Introduction

Since its first description in 1979 (Peterson M.E., Johnson J.G., Andrews L.K. Spontaneous hyperthyroidism in the cat. [abstract]. American College of Veterinary Internal Medicine. Seattle WA, 1979:108), the prevalence of hyperthyroidism has dramatically increased making it the most commonly diagnosed endocrinopathy in the cat [1–3]. The diagnosis of hyperthyroidism relies on evidence of thyroid hyperfunction (presence of compatible clinical signs, a palpable thyroid nodule, and elevated total thyroxine (TT4) concentrations) [4]. Clinicians rely heavily upon the TT4 concentration reference interval (RI) provided by diagnostic laboratories to make diagnostic and therapeutic decisions. These RIs are generally based on a relatively small population of healthy, young cats, a practice which is commonplace in veterinary medicine [5]. Reference intervals are generally determined by the nonparametric method of percentile estimates with confidence intervals to determine the central 95th percentile interval (i.e., 2.5^th through 97.5^th percentile range) for results from > 120 clinically normal subjects.

In studies where the health of the patients cannot be established, the data is often evaluated using Tukey’s method [6], which is appropriate for analyzing large univariate data sets. This methodology is valid and useful for establishing expected and unexpected values in a large population of patients, even when data is derived from patients with an unknown health status [7–8]. For each data set the median is defined as the middle value of the distribution where 50% of data results are higher and 50% are lower. The interquartile range (IQR) is defined as the range of values that reside in the middle 50% of concentrations between the lower quartile (25^th percentile) and upper quartile (75^th percentile). The expected distribution of TT4 concentration is defined as those contained within an upper value (1.5*IQR+ upper quartile) and a lower value (1.5*IQR-lower quartile). An unexpected TT4 concentration signifies an outlier, even if the value is within the laboratory’s RI, and is defined as a data point that is outside of the expected concentration for the sampled population [9–10].

Clinicians frequently encounter patients with compatible physical examination findings (enlarged thyroid) and/or clinical signs of hyperthyroidism but with TT4 concentrations within the laboratory’s RI. Such cats may be truly normal (non-hyperthyroid) or abnormal (hyperthyroid with concurrent non-thyroidal disease or early, emerging disease) requiring additional conformational testing or monitoring [11]. The identification of an unexpected TT4 value above which patients are considered statistical outliers would allow practitioners to recognize potentially hyperthyroid patients even when values come back within the RI. Recognizing unexpected TT4 concentrations may be a clinically useful piece of information, as it may prompt the clinician to pursue additional endocrine tests in a cat with clinical signs of hyperthyroidism or more frequently monitor an asymptomatic cat that may eventually progress to hyperthyroidism. Previous authors have suggested that values in the upper 33–50% of the RI in patients suspected as being hyperthyroid should undergo additional testing but the basis for these recommendations has not been validated in a large population and no specific cut-offs have been available [12–14].

The objectives of the study reported here were to 1) Determine the expected distribution of TT4 concentration and outlier values based on a large population of cats encompassing all ages and breeds in the US, 2) compare these TT4 concentrations to the standard laboratory RI, and 3) determine if and when the unexpected TT4 concentration changes with age.

Materials and methods

Results

Using the search criteria above, an initial 543,861 TT4 concentrations were identified between January 2014 and July 2015. Data were compiled from a total of 8,673 hospitals in the US. Samples were categorized into the following groups based on sex and reproductive status; female intact (n = 12,672), female spayed (n = 266,532), male intact (n = 10,145), male neutered (n = 254,512). Of the total samples tested, 485,195 samples comprising 89% of the total reference values were categorized as domestic short hair, domestic medium hair, domestic long hair, not identified, unknown, or blank. The remaining 58,666 samples tested comprising 11.0% of the total reference values were represented by 41 different Cat Fanciers’ Association breeds. TT4 concentrations greater than 20 ug/dL (n = 4,135), TT4 concentrations belonging to cats older than 20 years of age (n = 2,309), and TT4 concentrations for which there were no assigned age or obvious misinformation (n = 5,652) were excluded from the data set leaving a total of 531,765 unique TT4 concentrations. In addition, removing cats > 20 years and those with results > 20 ug/dl did not change any of the data conclusions regarding upper fences or outliers.

The data was broken down into age categories including <1 year and years 1 through 20. For each age group, the median value, lower and upper quartiles, IQR, expected TT4 concentration, and outlier values were determined (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Expected and unexpected TT4 concentrations in cats grouped by age.

TT4 concentrations in each age group are reported. The values are described by 25^th percentile, 50^th percentile (median), 75^th percentile, IQR, expected TT4 concentration (low and high), and outlier values. The gray shading represents identical values for expected and unexpected TT4 concentrations in cats aged 1 through 9 years. The number of TT4 samples that exceed 3.5 ug/dL, and the percentage of these unexpected outliers in each age group are reported.

https://doi.org/10.1371/journal.pone.0213259.t001

For all cats in the data set, the median TT4 concentration was 1.95 ug/dL and the expected distribution of TT4 concentration was 0.5–4.3 ug/dL. For all cats between the age of 1 and 9 years (n = 141,294) the median TT4 concentration, lower and upper quartiles, IQR, expected TT4 concentration, and outlier values were identical (Table 1).

These values were determined to be; median 1.9 ug/dL, lower quartile 1.5 ug/dL, upper quartile 2.3 ug/dL, IQR 0.8, expected TT4 concentration 0.5–3.5 ug/dL, and outlier > 3.5 ug/dL. In cats <1 year of age, the values were determined to be; median 1.8 ug/dL, lower quartile 1.4 ug/dL, upper quartile 2.4 ug/dL, IQR 1.0, expected TT4 concentration 0.5–3.9 ug/dL, and outlier > 3.9 ug/dL. As age increased above 9 years, the distribution of values shifted to the right as these populations contained a greater percentage of cats with higher TT4 concentrations. The number of cats in each age group that had a TT4 concentration greater than 3.5 ug/dL, defining a statistical outlier, as well as the percentage of outliers calculated as the number of outliers divided by the total number of cats of a particular age group were determined (Table 1). The number and percentage of outliers increased with age, particularly in cats over the age of 9 years. A graphical representation of the percentage of outliers as a function of age is shown in Fig 1. The shape of the line representing this relationship was determined to have a 3^rd degree polynomial distribution y = -0.0001x³ + 0.0046x²–0.0292x + 0.0666; R² = 0.9956. Thus, 99.56% of the variation of values outside of 3.5 ug/dL can be explained as a function of age.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Percentage of unexpected outliers as a function of age.

The graph depicts the percentage of unexpected outliers as a function of age (red squares), which is best depicted as a 3^rd order polynomial equation (solid black). The 3^rd order polynomial equation and R² values are illustrated.

https://doi.org/10.1371/journal.pone.0213259.g001

There were 3 age groups for which the rate of change (slope) of unexpected outliers were identifiable, including 3–7 years, 7–14 years, and 14–20 years. The rate increase was highest from 7–14 years where the slope was 4.6 times faster than from 3–7 years of age. Cats 14–20 years of age had a rate increase that was 1.5 times faster than young cats but 3 times slower than cats aged 7–14 years (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Rate of change of unexpected feline TT4 concentration by age.

The graph depicts the slopes for 3 different age groups; young 3–7 years (blue diamonds), adult 7–14 years (orange squares), and senior 14–20 years (grey triangles). Equations calculating the slope of the line and R² values are illustrated.

https://doi.org/10.1371/journal.pone.0213259.g002

Discussion

This study differs from previous studies by examining a large population of cats of all ages throughout the United States. As the data was not normally distributed and the health of the patients was not established, the data was evaluated using Tukey’s method which is appropriate for analyzing large univariate data sets and such approaches are distribution independent. By defining exclusion bounds (outliers) rather than the inclusion bounds determined by RI’s, we are able to identify those patients at risk of hyperthyroidism even when their TT4 values fall within the laboratory’s RI.

By examining a total of 531,765 unique TT4 concentrations that were measured using the same validated immunoassay it was shown that for all cats in the data set, the median TT4 concentration was 1.95 ug/dL and the expected distribution of TT4 concentration was 0.5–4.3 ug/dL. This is very similar to the laboratory’s RI of 0.8–4.0 ug/dl which was determined in >1,000 healthy cats between the ages of 1–3 years¹⁵ and a similar RI of 0.8–3.8 ug/dl found when evaluating 131 healthy, euthyroid cats, over the age of 7 years [16]. This highlights that the use of such large datasets, which may allow for the inclusion of some ill and/or hyperthyroid cats even with the exclusion criteria mentioned previously, in conjunction with appropriate statistical methods, carries sufficient statistical power to demonstrate the homogeneity of the TT4 data.

Perhaps the most notable finding in this paper was that cats between 1–9 years of age had an identical expected distribution of TT4 concentration (0.5–3.5 ug/dL). Cats with TT4 concentrations greater than 3.5 ug/dL represented unexpected outliers in the population even though TT4 concentrations between 3.5–4.0 ug/dl are considered to be within the established RI for the laboratory. Based on this information, the finding of a TT4 between 3.5–4.0 ug/dl, along with historical and physical examination findings suggestive of hyperthyroidism, indicates the need for additional testing or monitoring.

A previous study by Peterson et al [17] which compared TT4 concentrations in cats with untreated hyperthyroidism, non-thyroidal disease, and clinically normal cats demonstrated that there existed a population of cats with mild hyperthyroidism with TT4 concentrations within the RI for the laboratory. These cats typically had a TT4 value greater than 2.3 ug/dl, which was within the laboratory's established RI. Multiple authors have made the recommendation that practitioners consider the possibility of hyperthyroidism when TT4 values are in the middle to upper range of the RI [12–14], however this finding has not been validated in a larger data set, nor has a different statistical methodology been performed to further substantiate that approach. Interestingly, our study which evaluated expected and unexpected TT4 concentrations using a population of 531,765 cats with unknown health status resulted in a similar finding despite the utilization of a completely different statistical methodology and TT4 assay. Based on the results of this very large data set, TT4 values >3.5 ug/dl represent statistical outliers in the population lending support to the recommendation that TT4 values above this cutoff using this TT4 assay should prompt further evaluation by the practitioner. The evaluation of the whole population within an age bracket in detecting outliers represents both physiologic and pathologic conditions. As the number and percentage of outliers was identical through age 9, the change in outliers with increasing age more likely reflects pathology and is consistent with the known natural history of hyperthyroidism and the observed increase in the rate of change (slope) of unexpected outliers with age observed in this study.

The development of RIs in a given population was a concept first introduced in human medicine and later applied to veterinary medicine [18–21]. Laboratory-generated RIs are commonly used to assist in the interpretation of a given observation when the same laboratory technique is used in a corresponding population of animals of the same species.¹⁹ In veterinary medicine, RIs are generally based on a relatively small sample size. Although there are conventional methodologies for establishing RIs using a small number (i.e. <120) of reference values [5,22], the current study intended to more accurately describe the expected TT4 concentration in cats using a vast data set with greater than 500,000 reference values of TT4. The intent of the present study was not to establish a RI, but rather to utilize the above statistical methodology to determine which TT4 concentrations are expected and which represent outliers, even when such outliers are within the laboratory’s established RI. The samples utilized to determine expected TT4 concentrations in the present study were obtained from cats of all ages and breeds from every part of the US, but without accompanying information regarding health status or clinical circumstance. Generally, when reference values are obtained from individuals in well-defined populations or with established health status, all reference values tend to be retained. Alternatively, in instances where reference values are selected by convenience, where health status is not established or confirmed, or when field methods introduce high levels of inaccuracy, reference values located at extremities are more likely to be excluded [5].

The expected TT4 values were determined using non-parametric methods of percentile estimates with confidence intervals to determine the central 95th percentile interval (i.e., 2.5^th through 97.5^th percentile range) and the expected TT4 concentrations in the 531,765 cats was determined using Tukey’s method which can be used for data sets where the health status of the population is unknown such as determination of RIs in wildlife species [5,22–23].

Limitations of the present study include its retrospective and observational design. Certain necessary assumptions were made in the present study with regards to the clinical subjects, recording of information, and accuracy of the analysis so that conclusions could be drawn from the data. The first assumption was that younger cats are generally healthy, euthyroid, and without concomitant severe disease that substantially alters their TT4 concentrations when compared to older cats. This assumption is supported by the fact that hyperthyroidism tends to be a disease of geriatric cats. The average age of hyperthyroid cats is 13 years, and the disease uncommonly occurs in cats younger than 10 years of age [3,24–25]. Undoubtedly, a small subset of young cats with elevated TT4 concentrations and clinical signs of hyperthyroidism were included in the current data set, however these values did not meaningfully influence the expected TT4 concentration as cats 1–9 years of age had an identical median and expected distribution of TT4 concentration (0.5–3.5 ug/dL).

An assumption was made about the accuracy of the reported age, sex, reproductive status, and breed by the submitting hospital. While there are recording errors that do occur, the signalment corresponding to the individual data points was accepted to be accurate in the current study. Moreover, while there may have been errors in TT4 concentrations due to pre-analytical errors (i.e. sample handling, shipment, storage), analytical errors, or post-analytical (transcription) errors [5], the TT4 concentrations were assumed to be correct and reflective of the true TT4 concentration in that patient. Additionally, while it was our intent to include only unique TT4 concentrations into the data set, we cannot exclude the possibility of redundant samples in which a specific patient had TT4 concentration performed at two or more separate hospitals using the same reference laboratory during the study period. While these assumptions were not absolute across all measurements, the robust statistical power of this particular data set precludes these exceptions from strongly influencing the overall results. It is important to reiterate that results of the current study pertain only to TT4 concentrations measured by an homogenous EIA method^c [15–16,26] at a major commercial veterinary laboratory^b, and not to TT4 concentrations measured by other assay types (i.e. RIA, chemiluminescent enzyme immunoassay, enzyme-linked immunosorbent assay) [4,27–29]. Despite, the reasonable correlation among varying assay types [30–31], in order for the present interpretations to be extrapolated to other assays, individual diagnostic laboratories would need to perform similar evaluations using their specific, corresponding TT4 assay.

Diagnosis of feline hyperthyroidism relies on compatible clinical signs (i.e. polyuria, polydipsia, polyphagia, weight-loss, altered behavior), identification of a thyroid nodule, and an elevation in circulating thyroid hormone levels. Treatment of hyperthyroidism regardless of modality (i.e. radioactive iodine, pharmacological management with anti-thyroid medications, surgical thyroidectomy, or an iodine-restricted diet) carries a favorable prognosis [15,32–36]. Thus, it is logical to surmise that earlier identification of at-risk patients should improve the overall clinical outcome. The hallmark for diagnosis of hyperthyroidism remains an elevated circulating TT4 concentration, and clinicians rely heavily upon the TT4 RI provided by diagnostic laboratories to make clinical decisions. The majority of overtly hyperthyroid cats have TT4 levels above the individual laboratory’s RI [17–37], although there are also a subset of cats with clinical signs referable to hyperthyroidism that have TT4 concentrations within the “normal” RI, a disease termed occult hyperthyroidism [38]. Based on our data which defines an outlier TT4 concentration as > 3.5 ug/dl, which is lower than the upper end of the RI generated by the reference laboratory (e.g. 0.8–4.0 ug/dL), it is likely that practitioners will identify emerging and occult hyperthyroid cats earlier than they would otherwise, thereby potentially improving the clinical outcome for their patients.

This study demonstrated that age was a strong predictor for the number and percentage of unexpected outliers in the population, accounting for 99.5% of the variation in outlier TT4 concentrations. The greatest change in the number and percentage of outliers occurred in cats aged 7–14 years, which suggests that cats in this age group are more likely to become outliers, and thus should be more closely scrutinized for abnormal TT4 concentrations and clinical signs of hyperthyroidism compared to cats of other age groups. These results are consistent with the observation that hyperthyroidism occurs most commonly in older cats [3,24].

Since its first description in 1979, the incidence of hyperthyroidism has dramatically increased prompting veterinarians and researchers to hypothesize whether exposure to environmental thyroid-disruptor chemicals or other environmental, genetic or dietary factors are involved in the pathogenesis of hyperthyroidism [1]. Potential exposure to several substances have been implicated including organohalogen compounds such as polychlorinated biphenyls and polybrominated diphenyl ethers [39], fertilizers [40], soy isoflavones [38–42], bisphenol-A primarily released from “pop-top” canned cat food lids [43–44], and consumption of commercial canned food [40,43]. One theory as to why the number and percentage of unexpected outliers becomes accelerated over the age of 9 years is that it may take several years of exposure to such environmental, dietary and genetic factors before they express themselves clinically and hyperthyroidism ensues, although this topic requires further investigation.

Based on the current study, the identification of a TT4 concentration greater than 3.5 ug/dL is an unexpected finding and characterizes an outlier in the general population. Irrespective of clinical signs of hyperthyroidism, the identification of an outlier warrants suspicion for the possibility of hyperthyroidism. If the TT4 concentration is greater than 3.5 ug/dL but less than the upper limit of the reporting laboratory’s RI, and the patient exhibits clinical signs of hyperthyroidism, further evaluation is warranted to rule in or rule out hyperthyroidism such as by measurement of free T4 by equilibrium dialysis (fT4ED) or thyroid stimulating hormone (TSH) where a normal TSH concentration can be used to rule-out hyperthyroidism [16]. The identification of a TT4 concentration greater than 3.5 ug/dL in the absence of clinical signs of hyperthyroidism may represent a cat with early or emerging hyperthyroidism. This finding should prompt a high index of suspicion for hyperthyroidism going forward, which may include measurement of fT4ED and/or TSH or close monitoring of that patient’s TT4 and clinical signs in the subsequent weeks to months.

With the establishment of this new expected TT4 concentration in the cat, future studies will focus on determining how breed and geographical distribution within the US affects the expected TT4 concentration. It would be of interest to determine the true incidence and prevalence of hyperthyroidism in the US and different parts of the world, and to evaluate changes in prevalence over time by repeating this study several years from now to determine if the expected TT4 concentration or outlier population changes.

Acknowledgments

The authors thank Alison Motsinger-Reif Ph.D. for assisting with statistical analyses, and Carolyn Cray Ph.D. for her critical review of the manuscript.