Women undergoing assisted fertilisation and high-intensity interval training: a pilot randomised controlled trial

Participants and design

This was a randomised controlled trial on HIT compared with standard procedures prior to assisted fertilisation, undertaken at the Norwegian University of Science and Technology (NTNU) and St. Olavs Hospital, in Trondheim, Norway. The study protocol is previously published.20 The study is registered in ClinicalTrial.gov (NCT01933633).

We made the following changes to the protocol after trial initiation: (1) from February 2016, participants allocated to the control group no longer waited for 10 weeks before the assisted fertilisation treatment started, as we experienced that participants declined to participate if their standard treatment could be delayed. (2) From February 2016, we also opened the study to patients from a private fertility clinic in Trondheim (Spiren Fertility Clinic) to increase the recruitment. Changes to the original study protocol were approved by Regional Committee for Medical and Health Research Ethics.

After recruitment and screening, participants received written information and signed an informed consent before entering the study. Inclusion criteria were age>18 years, BMI>25 kg/m², accepted for assisted fertilisation, willing to come for study assessments and attend supervised exercise sessions for 10 weeks. Exclusion criteria were high-intensity exercise >2 times per week, current or previous Metformin use (with a washout period of >4 weeks), physical impairments limiting exercise and unwilling to delay fertility treatment for 10 weeks. Participants randomised to the HIT group received a free membership at the local gym during the intervention period, and participants in the control group received a gift from the same local gym worth US$85.

After baseline assessments, the participants were stratified for polycystic ovary syndrome and randomly allocated (1:1) to the intervention group or the control group (as previously described).20

The fertility treatment, blood pressure measurements, blood sampling and hyperinsulinemic euglycaemic clamp assessments were done blinded for group allocation. The investigators were not blinded for group-allocation on measurements of VO₂peak, body composition and height or in intervention administration.

Intervention

The exercise programme consisted of 3 weekly HIT sessions for 10 weeks, as previously described.20 Two of the weekly sessions were 4×4 min HIT at 85%–95% of individual heart rate maximum (HR_max) in the work-bouts. The third weekly exercise session was 10×1 min HIT with maximum intensity in the work-bouts. All participants in the HIT group attended supervised sessions until they were familiar with the HIT protocols. All participants wore heart rate monitors (Polar RCX3, Polar, Oulu, Finland) and documented their exercise sessions in a training diary. After 10 weeks, the participants followed standard care assisted fertilisation. Participants in the HIT group were encouraged to continue with the exercise during the fertility treatment, until ovulation induction. Participants in the control group received regular advice from the hospital staff about physical activity (usual care) and were not discouraged from being physically active. None of the groups received any nutrition advice, but were encouraged to adhere to the current Norwegian diet recommendations.

Fertility treatment

The fertility treatment included either a short or a long ovarian stimulation protocol individualised for each participant according to standardised routines at the Fertility Unit, St. Olavs Hospital (as described previously).20

Measurements

All participants underwent the same assessments at baseline and after the 10 weeks intervention period.

The primary outcome was ongoing pregnancy, defined as the sonographic evidence of intrauterine gestational sac and fetal heart activity at week 7 to 8 of gestation. No active intervention took place after the initial cycle after randomisation, and we only included results from the first cycle in our trial. Secondary outcomes were insulin sensitivity, reproduction-related hormones, lipids, VO₂peak, body weight, body composition, blood pressure and resting heart rate. In addition, we recorded physical activity and diet at baseline and after 10 weeks.

We obtained blood samples in the morning after ≥10 hours overnight fast, followed by a hyperinsulinemic euglycaemic clamp according to a modified version of the method originally described by De Fronzo et al 21 and as previously reported.20 Steady-state glucose infusion rate (M-value) for the last 30 min of the test was calculated and expressed as mg glucose per body surface area (m2 per minute, and the M-value during this period represents the whole-body glucose disposal rate. The insulin sensitivity index (SI_Clamp) was calculated as M/(G × ∆I), where M is the glucose infusion rate (mg/min), G is steady-state blood glucose concentration (mg/dL) and ∆I is the difference between fasting insulin and the last plasma insulin concentration in the insulin-stimulated steady state (μU/mL).22 23 Serum insulin was analysed on ELISA (IBL International) using a DS2 ELISA processing system, Dynex Technologies, Chantilly, USA. We also calculated homeostasis model assessment of insulin resistance (HOMA2-IR) as (glucose × insulin) /22.5.24

Concentrations of lipids, glucose, haemoglobin, glycated haemoglobin (HbA1c), albumin, high-sensitive C-reactive protein were measured using Advia Chemistry XPT, Siemens, Erlangen, Germany. Insulin c-peptide was measured using DPC Immulite 2000, Siemens, Erlangen, Germany. The hormone assays included luteinising hormone, follicle-stimulating hormone, prolactin, sex hormone-binding globulin (SHBG) and thyroid-stimulating hormone, analysed on Advia Centaur XPT, Siemens, Erlangen, Germany. Antimüllarian hormone was analysed on Cobas 8000, Roche, Basel, Switzerland. Testosterone was analysed on an Agilent 1290 with 6410 Triple Quad LC/MS-Ms detector, Agilent, Santa Clara, USA. Free androgen index was calculated as 100 × (total testosterone/SHBG). We also collected and stored (at −80°C) blood and urine in the Regional Biobank 1 of Central Norway, with the data solution BioByte. The biobank is approved by the Data Inspectorate of Norway and by the Regional Committee for Medical Research Ethics.

Cardiorespiratory fitness was measured as peak oxygen uptake (VO₂peak) on a treadmill (Woodway USA, Waukesha, Wisconsin, USA) using an incremental test to exhaustion. Expired gasses were analysed using direct ergospirometry with a mixing chamber (Oxygen Pro, Erich Jaeger GmbH, Hoechberg, Germany). VO₂peak was calculated as the average of the three highest consecutive 10 s measurements.

Body weight and body composition were measured using bioelectrical impedance analysis (InBody 720, Biospace, Seoul, Korea). Waist-and-hip circumference was measured to the nearest 0.5 cm using a metric tape with the participants in standing position and at normal expiration.

We measured systolic and diastolic blood pressure on the right arm after the participants had rested in the supine position for 15 min using an automatic blood pressure monitor (IntelliVue MP50, Philips House, Dublin, Ireland). The average of three measurements, with 2 min intervals between, was calculated.

Physical activity was registered by questionnaires and by activity monitors (SenseWear, BodyMedia, Pittsburgh, Pennsylvania, USA) for 5 days (3 weekdays and 2 weekend days) at baseline and after 10 weeks. An electronic standardised food frequency registration system25 was used to register diet for 3 days (2 weekdays and 1 weekend day).

Sample size

The power calculation was estimated from a pilot study26 giving advice about exercise and diet to improve pregnancy rates in overweight and obese participants. We assumed an increase in clinical pregnancy rate from 0.30 to 0.55 during the first treatment cycle in the intervention group as clinically relevant. With a statistical power of 0.8 and a 5% level of significance, it was estimated that we needed 61 participants in each group.20 To allow for an expected 15% dropout, we aimed to include 140 participants in the study.

Statistical methods

All available data were used in both time points. Baseline data were tested for normality. We assumed no systematic differences between groups at baseline due to the randomisation model; however, because of a smaller sample size than expected, differences between groups at baseline were tested using independent sample t-tests and Fisher’s exact tests. To test the between-group difference in ongoing pregnancy rates, we used Fisher’s exact tests. For the remaining outcome variables, the effects of intervention were analysed using mixed linear models for continuous outcomes. The effect of time and group allocation was set as fixed effect with the levels: baseline, exercise postintervention and control postintervention. Participant ID was set as a random effect to account for repeated measurements. We performed all statistical analysis using IMB SPSS Statistics 22. Baseline data are given as mean±SD or as number of participants (percentage). Comparisons between groups are reported as estimated means with 95% CIs or as number of participants (percentage). We considered a p value less than 0.05 as statistically significant.

Pregnancy rate

There was no difference in pregnancy rate between the HIT group and the control group (p=0.6) (table 2). One participant in the control group interrupted the treatment.

Secondary outcomes

Insulin sensitivity, measured as glucose infusion rate (M-value, mg/min/m²) improved significantly after HIT by 23% (within-group difference 60.6, 95% CI 11.4 to 109.8, p=0.02), with no change in the control group (between-group difference, p=0.04) (table 3 and figure 2). Three participants in the control group versus two participants in the HIT group were insulin-resistant at baseline according to a cut-off value of <6.5 mg/min/fat-free mass.27 After 10 weeks, only one participant in the control group was insulin-resistant. Fasting glucose was significantly decreased in the HIT group (within-group difference 0.2 mmol/L, 95% CI −0.4 to 0.0), however, the between-group difference was not significant. We found no statistically significant changes in other measures of insulin sensitivity or blood markers (table 3).

Request permission

We found no statistically significant changes in body composition between groups, but we observed tendencies for reduced fat mass (p=0.06) and visceral fat mass (p=0.08) in the HIT group compared with the control group (table 3 and figure 2). Visceral fat decreased significantly after HIT (within-group differences −7.5 m², 95% CI –14.9 to -0.1) and so did waist circumference (within-group differences −6.1 cm, 95% CI –10.8 to -1.4).

VO₂peak increased significantly after HIT with 2.6 mL/kg/min (95% CI 0.6 to 4.6), with no between-group differences.

We observed no between-groups differences at baseline in physical activity levels or diet, and no within-group or between-group differences after 10 weeks.

Main findings

We were not able to test our primary hypothesis due to the low number of participants and therefore low statistical power. Nevertheless, the HIT group had a significant improvement in insulin sensitivity, measured by hyperinsulinemic euglycaemic clamp, compared with the control group. This improvement was seen despite no body weight loss. We also found a statistically significant decrease in fasting glucose, visceral fat area and waist circumference as well as an increase in VO₂peak after 10 weeks of HIT.

Pregnancy rate

In our study, we observed a relatively high pregnancy rate in both groups compared with previous reports.13 14 28 29 Since participation was voluntary and the intervention was HIT, we may have recruited a healthier study population. There were conflicting results of previous studies on the effect of exercise on fertility outcomes (without assistance).9 12 15 19 30–32 A recent, well-powered randomised controlled trial of a lifestyle intervention programme among obese infertile women found no effect of a 6-month programme of hypocaloric diet and moderate-intensity physical activity on birth rates of a healthy singleton at term during 24 months follow-up.14 Of note, however, significantly more women in the intervention group conceived naturally.14 Several smaller randomised controlled trials have indicated an effect of either hypocaloric diet,18 19 32 exercise18 19 or the combination of dietary intervention and exercise,19 32 33 on fertility outcomes. There is, however, a lack of good-quality randomised controlled trials comparing types of physical activity, intensity and durations on reproductive outcomes and the focus have been mainly on weight loss.15

Insulin sensitivity and other outcomes

We argue that the significant improvement in insulin sensitivity of 23% in the HIT group is likely to be clinically important for metabolic and reproductive health.6 13 15 34 35 A systematic review on euglycaemic hyperinsulinaemic clamp studies in women with and without polycystic ovary syndrome reported the following thresholds for an improvement in insulin sensitivity: small (3.8%), moderate (12%), large (25%) and very large (46%).36

Insulin sensitivity is closely related to reproductive function2 6 10 and insulin resistance affects ovarian responsiveness to assisted fertilisation.6 30 Furthermore, an improvement in insulin sensitivity is associated with more regular menstrual periods and ovulation after lifestyle interventions including exercise and/or diet.18 30 37 Combining exercise and diet in lifestyle interventions makes it difficult to separate the potential different effects exercise and diet may have on fertility-related outcomes. In addition, such complex interventions have higher dropout rates than isolated exercise training programmes.19 26 We did not observe statistically significant improvements in HbA1c, fastening c-peptide, SI_Clamp or HOMA2-IR, potentially caused by less sensitive measurements and thus lower statistical power.

Strengths and limitations

The strengths of our study are the randomised controlled trial study design, the use of hyperinsulinemic euglycaemic clamp methodology to measure insulin sensitivity, well established methods used for fertility treatment and measures of cardiorespiratory fitness and the well-controlled exercise intervention. Since exercise was the only intervention, we were able to assess the isolated effects of HIT on the reported outcomes. The adherence to the HIT programme was high, indicating that this could be a feasible exercise regime for this population. HIT is time-efficient and reported to be enjoyable for obese inactive women.38 39 Exercise interventions are shown to have a lower dropout rate compared with dietary interventions and better long-term weight maintenance—two important factors regarding adherence to lifestyle interventions.15 30 However, the fact that a high number of eligible women declined to participate in our trial, needs to be further explored.

The main limitation of our trial is the low number of participants. We aimed to include 140 participants in the study, while only 18 women agreed to participate. Hence, the lack of power on our primary outcome analysis represents a major limitation and risk of a statistical type 2 error. Blinding was not possible in all measurements; however, the treating physician who selected type of fertility treatment, and the personnel who performed the clamp and blood measurements, were blinded for group allocation. Furthermore, all baseline measurements were undertaken before randomisation.

Clinical relevance

Although most studies on exercise and fertility-related outcomes seem promising, these exercise-induced improvements are multifactorial and complex.14 15 30 40 There exist no previous clinical trials investigating the effects of HIT as the only intervention prior to assisted fertilisation. Several systematic reviews have found exercise to improve fertility,12 15 however, highlighting the need of more comprehensive clinical trials on the isolated effect of exercise.11 12 15 We argue that our findings regarding improvements in insulin sensitivity, and the positive tendencies to improvements in several of our other secondary outcomes, are of clinical relevance. These potential positive effects could bring significant benefits to the women and to society as an inexpensive and accessible strategy to improve assisted fertilisation success rate among overweight and obese women.