modified by the availability of data in the CPRD by 2 Traditional therapy for XLH includes activated experts in familial hypophosphatemia in children (NS) and vitamin D and oral phosphate, which while effective in adults (MKJ). A list of potential diseases that cause rickets/ increasing childhood growth, the therapy is poorly tol- osteomalacia and could represent XLH was used to extract erated and of unknown efficacy in adults with XLH (7). potential cases and their controls from the CPRD dataset Burosumab, a neutralizing antibody of FGF23, increases (Table 1). Next, an algorithm was developed to grade cases (“highly likely,” “likely,” “possible,” “unlikely” or “unable to serum phosphorus and improves rickets and linear determine”) using other primary care data that included: growth (8). In adults, burosumab also significantly im- proves serum phosphate as well as fracture healing, (a) Terms for other diagnoses that could mimic XLH, such as pain, and stiffness (9). However, this is a relatively novel cystinosis or fibrous dysplasia (Supplementary Item 1) (16), (b) Laboratory tests performed in the primary care setting therapy, and, in many countries, policy makers are un- including counts of: normal, low, and very low (< 75% clear about which adults should be eligible for therapy. of lower limit of normal range) serum phosphate; normal This is compounded by the scarcity of data on the preva- or high, low alkaline phosphatase (ALP); normal, low, lence and outcomes of adults with XLH. and high serum adjusted calcium; normal or high serum Three previous studies of the prevalence of XLH in parathyroid (PTH); estimated glomerular filtration rate (eGFR) < 15 mL/min/m2 or < 30–15mL/min/m2; normal children have used a mixture of hospital surveys and or high alanine transferase (ALT) values; normal or low registry data with conflicting prevalence rates (10–12), serum bicarbonate values. Existing CPRD reference ranges due in part to differences in criteria for case identifica- for laboratory values were used to estimate values for the tion and validation. Accurate data for adults is com- 20% of patients missing these ranges (Supplementary Item pounded by the lack of any standard management for 2) (16). adults with XLH in terms of monitoring laboratory (c) Prescriptions for more than 12 months of any activated vitamin D preparation, oral phosphate preparations, bicar- values, skeletal status, and other characteristics. Hence, bonate, potassium or calcium supplements (see Supplemen- it is possible that in the United Kingdom (UK), most tary Item 3 for drug codes) (16). adults are managed principally in the primary care set- A key feature of XLH is low serum phosphate. From the ex- ting. The National Health Service (NHS) healthcare tracted cases, individuals with 2 or more low phosphate values system within the UK has near universal coverage and or 50% or more low phosphate values if fewer than 3 values represents an opportune data resource to explore the were recorded or no phosphate results recorded were selected prevalence of a rare disease such as XLH and its as- for grading. Major criteria, unexpected criteria, and red flags sociated mortality rate. Our aim was to determine the were developed. Major criteria were high ALP, or more than 1 year of activated vitamin D therapy or phosphate therapy. prevalence of XLH in both children and adults and to Unexpected criteria included less than 50% serum phosphate describe survival across the life course using routinely values below the minimum reference range, presence of low collected medical data. ALP values or only normal ALP values, or a low adjusted serum calcium or bicarbonate value. Criteria that would discount a diagnosis of XLH were only low ALP values, 2 or Methods more low adjusted calcium values, 2 or more low bicarbonate values, treatment with bicarbonate or potassium supplements Study design and participants or a competing diagnosis. NS and MKJ independently graded This UK-based study used primary care health data obtained cases using these criteria (Supplementary Item 4) (16). Patients from the UK Clinical Practice Research Datalink (CPRD) with no recorded phosphate values and no unexpected or red GOLD dataset from 1995 to 2016. As of 2013, CPRD GOLD flags were designated as not able to be graded. covered more than 11.3 million patients from 674 general Given the expected excess of females with XLH, graders practitioner (GP) practices and had a representative coverage were blinded to the gender of potential patients. After the first of approximately 7% of the UK population (13). Only GP round of grading, a more detailed record was created for each practices that successfully complete the up to standard process patient discordantly graded highly likely/likely versus other are then included in the CPRD GOLD dataset. CPRD uses grades in a second round with the complete clinical record the Read code system, devised by Dr Read, a UK GP who pi- made available. In a post hoc analysis we explored the pres- oneered in data coding(14), and includes more than 100 000 ence of Read codes for renal stones and/or hypercalciuria codes for clinical events in primary care (15). Mortality data (codes in Supplementary Item 3 (16)). doi:10.1210/clinem/dgz203 https://academic.oup.com/jcem  e873 A physician verification questionnaire (Supplementary for cases and controls. Kaplan-Meier estimates were plotted Item 5 (16) was also sent to GPs of eligible patients with following univariable extended Cox regression using patient potential diseases that cause rickets/osteomalacia (codes in- age as the time axis (17). cluded in Table 1) and who were still alive and in a practice Prevalence and mortality analyses were repeated using only that agreed to participate in a survey consisting of 4 ques- cases (and their matched controls) graded as very likely or tions. The questionnaire was sent twice with a reminder likely, to assess if the main findings persisted when using a by CPRD. conservative case definition, albeit with a smaller sample size. Downloaded from https://academic.oup.com/jcem/article/105/3/e871/5626435 by Universitat Autònoma de Barcelona user on 27 September 2021 Up to 4 non-XLH controls (without any of the Read codes A further sensitivity analysis was performed for the mortality of interest) of same age, gender, and GP practice were matched analysis, in which eligible patients in England and Wales were to each potential XLH case. Data on demographics, prescrip- followed from index date until the earliest of death date or tions, laboratory test results, and comorbidities were extracted data extraction date (irrespective of transference out of prac- from CPRD GOLD for both potential cases and controls. tice and last collection date in CPRD) in order to make use of the continued linkage to ONS data. Statistical analysis To evaluate the reproducibility of the XLH likelihood Role of the funding source and ethics committee grading by the 2 clinicians, a weighted Kappa was calculated approval to assess concordance across the 5 different categories (after The funder of the study had no role in the study design, both the first and second round of grading). Descriptive stat- data collection, data analysis, data interpretation, or writing istics (counts and percentages) were used to assess agree- of the report. The corresponding author had full access to all ment between final likelihood gradings and returned GP of the data in the study and had final responsibility for the questionnaires. Temporal trends in XLH prevalence and decision to submit for publication. The Independent scientific associated 95% confidence intervals (CI) were estimated Advisory Committee (ISAC) approved the study and since by dividing mid-year counts of XLH cases by CPRD an- only de-identifiable routinely-collected data was used, no fur- nual denominator data. Prevalence estimates were gener- ther approvals were required. ated using all included cases and stratified by age in order to explore prevalence in childhood vs adulthood: younger than 16 years vs 16 years and older. Annual estimates were Results depicted graphically, and estimates reported separately for the first 5 years of the study period (1995–1999) versus last A total of 522 patients were identified using the initial 5 years (2012–2016). CPRD codes. Following the first round of independent An index date for each XLH patient was defined as the grading of potential cases, the weighted Kappa for first relevant Read code date after the GP practice of regis- inter-grader agreement was 0.88 (95% CI, 0.85–0.91). tration was flagged as “up to standard” by CPRD for clin- ical research. Index dates for controls were anchored to match Following independent re-grading of discordant cases, those of cases. Patients were followed from index date until it increased to 0.92 (95% CI, 0.89–0.94) when using 5 the earliest of: death, other types of transfer out of practice, or likelihood categories and to 0.98 (95% CI, 0.96–1.00) end of study. Deaths (and causes where available) were then identified and mortality incidence rates calculated separately Table 1. List of Read Codes Used to Identify Individuals With a Potential Diagnosis of XLH: Stratified by Final Grading for Likelihood of XLH Frequency Read Highly Likely Possible/ Unlikely Unable to TOTAL PPV Code Read Code Label likely potential determine (%) 2374 O/E —rickety rosary 0 0 0 4 2 6 0 C280.00 Active rickets 0 0 0 9 8 17 0 C281.00 Late effect of rickets 0 1 0 4 0 5 20 C282.00 Osteomalacia, unspecified 0 0 3 75 15 93 0 C28A.00 Vitamin D dependent rickets 0 0 0 8 0 8 0 C353000 Hypophosphatasia 3 6 45 129 57 240 3.8 C353100 Hypophosphatasia rickets 3 10 1 6 8 28 46.4 C353200 Vitamin-D-resistant rickets 3 3 2 9 9 26 23.1 C353211 Hypophosphatemic rickets 17 17 5 15 13 67 50.7 C353700 X-linked Hyposphataemic rickets 1 0 1 0 0 2 50.0 C353800 Autosomal dominant 0 0 0 1 1 2 0 hypophosphatemic rickets C353z00 Disorders of phosphorus 0 0 0 11 6 17 0 metabolism NOS K080300 Renal rickets 0 0 1 6 4 11 0 ALL 27 37 58 277 123 522 Abbreviations: O/E, on examination; PPV, positive predictive value (using highly likely and likely as reference standard). e874  Hawley et al   XLH: Prevalence and Associated Mortality J Clin Endocrinol Metab, March 2020, 105(3):e871–e878 when using 2 categories (likely and very likely vs other When using the more conservative sensitivity definition grades). of only highly likely and likely cases, prevalence esti- Of the 522 initially identified potential cases, 122 mates increased from 3.0 (95% CI, 1.4–6.5) per million were used in main analyses: 27 highly likely, 37 likely in the first 5 years to 8.1 (95% CI, 5.8–11.4) per mil- and 58 possible (Table 2). An expected preponderance lion in the last 5 years of the study period, Fig. 1B. As of females (70%), childhood diagnoses (median age per main analyses, prevalence per million increased in Downloaded from https://academic.oup.com/jcem/article/105/3/e871/5626435 by Universitat Autònoma de Barcelona user on 27 September 2021 at earliest Read code < 10 years) and phosphate sup- both pediatric (8.5 [95% CI, 2.7–26.4] to 14.6 [95% plementation (> 95%) was observed amongst the very CI, 8.1–26.6]) and adult (1.8 [95% CI, 0.6–5.4] to 6.7 likely and likely cases, although not in the “possible” [95% CI, 4.5–10.2]) populations. ones (Table 2 and Supplementary Item 6) (16). From the GP verification, there were 69/106 (65%) Mortality data questionnaires returned, of which a total of 17 (25%) Of the 610 patients included in the Cox model, 9 reported a definite/possible familial hypophosphatemic (7.4%) of the 122 cases died during a median of 4.6 condition (results shown in Supplementary Item (interquartile range [IQR], 2.1–8.6) years of follow-up, 7) (16). at a median age of 64 (IQR, 58–74) years. This equated to a rate of 12.1 (95% CI, 6.3–23.3)/1000 Prevalence person-years. Fourteen (2.9%) controls died during a In the main analysis, prevalence of XLH increased median of 4.2 (IQR, 2.84–8.09) years of follow-up, at from 3.1 (95% CI, 1.5–6.7) per million in the first median age of 72.5 (IQR, 29–71) years. This equated 5 years of the study (1995–1999) to 14.0 (95% CI, to a rate of 4.8 (95% CI, 2.8–8.1)/1000 person-years 10.8–18.1) per million in the last 5 years (2012–2016), and a hazard ratio of 2.93 (95% CI, 1.24–6.91) (Fig. (Fig. 1A). Likewise, prevalence per million increased in 2A). Causes of death were available in 9 (48%) deaths. both pediatric (9.4 [95% CI, 3.2–27.2] to 17.0 [95% Among cases, causes were: arthropathy (or related CI, 9.7–29.6]) and adult (1.8 [95% CI, 0.6–5.4] to 13.3 disorder), thyroid cancer, prostate cancer, and pneu- [95% CI, 9.9–17.9]) populations over the study period. monia. Amongst controls, the causes were external/ Table 2. Patient Characteristics of Initial Sample of 522 Potential XLH cases: Stratified by Final Grading for Likelihood of XLH (1995–2016) Grading Unable to Highly Likely Likely Possible Unlikely Determine TOTAL n 27 37 58 277 123 522 Female, n (%) 19 (70.4) 26 (70.3) 24 (41.4) 159 (57.4) 46 (37.4) 274 (52.5) Median age at earliest 7 (1–25) 3 (1–7) 47.5 (26–61) 37 (18–57) 2 (0–27) 29 (3–52) Read code, years (IQR) Median duration of 8.28 6.36 2.88 3.99 1.64 3.51 follow-up, years (IQR) (5.40–14.01) (2.29–10.01) (1.15–5.53) (1.43–9.64) (0.47–4.92) (1.29–8.30) Prior hypercalciuria or 2 (7.4) 0 (0) 0 (0) 15 (5.4) 1 (0.8) 18 (3.5) renal stones, n (%) Low serum phosphate 19 (70.4) 11 (29.7) 49 (84.5) 63 (22.7) 9 (7.3) 151 (28.9) count, n (%) High alkaline phosphatase 4 (14.8) 3 (8.1) 3 (5.2) 24 (8.7) 1 (0.8) 35 (6.7) count, n (%) Activated vitamin 19 (70.4) 6 (16.2) 12 (20.7) 68 (24.6) 2 (1.6) 107 (20.5) D therapy (ever initiated), n (%) Activated vitamin D 4 (14.8) 3 (8.1) 12 (20.7) 49 (17.7) 3 (2.4) 71 (13.6) therapy duration, years (IQR)a Phosphate supplements 27 (100) 35 (95.6) 9 (15.5) 49 (17.7) 10 (8.1) 130 (24.9) (ever initiated), n (%) Phosphate supplements 8.56 8.30 1.29 4.04 - 5.35 duration, years (IQR)b (5.02–17.21) (3.17–11.07) (0.55–7.42) (1.98–7.45) (2.38–11.08) Abbreviations: IQR, interquartile range. a 23 Patients on activated vitamin D unable to determine duration given missing dates. b 16 patients on phosphate supplements unable to determine duration given missing dates. doi:10.1210/clinem/dgz203 https://academic.oup.com/jcem  e875 Downloaded from https://academic.oup.com/jcem/article/105/3/e871/5626435 by Universitat Autònoma de Barcelona user on 27 September 2021 Figure 1. Prevalence of XLH by calendar year and age band for: (A) all possible cases of XLH and (B) only likely and highly likely cases of XLH. unspecified event (x2), circulatory system disorder, COPD, lung cancer, aortic valve disorder, and senility. Under the analyses using the most conservative defin- Figure 2. Kaplan-Meier plot for survival across lifespan for: (A) all ition of likely or highly likely grading only, 4 (6.5%) possible cases of XLH and (B) only likely and highly likely cases of XLH. cases died during a median follow-up of 7.3 (IQR, 3.0–12.4]) years, at a median age of 61 [IQR, 56–66]. being 15.1 (95% CI, 11.3–20.1) per million and 15.7 Three (1.2%) controls died during a median follow-up (95% CI, 11.8–20.9) per million, respectively. We ob- of 6.0 (IQR, 2.6–11.0) years, at a median age of 68 served an unexpected reduction in survival among XLH [IQR, 29–71], yielding a hazard ratio of 6.65 (95% cases relative to controls (with average age at death CI, 1.44–30.72) (Fig. 2B). being approximately 8 years younger in cases relative to For the sensitivity analysis in which follow-up was controls), which was evident when using either the more extended until end of study period (irrespective of or less conservative case definition. transfers out of practice and last collection date) for the Three studies have previously estimated the preva- sample of 535 patients in England and Wales, a total of lence of XLH, although these should be cautiously com- 23 deaths occurred (8 cases and 15 controls), yielding pared with the present investigation due to differences in a hazard ratio of 2.88 (95% CI, 1.18–7.00); P = 0.020. study populations and methods used. Most recently, 21 children were identified with hereditary rickets and con- Discussion firmed genetically from pediatric hospitals in Norway in 2009 (12), giving a prevalence of 1 in 60 000 (16.6 per This nationally representative study provides for the million) among individuals from birth to18 years of age. first time an estimated prevalence of XLH in both child- These children were also identified using the ICD 10 hood and adulthood, with 2016 estimates for these code E83.3 for disorders of phosphorus metabolism and e876  Hawley et al   XLH: Prevalence and Associated Mortality J Clin Endocrinol Metab, March 2020, 105(3):e871–e878 phosphatases in the Norwegian Patient Register, a na- (18) and humans (8, 9), suggesting a key role of FGF23 tionwide health administrative registry after discounting in XLH. Since its discovery (19), various data have been hypophosphatasia and other causes such as hyperpara- submitted to support an association between elevated thyroidism, renal tubular disorders such as Fanconi, and FGF23 and a number of common diseases (20–22) vitamin D deficiency. A Danish study used hospital med- and mortality (23), although further work is needed to ical records and ICD8/10 codes for rickets to identify confirm and/or further elucidate these findings. For in- Downloaded from https://academic.oup.com/jcem/article/105/3/e871/5626435 by Universitat Autònoma de Barcelona user on 27 September 2021 patients referred or discharged over a 20-year period stance, changes in cardiac size have not been seen in hyp in Southern Denmark (10), and reported a prevalence mice up to the age of 30 weeks (24), and neutralizing of 4.8 per 100 000 (48 per million) among those from FGF23 a) did not prevent cardiac hypertrophy in rats birth to 14.9 years of age. This estimate was based on that underwent subtotal nephrectomy, and b) increased 15 cases with XLH, but excluded patients aged 15 years mortality, this may reflect different effects of FGF23 or older and so did not estimate adult prevalence or out- blockade based on the underlying mechanisms of excess comes. The medical records were reviewed to exclude (25). The direct vs indirect role of FGF23 in vascular rickets from renal insufficiency, renal tubular acidosis, calcification remains controversial (26, 27). Untangling liver/biliary disease, malabsorption, related syndromes, a potential direct pathogenic role for FGF23 is relevant or iatrogenic causes. If cases with suspected nutritional for patients with XLH, as the current therapies for this rickets had a serum 25-hydroxy vitamin D of more than rare disease include phosphate and activated vitamin D 50 nmol/L they were also excluded. Biochemical criteria supplementation, which are associated with increases in included: raised serum ALP, low serum phosphate, and FGF23 in both adults and children (28) and worsening a normal serum calcium. In Japan, a survey was used to enthesopathy in mice (29). The consequences of FGF23 ascertain the numbers of patients with FGF23-related on survival in other rare bone diseases that involve disorders from 2005 to 2009 (11). Forty-one patients FGF23, such as fibrous dysplasia (30) and cutaneous with genetic hypophosphatemic diseases (63% female) skeletal hypophosphatemia syndrome (31), are also were identified and the incidence of XLH was estimated unknown. to be 1 per 20 000 live births (50 per million). Inclusion In XLH, it is likely that the FGF23 pathway may not and exclusion criteria for cases were not described nor be the only molecular mechanism perturbed. In the ab- was the potential bias from a nonrandom response rate sence of an XLH-specific scale, little correlation has been considered. There remains inconsistency in the methods found between broad measures of phenotype severity, for case ascertainment across studies and further col- such as presence of bowing and dental abscesses, with the laborative work is needed to standardize procedures for type or location of genetic mutation between individuals case validation and selection of controls. or even within families (32). These findings suggest a pos- Prevalence estimates in the current analysis generally sible involvement of other genetic and environmental fac- increased over the study period (1995–2016), although tors in determining the clinical phenotype of individuals this was more marked among adults and in all possible with XLH; although whether these are also implicated cases vs the more conservative definition (likely and highly in the observed reduction in survival is unknown. Future likely). The secular trend is most likely driven by changes work will seek to further elucidate the findings from our in clinical practice for laboratory testing, documenta- study, including analyses to better ascertain the natural tion and quality of coding during this time. For instance, history of XLH in terms of the burden of comorbidity, among all (522) potential XLH cases, the proportion with therapies and associated medical procedures using sec- no serum phosphate testing declined during the study ondary care data, which will also likely offer insight into period from 54% (1995–1999) to 42% (2012–2016) (re- the elevated mortality rate here observed. sults not shown), while for serum alkaline phosphatase this decline was from 34% to 27% for the same years. Strengths and limitations The mechanism for the observed increase in mor- This is the first study to examine the prevalence and tality in adults with XLH is not known. It could be that prognosis of XLH in adulthood. A major strength is the reduced survival is indirect, driven by an imbalance the generalizability of the findings which arise from a in comorbidities and other associated characteristics of population-based, real-world dataset having linkage to XLH patients relative to controls. Furthermore, a differ- gold-standard death registry data and by taking a con- ence in the management of comorbidities between cases servative approach to case identification with 2 methods and controls may also exist. A direct FGF23 pathway is for case ascertainment. The specificity that these cases also possible. It has previously been shown that neutral- represent patients with XLH is supported by the ob- izing circulating FGF23 reverses several key biochemical served excess of females expected from an X-linked dis- and musculoskeletal features of XLH in both animals order, even though graders were blinded to sex status of doi:10.1210/clinem/dgz203 https://academic.oup.com/jcem  e877 potential cases. Another strength is the ability to iden- may have led to relatively severe cases being favored tify relevant controls that allowed the determination of for inclusion, which may have subsequently impacted outcomes attributable to XLH. the generalizability of mortality estimates. However, it A major limitation is that no validated algorithm is worth noting that main findings remained unchanged exists for confirming the diagnosis of XLH in the pri- in a post hoc sensitivity analysis comparing all 522 po- mary case setting. The potential for miscoding rare tential XLH cases with their controls (mortality hazard Downloaded from https://academic.oup.com/jcem/article/105/3/e871/5626435 by Universitat Autònoma de Barcelona user on 27 September 2021 disease is also likely higher in the primary care setting ratio of 2.87 [95% CI, 2.08–3.95]; P < 0.001). Linked where coding is completed by a clinician, nurse, or ad- ONS data were only available for England and Wales ministrative staff with little or no training in rare dis- and the linkage also depended on local practice agree- eases; this was highlighted by the responses from the ments being in place, so only 60% of practices in total GP verification questionnaires, where cases were found were linked, although CPRD mortality data have pre- to be discordant to expert clinical grading (results in viously been shown to be reliable (irrespective of ONS Supplementary Item 7) (16). Given the lack of genetic linkage) (14). and radiological data in the primary and secondary care datasets and restrictions to directly contact patients, we were unable to genotype identified cases for the PHEX Conclusion mutation or check their radiological findings (the gold We describe conservative estimates of XLH prevalence standards for case validation). Furthermore, we omitted in both childhood and adulthood. We find a concerning diagnosis of Dent’s disease as a potential mimicking reduction in survival in adults that has not been pre- disease in the initial extraction, which meant 11 pa- viously reported and which may have implications for tients subsequently identified in CPRD GOLD using the other FGF23-related disorders. Read code for Dent’s disease were not considered. Read codes for hypercalciuria and renal stones identified only 4 of 122 (3.3%) adjudicated cases (compared to 17/400 Acknowledgments [4.3%] in matched controls), of whom only 2 (1.6%) Financial Support: Kyowa Kirin International provided (vs 16/400 [4.0%] of controls) had such a diagnosis funding for this project to the University of Oxford. The preceding the date of potential XLH diagnosis. Future funder was issued reports at the end of each stage of work algorithms should consider Dent’s disease in the differ- but had no other influence on the development, conduct, and ential diagnosis of XLH. reporting of the project. MKJ was supported by the National In addition, the Read coding system does not include Institute for Health Research (NIHR) Oxford Biomedical coding for family history of XLH that would further Research Centre (BRC). support the diagnoses. Due to information governance, Author Contributions: MKJ, RPV, NS, and DPA designed analysis of free text in the clinical or radiological re- the study. AD provided management and advanced curation ports was not possible, which may have helped by of the data. MKJ and NS performed case ascertainment. SH, describing known radiological features of XLH such as RPV, and MKJ performed data cleaning and analysis. MKJ mineralizing enthesopathy. Another limitation is that by and SH drafted the manuscript. All authors revised the manu- taking a very conservative approach to case identifica- script and approved the final version tion, we have likely underestimated the prevalence in ex- change for specificity that was required for the matched Additional Infromation analysis for prognosis. Even after calculation of missing laboratory tests, a number of potential participants were Correspondence: Muhammad Kassim Javaid, MBBS, still unable to be graded because of missing laboratory BMedSci, FRCP. PhD, The Botnar Research Centre, NDORMS, values (Supplementary Item 6) (16). As laboratory tests University of Oxford, Windmill Road, Oxford, OX3 7YD, performed in the secondary care setting are not included UK. E-mail: kassim.javaid@ndorms.ox.ac.uk Disclosure Summary: Kyowa Kirin International pro- in CPRD GOLD, the prevalence of young children, who vided funding for this project to the University of Oxford. are unlikely to have blood tests in primary care would Outside the submitted work, SH, NS, and AD report no con- have been underestimated. Similarly, given there is no flicts of interest. DPA reports grants from AMGEN, UCB agreed guideline for the routine care of adults with Biopharma, and Les Laboratoires Servier. CC reports per- XLH, incorporating frequency of laboratory testing sonal fees from Alliance for Better Bone Health, Amgen, and treatment with activated vitamin D and phosphate Eli Lilly, GSK, Medtronic, Merck, Novartis, Pfizer, Roche, supplements into our case adjudication process may Servier, Takeda, and UCB. RPV has received research grants have missed patients with XLH, leading to a conser- from Kiowa Kirin, and consultancy fees outside the scope of vative prevalence of XLH. Were this to be the case, it this study. MKJ reports grants from AMGEN, Kyowa Kiran e878  Hawley et al   XLH: Prevalence and Associated Mortality J Clin Endocrinol Metab, March 2020, 105(3):e871–e878 Hakin, and consultancies from AMGEN, Internis, consilient 16. Supplementary material for “prevalence and mortality of individ- Health, Mereo Biopharma, Kyowa Kirin Hakin and UK uals with X-linked hypophosphataemia in the United Kingdom: a real-world analysis” [Internet]. 2019. Available from University Data Availability: Restrictions apply to the availability of of Oxford depository, date of deposit 21.09.2019: http://users. data generated or analyzed during this study to preserve patient ox.ac.uk/~orms0122/ confidentiality or because they were used under license. The cor- 17. Lamarca R, Alonso J, Gómez G, Muñoz A. Left-truncated responding author will on request detail the restrictions and any data with age as time scale: an alternative for survival ana- Downloaded from https://academic.oup.com/jcem/article/105/3/e871/5626435 by Universitat Autònoma de Barcelona user on 27 September 2021 conditions under which access to some data may be provided. lysis in the elderly population. J Gerontol A Biol Sci Med Sci. 1998;53(5):M337–M343. 18. Aono Y, Yamazaki Y, Yasutake J, et al. Therapeutic effects of anti- FGF23 antibodies in hypophosphatemic rickets/osteomalacia. J References Bone Miner Res. 2009;24(11):1879–1888. 19. Yamashita T, Yoshioka M, Itoh N. Identification of a novel 1. A gene (PEX) with homologies to endopeptidases is mutated fibroblast growth factor, FGF-23, preferentially expressed in the in patients with X-linked hypophosphatemic rickets. The HYP ventrolateral thalamic nucleus of the brain. Biochem Biophys Res Consortium. Nat Genet. 1995;11(2):130–136. Commun. 2000;277(2):494–498. 2. Carpenter TO, Imel EA, Holm IA, Jan de Beur SM, Insogna KL. 20. Gutiérrez OM, Januzzi JL, Isakova T, et al. Fibroblast growth A clinician’s guide to X-linked hypophosphatemia. J Bone Miner factor 23 and left ventricular hypertrophy in chronic kidney Res. 2011;26(7):1381–1388. disease. Circulation. 2009;119(19):2545–2552. 3. ADHR Consortium. Autosomal dominant hypophosphataemic 21. Mathew JS, Sachs MC, Katz R, et al. Fibroblast growth factor-23 rickets is associated with mutations in FGF23. Nat Genet. and incident atrial fibrillation: the Multi-Ethnic Study of 2000;26(3):345–8. Atherosclerosis (MESA) and the Cardiovascular Health Study 4. Cai Q, Hodgson SF, Kao PC, et al. Brief report: inhib- (CHS). Circulation. 2014;130(4):298–307. ition of renal phosphate transport by a tumor product 22. Yao XY, Li S, Zhang LG, Liu ZH, Bao JN, Wu ZY. Higher serum in a patient with oncogenic osteomalacia. N Engl J Med. fibroblast growth factor-23 levels and the risk of stroke and its 1994;330(23):1645–1649. subtypes: evidence from a meta-analysis of prospective studies. J 5. Shimada T, Mizutani S, Muto T, et al. Cloning and characteriza- Stroke Cerebrovasc Dis. 2018;27(11):3076–3083. tion of FGF23 as a causative factor of tumor-induced osteomal- 23. Souma N, Isakova T, Lipiszko D, et al. Fibroblast growth acia. Proc Natl Acad Sci U S A. 2001;98(11):6500–6505. factor 23 and cause-specific mortality in the general popula- 6. Pettifor JM. What’s new in hypophosphataemic rickets? Eur J tion: The Northern Manhattan Study. J Clin Endocrinol Metab. Pediatr. 2008;167(5):493–499. 2016;101(10):3779–3786. 7. Imel EA, White KE. Pharmacological management of 24. Liu ES, Thoonen R, Petit E, et al. Increased circulating FGF23 X-linked hypophosphataemia. Br J Clin Pharmacol. does not lead to cardiac hypertrophy in the male hyp mouse 2019;85(6):1188–1198. model of XLH. Endocrinology. 2018;159(5):2165–2172. 8. Carpenter TO, Whyte MP, Imel EA, et al. Burosumab therapy 25. Shalhoub V, Shatzen EM, Ward SC, et al. FGF23 neutralization in children with X-Linked hypophosphatemia. N Engl J Med. improves chronic kidney disease-associated hyperparathyroidism 2018;378(21):1987–1998. yet increases mortality. J Clin Invest. 2012;122(7):2543–2553. 9. Insogna KL, Briot K, Imel EA, et al. A randomized, double-blind, 26. Yamada S, Giachelli CM. Vascular calcification in CKD- placebo-controlled, phase 3 trial evaluating the efficacy of MBD: Roles for phosphate, FGF23, and Klotho. Bone. burosumab, an anti-FGF23 antibody, in adults with X-linked 2017;100:87–93. hypophosphatemia: week 24 primary analysis. J Bone Miner Res. 27. Ketteler M, Biggar PH, Liangos O. FGF23 antagonism: the thin 2018;33(8):1383–1393. line between adaptation and maladaptation in chronic kidney 10. Beck-Nielsen SS, Brock-Jacobsen B, Gram J, Brixen K, Jensen TK. disease. Nephrol Dial Transplant. 2013;28(4):821–825. Incidence and prevalence of nutritional and hereditary rickets in 28. Imel EA, DiMeglio LA, Hui SL, Carpenter TO, Econs MJ. southern Denmark. Eur J Endocrinol. 2009;160(3):491–497. Treatment of X-linked hypophosphatemia with calcitriol 11. Endo I, Fukumoto S, Ozono K, et al. Nationwide survey of and phosphate increases circulating fibroblast growth fibroblast growth factor 23 (FGF23)-related hypophosphatemic factor 23 concentrations. J Clin Endocrinol Metab. diseases in Japan: prevalence, biochemical data and treatment. 2010;95(4):1846–1850. Endocr J. 2015;62(9):811–816. 29. Karaplis AC, Bai X, Falet JP, Macica CM. Mineralizing 12. Rafaelsen S, Johansson S, Ræder H, Bjerknes R. Hereditary enthesopathy is a common feature of renal phosphate- hypophosphatemia in Norway: a retrospective population-based wasting disorders attributed to FGF23 and is exacer- study of genotypes, phenotypes, and treatment complications. bated by standard therapy in hyp mice. Endocrinology. Eur J Endocrinol. 2016;174(2):125–136. 2012;153(12):5906–5917. 13. Herrett E, Gallagher AM, Bhaskaran K, et al. Data resource 30. Riminucci M, Collins MT, Fedarko NS, et al. FGF-23 in fibrous profile: clinical practice research datalink (CPRD). Int J dysplasia of bone and its relationship to renal phosphate wasting. Epidemiol. 2015;44(3):827–836. J Clin Invest. 2003;112(5):683–692. 14. Gallagher AM, Dedman D, Padmanabhan S, Leufkens HGM, 31. Lim YH, Ovejero D, Sugarman JS, et al. Multilineage somatic de Vries F. The accuracy of date of death recording in the Clinical activating mutations in HRAS and NRAS cause mosaic cutaneous Practice Research Datalink GOLD database in England com- and skeletal lesions, elevated FGF23 and hypophosphatemia. pared with the Office for National Statistics death registrations. Hum Mol Genet. 2014;23(2):397–407. Pharmacoepidemiol Drug Saf. 2019;28(5):563–569. 32. Holm IA, Nelson AE, Robinson BG, et al. Mutational ana- 15. Davé S, Petersen I. Creating medical and drug code lists to iden- lysis and genotype-phenotype correlation of the PHEX gene in tify cases in primary care databases. Pharmacoepidemiol Drug X-linked hypophosphatemic rickets. J Clin Endocrinol Metab. Saf. 2009;18(8):704–707. 2001;86(8):3889–3899.