determining the effective factors on tsh level in the
TRANSCRIPT
JP Journal of Biostatistics
© 2020 Pushpa Publishing House, Prayagraj, India
http://www.pphmj.com
http://dx.doi.org/10.17654/BS017020477
Volume 17, Number 2, 2020, Pages 477-490 ISSN: 0973-5143
Received: July 13, 2020; Accepted: August 12, 2020
Keywords and phrases: hyperthyroidism, radioactive iodine, generalized estimating equations,
mixed effects model.
∗Corresponding author
DETERMINING THE EFFECTIVE FACTORS ON TSH
LEVEL IN THE PATIENTS WITH GRAVES’
DISEASE: A LONGITUDINAL STUDY
Atefeh Malekhoseiny1, Mohammadreza Rezvanfar
2, Azam Moslemi
1,
Fatemeh Rafiei3, Mohammad Rafiei
1,*, Faezeh Rezvanfar
4 and
Kosar Hajdezfulian4
1Department of Biostatistics
School of Medical Sciences
Arak University of Medical Sciences
Arak, Iran
e-mail: [email protected]
2Department of Endocrinology
Thyroid Disorders Research Center
Arak University of Medical Sciences
Arak, Iran
e-mail: [email protected]
3Department of Biostatistics and Epidemiology
School of Health
Scientific Research Center
Tehran University of Medical Sciences
Tehran, Iran
e-mail: [email protected]
Atefeh Malekhoseiny et al. 478
4School of Medicine
Arak University of Medical Sciences
Arak, Iran
e-mail: [email protected]
Abstract
Introduction: One of the most important thyroid disorders is
hyperthyroidism resulting from Graves’s disease. One of the most
common therapies is using radioactive iodine, called iodine therapy.
This study aims to identify the factors effective on TSH level variation
in the patients with Graves’s disease under treatment with radioactive
iodine.
Materials and Methods: This is a longitudinal-observational study.
201 patients with Graves’s disease, who were under treatment with
radioactive iodine, were entered into the study. TSH levels in the
patients’ blood were measured prior to the treatment with radioactive
iodine, and 2, 4 and 6 months after the treatment. The marginal model
by generalized estimating equations, and mixed effects model were
fitted in order to analyze the effect of each predictive variable on the
trend of TSH levels. The statistical analysis was performed by using
R3.5.2 software, and for all the tests, the level of significance was
considered 0.05.
Results: Mean age of patients was 37.54 ± 13.66 years, out of which
61.7% were female. In both the methods, the interaction of time
duration, and baseline TSH were significant ( ).05.0<P In addition,
fitting the mixed effect model showed that the dose of Methimazole
had significant effect on the TSH level ( ).05.0<P
Conclusions: In this study, results of both the models were similar.
The effects of baseline TSH, dose of administered Methimazole and
the interaction of time and the dose of radioactive iodine led to the
increase of TSH in the patient’s bloodstream.
Determining the Effective Factors on TSH Level … 479
Introduction
The thyroid disorders belong to the most common endocrine disorders
in the world [1]. The thyroid gland controlling mechanism is as such: the
reduction in amount of the Thyroid hormones ( )43 and TT leading to the
increase of producing the thyroid stimulating hormone (TSH), and vice versa
(2, 3). Hyperthyroidism is a case where the thyroid gland is over-activated.
This may lead to the reduction or even lack of TSH in the patient’s
bloodstream. Its incidence occurs in women 10 times more than men [4, 5].
The incidence of hyperthyroidism in the world has been reported to be
0.2% to 1.3% [6]. In a study, it has been reported that the incidence of
hyperthyroidism in an Iranian adult population was 1.6% [7]. The major
causes of hyperthyroidism are Graves’s disease (GD), multi-nodular toxic
goiter, and the toxic adenomas of the thyroid gland [8].
GD is an auto-immunity disorder. Its main cause is the formation of
antibodies against thyrotropin receptors (TSHR) which activate the thyroid
cells to increase their activities. Often, genetic factors, stress, pregnancy, and
inflammation are regarded as the main causes of GD [9]. Surgical operation,
taking anti-thyroid drugs, and using radioactive iodine (I131) are the main
therapies for GD. The method of iodine therapy is more attracting due to its
ease of application, repeatability, cheapness, fewer side effects and the less
probability of relapse after the treatment. On the other hand, the therapeutic
methods are not completely safe, and it may result into some side-effects
such as the prevalence of hypothyroidism in the patients under treatment.
Also, the attempts of determining the optimum dose of I131 without
considering hypothyroidism side-effects, the treatment of hyperthyroidism
will remain useless [10, 11].
The longitudinal data is repeated observations of the studied individuals
over a time period. It is assumed that in this series of data the individuals
are independent; but, everyone’s responses over time are correlated. In
these studies, often, with the passage of time, correlation of everyone’s
observations reduces; and, the assumption of variance homogeneity is not
Atefeh Malekhoseiny et al. 480
set. Also, the possibility of lost data, and the number of unequal repetitions
especially in clinical trials cannot be ignored [12, 13].
The marginal model with generalized estimating equations (GEE), and
mixed effects model (MEM) are the best methods of analyzing the
longitudinal data [14-16].
The main objective of the present study is to identify the factors
effective on the variation of TSH levels in the patients with GD, who have
gone under treatment with radioactive iodine, by using the method of GEE,
and MEM.
Methods
201 patients with hyperthyroidism caused by GD, who were under
treatment with I131 according to the physician’s diagnosis, were entered in
this longitudinal study. The diagnosis of GD was based on elevated higher
free thyroxine (FT4) with suppressed TSH level and clinical examination
findings (such as diffusely enlarged goiter and typical signs and symptoms of
hyperthyroidism) with documented presence of thyroid autoantibodies. After
diagnosis of hyperthyroidism, the patients took at least for a month anti-
thyroid Methimazole drug (MMI) until the thyroid performance became
natural, i.e., euthyroidism and patients get ready to receive I131. After
euthyroidism, and prior to iodine therapy, every patient’s TSH level, was
recorded. Five days after discontinuity of MMI, iodine therapy started
(patients received no medication for 5 days, because antithyroid drugs
reduce iodine uptake by the thyroid, and it is necessary to stop taking drugs
before initiating iodine therapy). The patients were studied for a maximum
period of consecutive months. Every two months, the patients’ TSH levels in
their bloodstream were measured and recorded. Also, the variables of age
(year), gender (male or female), smoking status (smoker or non-smoker), and
the thyroid gland size 0 (small, no palpable or visible goiter), 1 (moderate, a
mass in the neck that is consistent with an enlarged thyroid which is palpable
but not visible when the neck is in the normal position) and 2 (large, a
swelling in the neck that is visible when the neck is in the normal position
Determining the Effective Factors on TSH Level … 481
and is consistent with an enlarged thyroid when the neck is palpated)) [17],
the administered dose of MMI (the number of tablets 5 mg), and the
administered dose of I131 (mCi) were recorded for every patient. Then the
marginal model of GEE used exchangeable correlation structure. The MEM
model only with the random effect of subject variable and fixed effect time
used the Gamma distribution, and the log link function. Both of the models
were fit to the data for investigation of the main effects of age, gender,
smoking status, thyroid gland size, administered dose of MMI, administered
dose of I131, TSH level after completion of consuming MMI (baseline
TSH), time and the interaction of time and dose of the I131 on TSH level
variations in the patients’ bloodstreams under treatment with iodine therapy
(TSH level). The statistical analysis was performed by using R3.5.2
software. The level of significance was considered 0.05 for all the tests.
Marginal model by generalized estimating equations and mixed effects
model
GEE method focuses on the average response variation of the population
over time, and on the effects of the predictive variables. In this method,
needless of assuming a specific parametric distribution for the response
variable, the mean structures and their variance/covariance can be separately
and without any limitation determined. The response variable’s mean can be
fitted into a link function as a linear function of the determined predictive
variables [18-20].
A MEM can be considered as a multi-level or hierarchical model,
in which the 1st level observations (everyone’s observations) are nested
in the 2nd level observations (different individuals). MEM investigates
individually the trend of response variations, and considers the heterogeneity
of individuals by adding into model the random effects of the predictive
variables in the analysis. Contrary to the method of GEE, in this model, there
should be considered a proper parametric distribution for each response
variable [20, 21].
Atefeh Malekhoseiny et al. 482
Ethical considerations
In this study, all ethical regulations have been observed. Participation in
this study had no financial burden on patients and patients’ private and
personal information had been kept confidential.
Results
The mean TSH level before MMI administration and at the time of
diagnosis was 0.024 ± 0.018. Table 1 shows the descriptive statistics of the
patients’ demographic and clinical specifications. As it can be seen, a
majority of the patients were non-smoker women. TSH level Pearson
correlation coefficients in the three times of measurement are shown in
Figure 1. Baseline TSH correlated significantly with TSH level 2 and 4
months after iodine therapy ( ).002.0,0001.0 == PP Also, TSH level 2
months after iodine therapy was significantly correlated with TSH level 4
and 6 months after iodine therapy ( ).009.0,0001.0=P
Table 1. The descriptive statistics of studied patients’ demographic and
clinical specifications
Variable Class No. Percentage Mean Standard deviation
Male 77 38.3 Gender
Female 124 61.7 - -
Non-smoker 176 87.6 Smoking status
Smoker 25 12.4 - -
0 21 10.4
1 46 22.9 Thyroid gland size
2 134 66.7
- -
Age (year) - - - 37.54 13.66
Dose of MMI - - - 4.44 1.24
Dose of I131 - - - 11.07 4.03
Baseline TSH - - - 3.21 7.47
Determining the Effective Factors on TSH Level … 483
Figure 1. The observed Pearson correlation coefficient of TSH levels during
different measurement intervals for patients with GD (P-value).
Mean TSH variation over time is shown in Figure 2 and TSH variation
over time for every patient is shown in Figure 3. As it can be seen from these
figures, TSH is increasing with time. The absence of some patients in the
fourth month for visit and exit, some other patients after this time of the
study created the gap in Figure 3.
Figure 2. Mean TSH level variation over time for patients with GD.
Atefeh Malekhoseiny et al. 484
Figure 3. TSH level variation over time for any patient with GD.
The results of fitting the marginal model by the method of GEE are
presented in Table 2, and those of fitting MEM are presented in Table 3.
Table 2. The results of fitting GEE model on TSH level of patients with GD
Variables Coefficient estimate Standard error P-value
Age(year) -0.003 0.01 0.710
Male - - - Gender
Female -0.26 0.30 0.387
Non-smoker - - - Smoking status
Smoker 0.60 0.43 0.165
0 - - -
1 0.19 0.35 0.589 Thyroid gland size
2 -0.12 0.39 0.760
Baseline TSH 0.05 0.01 0.0001
Dose of MMI 0.10 0.07 0.172
Dose of I131 0.05 0.06 0.405
Time 0.71 0.15 0.0001
Time∗ dose of I131 -0.03 0.01 0.030
QIC 1837.376
Determining the Effective Factors on TSH Level … 485
Table 3. The results of fitting MEM on TSH level of patients with GD
Variables Coefficient estimate Standard error P-value
Age (year) -0.007 0.01 0.467
Male - - - Gender
Female -0.10 0.32 0.770
Non-smoker - - - Smoking status
Smoker 0.66 0.45 0.244
0 - - -
1 -0.05 0.43 0.911 Thyroid gland size
2 -0.19 0.43 0.675
Baseline TSH 0.05 0.01 0.005
Dose of MMI 0.20 0.10 0.049
Dose of I131 0.10 0.06 0.127
Time 0.66 0.15 0.0001
Time∗ dose of I131 -0.05 0.01 0.0001
QIC 1970.476
The results of the method of GEE showed that there was a significant
relationship between TSH level, and baseline TSH, so that the patients who
had 1-unit higher baseline TSH, their TSH level mean were higher (1.05
times) after treatment with I131 ( ).0001.0=P If the patients received the
average dose of I131, then TSH level increased 2.07 times during every 2
months ( ).030.0=P
The results of fitting MEM showed that there was a significant
relationship between TSH level, and baseline TSH, so that patients who have
1-unit higher baseline TSH l, had higher (1.05 times) TSH level after
treatment with iodine therapy ( ).005.0=P In the patients who had received
1 more MMI tablet, TSH level was 1.22 times greater than other patients
( ).049.0=P If the patients received the average dose of I131, then TSH
level increased 1.19 times during every 2 months ( ).001.0=P
Discussion
In this study, the effects of baseline TSH, and the interaction of time
and the dose of I131 were found significant in both of the marginal and
mixed models. Also, the regression coefficients calculated were similar in
these models. These results indicated that after the initiation of iodine
Atefeh Malekhoseiny et al. 486
therapy with an average dose of I131 for each patient, TSH level in the
patients’ bloodstreams over time would increase. Also, the standard errors of
estimated coefficients were small and not different in these models.
According to the results of mixed effects model, the increase in the dose
of administered MMI for each patient will lead to the increase of TSH level.
This result confirms that MMI restrains the production of thyroid hormones,
as a result, and as a feedback there will be an increase of TSH.
Baek et al. in their study, examined 105 patients with major depressive
disorder over 3 months, and used linear mixed model to identify the
relationship between TSH level in the patients’ bloodstreams and
neurotropic factor brain-derived [22]. Abdi et al. conducted a study on the
individuals with natural performance of thyroid gland over 9 years, and they
used GEE to study the relationship between the performance of thyroid
gland, and the level of blood pressure in the adult population [23]. Ying et
al. discussed the linear regression methods in the analysis of ocular data in
order to determine the related factors of purview among the adolescent. The
results showed that in those studies that if the information of patients’ both
eyes is inserted; the ignorance of observations regression can lead to false
results. In such cases, GEE, and MEM show the utmost statistical power and
accuracy [24]. Strickland et al., in a study, used three methods: MEM, GEE,
and quadratic inference functions in the analysis of cluster data including the
amounts of genome and epigenome. In this study, all the three methods
showed acceptable statistical power; but, in the marginal model, diagonal
correction seemed necessary [25]. To investigate the relationship between
Stargardt’s disease characteristics, and visual ability, Kong et al. investigated
176 patients from Europe and the U.S.A. In this study, they used GEE
analytic method to estimate the temporary relationships, and they used MEM
analytic method to estimate the amount of visual ability loss over time [26].
Naseri et al., in a study, used GEE method to analyze menorrhagia
data. They also compared it with repeated measure-ANOVA, and MEM.
According to the results of this study, GEE, and MEM, having many
Determining the Effective Factors on TSH Level … 487
advantages, are very considerable for the analysis of repeated data. GEE,
especially is very appropriate to analyze the overall effects [21]. In the
present study, 62% of the patients with Graves’s disease were women. The
patients’ average age was 37.5 ± 13.7, which is in accordance to the study by
Younis [27]. But, in the study by Cheah et al., the patients’ average age with
Graves’s disease was reported 46.1 ± 1.5 [28]. Also, in the present study,
it has been shown that age, gender, and thyroid gland severity have no
significant effect on the patient’s response to treatment with radioactive
iodine, which is consistent with the study of Sun [29]. According to the
results of MEM, consuming Methimazole drug has a positive effect on the
increase of TSH level in the patients’ bloodstreams, which is consistent with
the study results by Younis [27]. Sheehan and Doi, in a study on the patients
with Graves’s disease who were under treatment with radioactive iodine,
examined the different types of hypothyroidism which occur after iodine
therapy in each patient [30]. Laurberg et al., in a study, investigated the
effect of common therapy of hyperthyroidism from Graves’s disease on
TSH-receptor autoimmunity. According to the results of this study, in all
the therapeutic methods, majority of the patients experienced remission of
TSH-receptor autoimmunity. But, in the treatment with radioactive iodine
remission of TSH-receptor autoimmunity was less prevalent [31].
Conclusions
Both the methods of GEE and MEM are very useful for the analysis of
the observations from iodine therapy in the patients with hyperthyroidism
over a given time period. In fact, in the longitudinal studies, added to
statistical inferences about the population mean over a given time by using
the method of GEE, some inferences also can be made about the paths of
response variations on the level of each individual by the help from MEM.
By this study, the effects of baseline TSH, the interaction of time and the
dose of I131, and dose of administered MMI were found significant. These
effects will increase the TSH in the patient’s bloodstream under treatment
with I131.
Atefeh Malekhoseiny et al. 488
Acknowledgement
The present article is extracted from the Master’s thesis. It is
financially supported by Arak University of medical sciences, coded
IR.ARAKMU.REC.1396.184. There is no conflict of interest with other
organizations or individuals. We thank everyone who provided insight and
expertise that greatly assisted the research.
References
[1] M. A. Walter, M. Briel, M. Christ-Crain, S. J. Bonnema, J. Connell, D. S. Cooper,
H. C. Bucher, J. Müller-Brand and B Müller, Effects of antithyroid drugs on
radioiodine treatment: systematic review and meta-analysis of randomised
controlled trials, BMJ 334(7592) (2007), 514.
[2] H. B. Burch and D. S. Cooper, Management of Graves disease: a review, JAMA
314(23) (2015), 2544-2554.
[3] M. I. Chiamolera and F. E. Wondisford, Thyrotropin-releasing hormone and the
thyroid hormone feedback mechanism, Endocrinology 150(3) (2009), 1091-1096.
[4] S. Melmed, Williams Textbook of Endocrinology, Elsevier Health Sciences, 2016.
[5] A. G. Unnikrishnan and U. V. Menon, Thyroid disorders in India: an
epidemiological perspective, Indian Journal of Endocrinology and Metabolism
15(Suppl2) (2011), S78-S81.
[6] P. N. Taylor, D. Albrecht, A. Scholz, G. Gutierrez-Buey, J. H. Lazarus, C. M.
Dayan and O. E. Okosieme, Global epidemiology of hyperthyroidism and
hypothyroidism, Nature Reviews Endocrinology 14 (2018), 301-316.
[7] A. Aminorroaya, R. Meamar, M. Amini, A. Feizi, A. Tabatabae and E. F. Imani,
Incidence of thyroid dysfunction in an Iranian adult population: the predictor role
of thyroid autoantibodies: results from a prospective population-based cohort
study, European Journal of Medical Research 22(1) (2017), 21.
[8] D. S. Ross et al., American thyroid association guidelines for diagnosis and
management of hyperthyroidism and other causes of thyrotoxicosis, Thyroid.
26(10) (2016), 1343-1421.
[9] B. Pandiyan, S. J. Merrill, F. Di Bari, A. Antonelli and S. Benvenga, A patient-
specific treatment model for Graves’ hyperthyroidism, Theoretical Biology and
Medical Modelling 15(1) (2018), 1.
Determining the Effective Factors on TSH Level … 489
[10] Y. A. Onimode, D. M. Dairo and A. Ellmann, Pattern of presentation of Graves’
disease and response to radioiodine therapy in South African men, The Pan
African Medical Journal 29 (2018), 48.
[11] C. Regalbuto, I. Marturano, A. Condorelli, A. Latina and V. Pezzino,
Radiometabolic treatment of hyperthyroidism with a calculated dose of 131-
iodine: results of one-year follow-up, Journal of Endocrinological Investigation
32(2) (2009), 134-138.
[12] Z. Geraili-Afra, A. Abadi, J. Yazdani-Charati, S. A. Gooraji, M. Zarghami and
S. Saadat, Comparison of efficiency GEE and QIF methods for predicting factors
affecting on bipolar I disorder under complete-case in a longitudinal studies, Acta
Informatica Medica 26(2) (2018), 111.
[13] Y. Ma, M. Mazumdar and S. G. Memtsoudis, Beyond repeated measures
ANOVA: advanced statistical methods for the analysis of longitudinal data in
anesthesia research, Regional Anesthesia and Pain Medicine 37(1) (2012), 99.
[14] E. Demidenko, Mixed Models: Theory and Applications with R, John Wiley and
Sons, 2013.
[15] P. Diggle, P. J. Diggle, P. Heagerty, P. J. Heagerty, K.-Y. Liang and S. Zeger,
Analysis of Longitudinal Data, Oxford University Press, 2002.
[16] S. L. Zeger and K.-Y. Liang, Longitudinal data analysis for discrete and
continuous outcomes, Biometrics 42 (1986), 121-130.
[17] W. L. Tay, C. L. Chng, C. S. Tien, K. S. Loke, W. W. Lam, S. M. Fook-Chong
and A. K. Tong, High thyroid stimulating receptor antibody titre and large goitre
size at first-time radioactive iodine treatment are associated with treatment failure
in Graves’ disease, Ann. Acad. Med. Singapore 48 (2019), 181-187.
[18] G. M. Fitzmaurice, N. M. Laird and J. H. Ware, Applied Longitudinal Analysis,
John Wiley and Sons, 2012.
[19] P. McCullagh and J. A. Nelder, Generalized Linear Models, CRC Press, 1989.
[20] L. Wu, Mixed Effects Models for Complex Data, Chapman and Hall/CRC, 2009.
[21] P. Naseri, H. A. Majd, N. Kariman and A. Sourtiji, Comparison of generalized
estimating equations (GEE), mixed effects models (MEM) and repeated measures
ANOVA in analysis of menorrhagia data, Journal of Paramedical Sciences
7(1) (2016), 32-40.
[22] J. H. Baek, E.-S. Kang, M. Fava, D. Mischoulon, A. A. Nierenberg, D. Lee, J.-Y.
Heo and H. J. Jeon, Thyroid stimulating hormone and serum, plasma, and platelet
brain-derived neurotrophic factor during a 3-month follow-up in patients with
major depressive disorder, Journal of Affective Disorders 169 (2014), 112-117.
Atefeh Malekhoseiny et al. 490
[23] H. Abdi, S. Gharibzadeh, E. Tasdighi, A. Amouzegar, L. Mehran, M. Tohidi and
F. Azizi, Associations between thyroid and blood pressure in euthyroid adults:
a 9-year longitudinal study, Hormone and Metabolic Research 50(3) (2018),
236-241.
[24] G.-S. Ying, M. G. Maguire, R. Glynn and B. Rosner, Tutorial on biostatistics:
linear regression analysis of continuous correlated eye data, Ophthalmic
Epidemiology 24(2) (2017), 130-140.
[25] J. C. Strickland, I.-C. Chen, C. Wang and D. W. Fardo, Longitudinal data methods
for evaluating genome-by-epigenome interactions in families, BMC Genetics
19(1) (2018), 82.
[26] X. Kong, R. W. Strauss, M. Michaelides, A. V. Cideciyan, J.-A. Sahel, B. Muñoz,
S. West and H. P. N. Scholl, Visual acuity loss and associated risk factors in the
retrospective progression of Stargardt disease study (ProgStar Report No. 2),
Ophthalmology 123(9) (2016), 1887-1897.
[27] J. Younis, The effect of anti thyroid medications on the therapeutic outcome of
I-131 in hyperthyroid patients with Graves’ disease, Egyptian J. Nucl. Med.
14(1) (2017), 28-39.
[28] S. Cheah, K. Aljenaee, N. Muhammad, L. Gavigan and M. Dooley, Outcomes
following fixed dose radioactive iodine therapy (RAI) in hyperthyroid patients
with Grave’s disease and toxic nodular disease, Endocrinol. Metab. Int. J.
3(6) (2016), 174-176.
[29] Q. Sun, S. Li, Z. Wu, J. Liu, H. Liu and K. Lu, The analysis of the affecting
factors on clinical outcome of early-onset hypothyroidism after the patients with
hyperthyroidism received 131I therapy, Journal of Nuclear Medicine 59(1) (2018),
239.
[30] M. T. Sheehan and S. A. Doi, Transient hypothyroidism after radioiodine
for Graves’ disease: challenges in interpreting thyroid function tests, Clinical
Medicine and Research 14 (2016), 40-45.
[31] P. Laurberg, G. Wallin, L. Tallstedt, M. Abraham-Nordling, G. Lundell and O.
Tørring, TSH-receptor autoimmunity in Graves’ disease after therapy with anti-
thyroid drugs, surgery, or radioiodine: a 5-year prospective randomized study,
European Journal of Endocrinology 158(1) (2008), 69-75.