Levothyroxine for a high-normal TSH in unexplained infertility

SUMMARY

Objective:

Unexplained infertility affects nearly 1/3 of infertile couples. Women with unexplained infertility are more likely to have a high-normal thyroid-stimulating hormone level (TSH: 2.5–5 mIU/L) compared to women with severe male factor infertility. Practice guidelines vary on whether treatment should be initiated for TSH levels >2.5mIU/L in women attempting conception because the effects of treating a high-normal TSH level with levothyroxine are not known. We evaluated conception rates and live birth rates in women with unexplained infertility and high-normal TSH levels.

Design, Patients and Measurements

Retrospective study including ninety-six women evaluated for unexplained infertility at a large academic medical center between 1/1/2000–6/30/2017 with high-normal TSH (TSH: 2.5–5 mIU/L and within normal range of the assay) who were prescribed (n=31) or not prescribed (n=65) levothyroxine. Conception and live birth rates were assessed.

Results:

Conception rate in the levothyroxine group was 100% compared to 90% in the untreated group (p=0.086 unadjusted; p<0.05 adjusted for age; p=0.370 adjusted for TSH; p=0.287 adjusted for age and TSH). Live birth rate was lower in the levothyroxine group (63%) compared to the untreated group (84%) (p=0.05 unadjusted; p=0.094 adjusted for age; p=0.035 adjusted for TSH; p=0.057 adjusted for age and TSH).

Conclusion:

Women with unexplained infertility and high-normal TSH levels treated with levothyroxine had a higher rate of conception but lower live birth rate compared to untreated women, with the limitation of a small sample size. These findings assert the need for prospective, randomized studies to determine whether treatment with levothyroxine in women with unexplained infertility and high-normal TSH is beneficial.

INTRODUCTION

Infertility – the inability to achieve a pregnancy after 1 year of regular, unprotected intercourse – afflicts between 7–19% of women between the ages of 15–49 in the US^1,2 and 48 million couples worldwide³. Infertility is associated with significant emotional costs as well as economic costs, with the global cost of infertility approaching 16 billion dollars⁴. Unexplained infertility, or infertility which is the result of no diagnosed cause, affects up to 1/3 of infertile couples^5,6. The most effective treatment for unexplained infertility is in vitro fertilization (IVF), an expensive and invasive form of assisted reproductive technology⁷. Therefore, there is a need to better understand potential etiologies of unexplained infertility, to develop more targeted treatment strategies.

In vitro studies, animal model studies and human studies suggest that thyroid stimulating hormone (TSH) and thyroid hormone are critical factors for fertility^8–12. Although a TSH between 2.5–5 mIU/L may be within the normal reference range, 95% of individuals with no evidence of thyroid autoimmunity and without a personal/family history of thyroid disease have a TSH level ≤ 2.5 mIU/L¹³. In this context, we have previously shown that women with unexplained infertility are nearly twice as likely to have TSH level of 2.5–5 mIU/L as compared to a population of women with infertility due to a male partner with severe oligospermia (sperm count < 1 million)¹⁴. High-normal TSH levels have been similarly associated with adverse pregnancy outcomes¹⁵, reinforcing the notion that values in the upper-half of the conventionally-defined normal range may confer an increased risk of reproductive pathology. Notably, despite these prior findings¹⁴, whether treating women with unexplained infertility with levothyroxine is beneficial is not currently known. This controversy is further reflected in society guidelines, which vary on whether to treat women attempting conception who have a TSH level > 2.5mIU/L^16–19, leading to substantial heterogeneity in clinical practice. Furthermore, a limited body of evidence suggests that levothyroxine treatment in the context of subclinical hypothyroidism may be associated with adverse pregnancy outcomes, including gestational hypertension, in women with a TSH in the high-normal range²⁰.

Taking advantage of the variability in levothyroxine prescribing patterns between providers, we performed an observational study using clinical data to investigate the utility of treatment with levothyroxine in women with unexplained infertility who have a TSH level between 2.5–5 mIU/L. We hypothesized that treatment with levothyroxine would result in improved conception rates, decreased time to conception and improved live birth rates in women with unexplained infertility.

MATERIALS AND METHODS

Study population and Follow-up Assessment

We used the clinical patient registry at a large academic health system (the Research Patient Data Registry of Mass General Brigham, formerly Partners HealthCare) to develop a database of women with unexplained infertility and high-normal TSH. In this regard, we first queried women between the ages of 18 and 39 years who presented to the healthcare system with a diagnosis of infertility but without irregular menstruation between 1 January 2000 and 30 June 2017. Medical records were then manually reviewed to identify women who met inclusion and exclusion criteria. We included women who did not conceive after at least 12 months (even if their infertility evaluation occurred prior to 12 months of infertility) with adequate exposure to sperm and a TSH level of 2.5 mIU/L < TSH < 5 mIU/L and also within the normal range of the TSH assay. All included women also had regular menstrual cycles, defined as a cycle length of 21–35 days with ≤5 days of inter-cycle variability, a prolactin level of ≤ 20 ng/mL, a day-3 follicle stimulating hormone (FSH) level of < 10 IU/mL with an estradiol level of ≤80 pg/mL²¹ if available, and no other known cause of infertility. Male partners had a normal semen analysis, defined as a concentration ≥15 million/mL, motility ≥40%, and normal forms ≥4% (if strict Tygerberg analysis was available) based on the WHO 2010 reference values²². Women with a known history of hypothyroidism or hyperthyroidism were excluded. We also excluded women with a body mass index (BMI) <18.5 kg/m² or ≥40 kg/m², women with a history of three or more consecutive miscarriages, and women with conditions that may affect reproduction and fertility [e.g., endometriosis or polycystic ovary syndrome (PCOS)]. A subset of women in this report were included in a prior study investigating TSH levels in euthyroid women with unexplained infertility¹⁴, but data on conception rate and live birth rate were not previously reported.

Once our database was established, medical records were reviewed to determine whether each patient was started on levothyroxine by a clinical care provider. All women were initiated on levothyroxine after their initial TSH level was measured. Medical records were also examined to determine if conception occurred (defined as a biochemical pregnancy), time to conception, and whether the pregnancy resulted in a live birth.

This study was approved by the Mass General Brigham (formerly Partners HealthCare) institutional review board.

Laboratory assessment

All diagnostic and laboratory testing was performed as part of routine clinical care.

Statistical analysis

The statistical analysis was performed using JMP Pro 14.0 (SAS Institute, Cary, NC) software and R version 4.2.0. We report means and standard deviation measurements and these values were compared using the Student’s t-test for normally distributed data. If the data were non-normally distributed, we report the median and interquartile range and compare the values using the Wilcoxon rank-sum test. For categorical outcomes, percentages were compared using the Fisher exact test. Due to the high rate of conception in our study, Firth bias-reduced logistic regression²³ was used to control for covariates (TSH and age). Time to conception between groups was compared using the log-rank test. A two-sided p-value of < 0.05 was considered statistically significant.

RESULTS

Clinical characteristics

A total of 96 women met our eligibility criteria of unexplained infertility with a TSH level of 2.5–5 mIU/L and within the normal range of the assay. Thirty-one of these women (32%) were started on levothyroxine by a clinical care provider after measurement of the TSH level. Clinical characteristics of the study population are listed in Table 1. Women who were treated with levothyroxine (LT4+) were slightly older than the untreated (LT4-) women (LT4+: 32.3 ± 2.5 years versus LT4-: 30.9 ± 3.1 years, p = 0.02) and had a higher TSH level than untreated women (TSH in LT4+: 3.45 mIU/L [3.07, 4.19] versus TSH in LT4-: 2.95 mIU/L [2.70, 3.38], p=0.0003). Likely due to the fact that current guidelines do not recommend testing for thyroid autoimmunity in euthyroid women with infertility^24,25, data on thyroid antibodies were not available for the majority of the women in the study.

Table 1:

Characteristics of the study population

	Treated with levothyroxine (n=31)	Not treated (n=65)	p-value
Age (years)	32.3 ± 2.5	30.9 ± 3.1	0.02
BMI kg/m 2 *	24.2 [21.7, 28.4]	23.1 [20.9, 26.8]	0.26
Tobacco
% Current use (n)	6.7% (2)	4.6% (3)	0.65
% Past or present use (n)	13% (4)	13.8% (9)	0.99
Age at menarche (years) *	12 [11, 13]	12 [12, 14]	0.18
Duration of infertility (months) *	17.5 [12, 24]	14 [12, 24]	0.95
% Secondary infertility (n)	32.2% (10)	18.5% (12)	0.19
% with prior parity (n)	12.9% (4)	9.2% (6)	0.72
Day 3 FSH (IU/mL)	6.5 ± 1.5 (range: 4.1–9)	6.7 ± 1.6 (range: 2.3–9.6)	0.60
Prolactin (ng/mL)	12.7 ± 4.2	12.2 ± 4.2	0.58
TSH (mIU/L) *	3.45 [3.07, 4.19]	2.95 [2.70, 3.38]	0.0003

Reported as mean ± SD or *median [interquartile range]

Levothyroxine dosage data was available for nine of the treated women. Levothyroxine dose ranged from 25 mcg to 75 mcg daily, with the majority of women being prescribed 50 mcg daily (n=7). Follow-up TSH values measured at least four weeks after initiation for eight of these women ranged from 1.53 mIU/L to 4.6 mIU/L. Four of the eight women had a TSH value < 2.5 mIU/L on follow-up measurement, whereas four had a TSH value between 2.5 mIU/L and 4.6 mIU/L.

Conception rate

Follow-up data were available for 88 of the 96 women (n=29 in the LT4+ group and n=59 in the LT4- group) (Figure 1); Supplementary Table 1 compares characteristics of women for whom conception data were available versus those for whom data were not available. Overall, the conception rate was high in both the LT4+ and LT4- groups. The conception rate in the LT4+ group was 100% (n=29) and the conception rate in the LT4- group was 90% (n=53), p=0.17 with Fisher’s exact test and p=0.086 with Firth logistic regression (odds ratio estimate with 95% confidence interval: 7.17 (0.80, 947.08) (Table 2). As age was slightly higher in the LT4+ group, when we controlled for age, the conception rate was statistically significantly higher in the LT4+ group compared to the LT4- group (p=0.046, Firth logistic regression). As baseline TSH was also higher in the LT4+ group, when we controlled for TSH and both TSH and age, the difference in conception rate was no longer significantly different (p=0.370 and p=0.287, respectively, Firth logistic regression) (Table 2).

Figure 1:

Flow chart illustrating number of women for whom data were available at each stage of the analysis.

Table 2:

Odds ratio estimates (95% confidence intervals) and their p-values for conception in the setting of treatment with levothyroxine (Firth logistic regression analysis; n=88). Model A: unadjusted analysis. Model B: adjustment for age. Model C: adjustment for baseline TSH level. Model D: adjustment for age and baseline TSH level. Women not treated with levothyroxine are the reference group. p < 0.05 indicates significance.

	Model A	Model B	Model C	Model D
Levothyroxine (treated)	7.17 (0.80, 947.08) p=0.086	9.62 (1.03, 1285.0) p=0.046	3.33 (0.31, 454.2) p=0.370	4.09 (0.37, 562.06) p=0.287
Age (years)		0.79 (0.56, 1.06) p=0.116		0.81 (0.58, 1.11) p=0.187
TSH (mIU/L)			15.28 (1.04, 1772.08) p=0.045	9.42 (0.82, 836.51) p=0.08

Data regarding method of conception were available for all the women who conceived except for one in the LT4- group. Five (17%) of the women in the LT4+ group who conceived had a spontaneous conception as compared to 18 of the women (34%) not treated with levothyroxine pre-conception (p=0.13, Fisher’s exact test), with the remainder conceiving via assisted reproductive technologies. Sixteen women (55%) who conceived in the LT4+ group conceived either with the use of IVF or IVF with intra-cytoplasmic sperm injection (ICSI) versus 43% in the untreated group (LT4) (p= 0.36, Fisher’s exact test).

For the women who conceived, median time to conceive (calculated as the number of months from infertility evaluation to first day of last menstrual period before conception) was 9.3 months [2.9, 18.1] in the LT4+ group and 6.2 months [1.6, 10.5] in the LT4- group (p=0.10). For the women who did not conceive (all in the LT4- group), we were able to collect follow-up data (ranging from 1 month-1.6 years after initial fertility evaluation); time to conception was not significantly different between groups using the log-rank test (p=0.23).

Live birth rate

Live birth data was available for 68 of the 82 women who conceived (n=24 in the LT4+ group and n=44 in the LT4- group) (Figure 1); Supplementary Table 2 compares characteristics of women for whom live birth data were available versus those for whom data were not available. Overall, 76% (n=52) of the women who conceived had a live birth. Of the women treated with levothyroxine (LT4+), 63% (n=15) had a live birth compared to 84% (n=37) of the women not treated with levothyroxine (LT4-) [p=0.07 with Fisher’s exact test and p=0.05 with Firth logistic regression, odds ratio estimate with 95% confidence interval: 0.33 (0.10, 0.998)] (Table 3). When we controlled for age (p=0.094), TSH (p=0.035), or both age and TSH (p=0.057), there was either a trend towards or a statistically significant higher rate of live births in the untreated group. Of the women who underwent IVF or IVF with ICSI, there was not a significant difference in live birth rate in those treated with levothyroxine (LT4+: 73% live birth rate) as compared to those not treated with levothyroxine (LT4-: 89%; p=0.37, Fisher’s exact test).

Table 3:

Odds ratio estimates (95% confidence intervals) and their p-values for live birth in the setting of treatment with levothyroxine (Firth logistic regression analysis; n=68). Model A: unadjusted analysis. Model B: adjustment for age. Model C: adjustment for baseline TSH level. Model D: adjustment for age and baseline TSH level. Women not treated with levothyroxine are the reference group. p < 0.05 indicates significance.

	Model A	Model B	Model C	Model D
Levothyroxine (treated)	0.33 (0.10, 0.998) p=0.05	0.38 (0.12, 1.18) p=0.094	0.28 (0.08, 0.91) p=0.035	0.31 (0.09, 1.04) p=0.057
Age (years)		0.87 (0.69, 1.09) p=0.231		0.86 (0.68, 1.07) p=0.185
TSH (mIU/L)			1.44 (0.57, 4.01) p=0.449	1.58 (0.62, 4.40) p=0.344

Of the 52 women who had a live birth, data were available in 48 as to whether the child was born at term or pre-term (defined as gestational age < 37 weeks). A similar percentage of women in the levothyroxine-treated group as compared to the untreated group had pre-term births. A total of 15% (n=2 of 13) in the LT4+ group had a pre-term birth compared to 14% (n=5 of 35) in the untreated group (p=0.99, Fisher’s exact test).

DISCUSSION

Women with unexplained infertility are more likely to have high-normal TSH levels than a comparator group of women whose partner has severe male factor infertility¹⁴. Whether treatment with levothyroxine is beneficial with respect to conception in this population of women with infertility and high-normal TSH levels is not known, but a recent cohort study including over 19,000 women found that 18.7% of reproductive aged women with infertility have a TSH level in the high-normal range (TSH: 2.5–4.5)²⁶ and therefore this question affects a significant proportion of women experiencing infertility. Our data suggest that while there may be a slight benefit with respect to conception, treatment with levothyroxine may be associated with a lower live birth rate in women with unexplained infertility and a high-normal TSH level.

Numerous prior studies have grappled with the question of whether treatment with levothyroxine in women with normal range TSH levels is beneficial to pregnancy outcomes^27,28. This question has been most rigorously studied in euthyroid women with thyroid autoimmunity. Despite data demonstrating that euthyroid women, including those with a history of miscarriages or infertility, who are thyroid antibody positive have higher miscarriage rates²⁹, treatment with levothyroxine during pregnancy does not increase the conception rate or live birth rate in euthyroid TPO-antibody positive women with a history of miscarriage or infertility³⁰ or those with recurrent pregnancy loss³¹ compared to placebo.

A prior retrospective study investigating pregnancy outcomes in pregnant women with subclinical hypothyroidism found that in women with a TSH level between 2.5–4 mIU/L, those treated with levothyroxine had a similar risk of pregnancy loss, but a higher risk of gestational hypertension as compared to untreated women²⁰. Therefore, there was a signal of harm in pregnant women with high-normal TSH levels treated with levothyroxine. Similarly, in a subgroup analysis of the TABLET study, a study in which euthyroid TPO-antibody positive women were randomized to either levothyroxine or placebo, women with a TSH of > 2.5 mIU/L randomized to treatment with levothyroxine were not more likely to have a live birth as compared to those randomized to placebo³⁰. Our data also suggest that while treatment with levothyroxine may lead to a slightly higher conception rate, live birth rate is not higher and in fact may be lower in women treated with levothyroxine. Therefore, the totality of evidence to date suggests that there may be no benefit of treatment with levothyroxine with respect to live birth rates in the population of women with high-normal TSH levels during pregnancy.

We did find a possible benefit with respect to conception rate in women treated with levothyroxine as compared to those not treated. Pre-clinical studies indicate that thyroid hormone may play a role in oocyte development^8,32. In women undergoing in vitro fertilization, women with oocyte fertilization rates of less than 50% had significantly higher TSH levels as compared to those with fertilization rates of greater than or equal to 50%³³. Similarly, a study in women undergoing in vitro fertilization with donor oocytes also supports the concept that thyroid hormone may play an important role through fertilization but not after, as TSH levels ≥ 2.5 in the oocyte donor but not the recipient were found to be associated with a lower rate of biochemical and clinical pregnancy as compared to oocyte donors with a TSH < 2.5¹². Our data demonstrating a slightly increased conception rate with a slightly lower live birth rate support the concept that thyroid hormone may play a significant role through conception but not thereafter.

These data are an important addition to the current literature, as there are varying recommendations regarding the treatment of high-normal TSH levels in women with infertility. Whereas some society guidelines recommend treating women attempting conception who have a TSH level > 2.5 mIU/L, others recommend only initiating treatment in those undergoing in vitro fertilization, whereas other guidelines recommend treatment in women with a TSH > 4 mIU/L^16–19. These recommendations are based on available data and expert opinion that treatment of a high-normal TSH level with levothyroxine is unlikely to result in harm. Our data, provide further support that this question remains unanswered and that treatment with levothyroxine in this population may not be a benign undertaking.

A strength of our study is our rigorous inclusion criteria; we only included women with unexplained infertility to minimize the effects of other causes of infertility, including subtle male factor infertility, on our outcomes. Limitations of this study include the fact that it consists of retrospective clinical data and our carefully phenotyped population resulted in a small sample-size. The non-randomized study design makes it possible that other factors, including socio-economic status and prior gestational outcomes, may be unaccounted for and affecting the outcomes. Our study does highlight the fact that the question of whether women with a high-normal TSH should be treated with levothyroxine remains unanswered, given the significant divergence in practice with some women being treated with levothyroxine and others not. We also do not have data on follow-up TSH levels for the majority of women in the study and therefore are not able to determine if outcomes are associated with changes in TSH. Lastly, we do not have data on thyroid antibodies in the majority of women in this study, which is not surprising given the fact that current guidelines do not recommend testing for thyroid autoimmunity in euthyroid women with infertility^24,25. Further, a recent cohort study including over 19,000 reproductive-aged women with a history of infertility or miscarriage found that 83% of euthyroid women with a TSH level in the high-normal range are TPO antibody negative²⁶, suggesting TPO antibody positivity is not very common in a population of euthyroid women.

In conclusion, our study highlights that 1) there is great diversity in society guideline recommendations and in practice with respect to treating women with unexplained infertility with high-normal TSH levels, with some providers treating euthyroid women with high-normal TSH levels and others not treating them and 2) there may be no benefit to levothyroxine treatment in euthyroid women with TSH levels > 2.5 mIU/L and in fact, treatment may be associated with a lower live birth rate. Therefore, this remains an important question in the field and warrants rigorous prospective study.

Data availability statement:

Deidentified data will be made available upon reasonable request.