Development and preliminary validation of the fertility status awareness tool: FertiSTAT

Abstract

BACKGROUND

The aim of our research was to create a fertility status awareness tool (FertiSTAT) that would enable women to gain personalized guidance about reducing risks to their fertility and seeking timely fertility medical advice based on their own lifestyle and reproductive profile.

METHODS

Independent risk factors associated with female fertility impairment were identified. Associations between risk indicator and fertility status were examined in 1073 women who completed the Fertility Risk Factors Survey (FRFS) online or in pregnancy termination, antenatal or infertility clinics in the UK, consisting of the FertiSTAT indicators; 49.58% (n = 532) were currently pregnant (78.82% ≥12 weeks pregnant) and 15.66% (n = 168) were currently infertile (trying to conceive >12 or 6 months if >34 years of age).

RESULTS

Twenty-two risk factors were identified from the literature review and expert Delphi consultation. Prevalence of risk factors in the validation sample was similar to general population. Most risks were independently associated with fertility status in logistic regressions and in the expected direction. Discriminant analysis demonstrated that the set of FertiSTAT indicators could correctly classify whether women were currently pregnant or infertile [χ²(19) = 204.209, P < 0.001] with a correct classification rate for the overall sample of 85.8% (326/380), 91.0% (n = 243/267) for the currently pregnant and 73.5% (n = 83/113) for the currently infertile.

CONCLUSIONS

The main result was the generation of a self-administered, multifactorial tool that can enable women to get personalized fertility guidance. This research and the FertiSTAT provide foundational work for public health campaigns to increase awareness about fertility health.

risk factors, infertility, risk correlates, raising awareness, fertility potential

Introduction

Research suggests that people do not behave optimally when it comes to their fertility. First, negative lifestyle habits (e.g. obesity, illicit drug use, alcohol consumption) or indicators of such [e.g. sexually transmitted diseases (STDs)] have increased markedly over the past decade especially in young people (WHO, 2006; Health Protection Agency, 2007). Second, people do not seek help in a timely fashion when they have difficulty conceiving. In one study, almost 20% of women currently trying to get pregnant naturally met the criteria for infertility, yet despite a strong desire for children had never sought any medical advice (Bunting and Boivin, 2007). One reason for these findings may be that people do not know that their behaviour (poor health habits, not getting medical advice) puts them at a disadvantage when it comes to their fertility. Further, although people may have some basic knowledge of risk (e.g. smoking is bad for health therefore bad for fertility), they may lack the precise details of critical thresholds of when behaviour becomes a problem. For example, they know age is associated with reduced fertility but do not know the precise age at which fertility begins to decline, making it difficult for them to modify their behaviour and safeguard their fertility (Bunting and Boivin, 2008). A fertility awareness tool that provides risks and critical thresholds could reduce uncertainty about risk factors for reduced fertility. In 2001, the American Society for Reproductive Medicine (ASRM) tried to raise fertility awareness in an advertising campaign about certain risks (e.g. older age, smoking) that had modest effects (Robert W. Rebar, 2008, personal communication), most likely because it was a general campaign. Health research shows that people are much more likely to reduce risk behaviour (or take appropriate action) if they are provided with personalized information rather than exposed to general campaigns. So-called tailored communications (Rimer and Glassman, 1999) are materials that allow the provision of individually relevant and appropriate information based on knowledge about that person (e.g. via completed symptom checklists) (see, for example, cardiovascular risk calculators: Wilson et al., 1998). These tools are most effective (and less alarming) if coupled with guidance about what to do to reduce risk or seek help (Soames, 1988). We propose that if women had access to a personal tool for fertility then it would (i) allow them to make informed decisions about current lifestyle and reproductive behaviour, (ii) potentially help them take action to safeguard future fertility where risk exists (e.g. quit smoking) and (iii) motivate them to seek timely medical advice (if desired) when clear symptoms of disease (e.g. absence of period) are, possibly unknowingly, undermining current attempts to conceive. To our knowledge, a validated self-administered, multifactorial tool that enables women to get fertility guidance based on their own lifestyle and reproductive profile does not yet exist.

The research presented in this paper consists of the first two phases of the FertiSTAT research programme. The aim of Phase I was to generate a list of independent risk indicators for reduced fertility from the empirical literature, which would be consistent with indicators used in clinical practice. The aim of Phase II was to provide additional univariate and multivariate evidence of the validity of the set of risk indicators as correlates of female fertility potential. The final tool and its guidance (FertiSTAT) are presented in Appendix.

Phase I: identification of risk factors

There are a number of personal characteristics that could indicate a woman's fertility potential and that have clinical relevance, for example, age, lifestyle factors (e.g. smoking) and/or a history of reproductive disease (or other medical diseases) (Hassan and Killick, 2004; Khadem and Mazlouman 2004; Axmon et al., 2006). Fertility potential (or lack thereof) can be measured in a number of ways, e.g. presence of a pregnancy or delay in conception. In order to generate a list of indicators that could be used in the FertiSTAT, we conducted a literature search to identify female fertility indicators, presenting the results to 25 experts in fertility and reproductive health, who discussed each, confirmed their relevance to female fertility and assisted in the generation of guidance that would be appropriate for women, irrespective of whether they were currently trying to get pregnant.

Materials and methods

Search strategy and selection criteria

A full review of the results from the literature search will be presented in a separate paper, but is summarized here. First, a comprehensive review was conducted using PubMed and specific reproductive health references and guidelines [e.g. National Institute for Clinical Excellence (NICE), WHO] to identify risk indicators. Studies were included if their results described a significant association between a risk factor and at least one reproductive outcome [e.g. time to pregnancy (TTP); short (<21 days) or long (>35 days) menstrual cycles and/or sporadic or unpredictable periods; tubal factor infertility]. A risk factor was considered for inclusion in the FertiSTAT if a statistical test and/or confidence interval (CI) linking the risk factor to a reproductive outcome was significant in at least one study. Risk factors that did not achieve statistical significance in the report were not considered for inclusion. Studies that concerned effects on treatment outcome were excluded because associations with risk factors could be affected by treatment. Initially, we aimed to have a tool that was suitable for men, women and couples; however, preliminary work showed that the evidence base for men was less substantial and not presenting a consistent picture of risk. Therefore, the search mainly focused on female factors, but risks identified in the preliminary male work were also presented to fertility experts for opinion.

Second, the set of empirically identified risk factors were presented to a panel of experts in reproductive health with the critical thresholds used in the original study. The critical threshold referred to the level of exposure that produced a significant impact on female fertility potential (e.g. exact number of cigarettes smoked daily or units of alcohol consumption per week). Medical experts were asked to confirm the relevance of each fertility indicator for reduced fertility as per clinical practice and examine the preliminary work on potential male risk indicators and wording of guidance via discussion similar to that of the Delphi method. The expert panel consisted of 10 medical doctors, 1 psychologist, 1 social worker and 8 leaders of patient advocacy groups (see Supplementary Table S1 for a full list of experts), all of whom were experts in reproductive health. The experts were members of the Assisted Conception Taskforce (ACT), a non-profit organization for people with fertility problems. The discussion about risk factors took place at their annual meeting (December, 2006). L.B. carried out the comprehensive literature review but did not attend this meeting. J.B. attended the meeting to answer queries, clarifying aspects of the research and contributing to the discussion where relevant.

Results

Identification of risk factors from literature

In total, 58 studies were reviewed (46 original articles and 1 review paper that contained 12 studies), of which 45 (78%) were retrospective in design and 13 (22%) prospective. Twenty-four studies used infertility as an outcome, 23 TTP, 7 reduced conception rate, 2 menstrual irregularities and 2 studies concerned specific diagnoses [e.g. risk of pelvic inflammatory disease (PID)].

In total, 31 risk factors were identified from the literature review and these were grouped into the following categories: demographic (n = 3 factors), reproductive (n = 7), lifestyle (n = 9) and medical (n = 12) (Table I). From this original list, the experts confirmed 14 as being mainly independent risk factors for reduced fertility as per clinical practice. These factors are identified in bold and with an asterisk in Table I. We included time trying to conceive because this variable was in clinical guidelines and is associated with reduced conception in women ≤34 years (trying >12 months) or >34 years (trying >6 months) (NICE, 2004). A total of 20 risk indicators from the literature review were included in the FertiSTAT.

Table I

Risk factors associated with reduced female fertility impairment extracted from the literature review and (*) included in the FertiSTAT.

Factor	Critical threshold
Demographic factors
*Age	>34 years
Ethnicity
Occupational and environmental exposure
Reproductive factorsa
*Endometriosis
*Menstrual cycle irregularities (<21 days*, >35 days*, amenorrhoea*, oligomenorrhoea*)
*Pelvic inflammatory disease
Polycystic ovary syndrome
*Sexually transmitted diseases (e.g. Chlamydia, gonorrhoea)
*Severe menstrual pain (i.e. dysmenorrhoea)
*Pelvic surgery
Lifestyle factors
*Alcohol	>14 units per weekb
*Caffeine	≥7 units per dayc
Contraceptive agents
Exercise
Coitus
*[Class A] illegal drug use (e.g. cocaine, heroin, anabolic steroid use*)*
*Smoking (tobacco*, marijuana*)	>10 cigarettes per day (tobacco) and >4 times per week (marijuana)
*Stress*	Stress one cannot cope with
*Weight (underweight body mass index [BMI] < 19, overweight BMI > 25*)	BMI > 25
*Unprotected sexual intercourse with multiple partners
Non-reproductive medical factors
Cancer and cancer therapies (e.g. chemotherapy, radiotherapy)
Coeliac
Diabetes insipidus/diabetes mellitus
Epilepsy
Heart diseases
Kidney diseases
Systemic lupus erythematosus
Anti-depressive agents/anti-depressants
Anti-inflammatory agents
Asthma
Anaemia, sickle cell
Thrombophilia/deep vein thrombosis

^aTime trying to conceive also included in FertiSTAT based on clinical guidelines.

^bOne unit of alcohol is 10 ml or 8 g of pure ethanol and refers to around half a pint of larger, beer or cider, 125 ml glass of wine or a 25 ml measure of spirit.

^cSeven units of caffeine per day equate to ≈700 mg of caffeine; 1 unit refers to a cup of coffee, 1/2 unit equals a cup of tea or a can of caffeinated soft drink.

*Confirmed by experts as independent factor associated with reduced fertility as per clinical practice.

There were four reasons why medical experts excluded factors in the final FertiSTAT. First, it was decided that questions about the quality of the menstrual cycle (presence, frequency, duration) would be more informative than questions about the causes of dysfunction that may or may not produce effects in individual cases. Five factors were excluded for this reason: exercise, underweight (BMI <19), ethnicity, polycystic ovarian syndrome and epilepsy. Second, some factors (i.e. asthma medication, occupational/environmental factors, contraception use and prescribed drug use) were eliminated because the evidence was perceived to be too weak or inconsistent after rigorous discussions by all experts. The experts decided that the evidence supporting a link between moderate to large amounts of alcohol consumption and reduced female fertility was sufficient, despite inconsistencies in the literature (e.g. Juhl et al., 2003 versus Axmon et al., 2006) and this factor was included in the final list. Third, medical experts decided that for fertility awareness, the emphasis should be on factors impacting on conception rather than ability to carry a pregnancy to term and three factors (i.e. heart disease, celiac disease, thrombophilia) were excluded because their primary effect is an increased risk of miscarriage, ectopic pregnancy, genetic abnormalities and/or perinatal risks (Molteni et al., 1990; Sher and Mayberry, 1994; Buchholz and Thaler, 2003). Finally, all the non-reproductive medical diseases not already excluded were removed [i.e. sickle-cell anaemia, systemic lupus erythematosus, cancer, diabetes, kidney disease and transplantation) because it would be impractical for a tool with the aims proposed to have an exhaustive list of relatively rare diseases and because the detrimental impact of conditions/treatment on fertility would most likely be conveyed to the individual through information provided in specialist clinics and consenting procedures.

The experts were actively encouraged to suggest other factors they considered important, that might not have been picked up in the literature searches. No new female risk factors were identified through this process, but the experts recommended the inclusion of a section in the tool for women to indicate the presence of male risk indicators (i.e. mumps after puberty, undescended testicles) to augment knowledge they would derive from the FertiSTAT if they had a partner. Therefore, the total number of risk indicators included in the FertiSTAT was 22 (20 female and 2 male).

Development of guidance

To make the tool personally relevant, guidance specific to the type of risk present was generated. It was recognized that not all risks were weighted equally in relation to reduced fertility; that some risks could be eliminated (smoking), whereas others could not (having ever used class A drugs); that the action (guidance) required (behaviour change, medical consultation) depended on the risk category (e.g. weight problem versus amenorrhea); and that action would also depend on whether one was currently trying to get pregnant or not (e.g. take action for endometriosis but only if trying to get pregnant). We assume for all indicators that people would seek help if they were experiencing significant pain or physical distress.

Several possible ways of providing guidance were discussed and tested based on these considerations, with the main ones being a numbering system to generate a total fertility score like the cardiovascular risk calculator and Quetelet matrix like the body mass index. These options were inappropriate due to multiplicative effects (e.g. age × time trying) and too cumbersome to use without additional guidance. The first draft of guidance used colour-coded indicators according to impact on fertility and type of action required based on guidance according to NICE guidelines (NICE, 2004).

The draft was reviewed by four medical fertility experts from ACT with exact wording finalized after pilot testing with 15 women of different ages and in different phases of the reproductive life cycle.

The layout of the final FertiSTAT was in two sections (see Appendix). Section 1 comprises the list of risk indicators for reduced fertility colour coded according to the type of risk. Instructions on the FertiSTAT ask women to tick all the risk indicators that apply to them. Section 2 of the FertiSTAT is the set of matched colour-coded guidance to inform women of what to do to safeguard their fertility. There are four categories of guidance (blue, yellow, orange and red) and each could be relevant to a participant. The guidance contained in the ‘blue’ category applies only to women who have been trying to conceive for <12 months (or <6 months if >34 years) and who have not ticked any other risk indicator. The guidance here specifies that women have not ticked any risk indicator, but that they should continue to monitor the situation regularly because fertility declines with age. Guidance in the ‘yellow’ category applies to anyone who ticked a negative lifestyle factor (e.g. smoking, alcohol consumption) and it specifies that the person should consider modifying their health habits because these risks reduce fertility. Guidance in the ‘orange’ category applies to anyone who has ticked a reproductive factor (e.g. severe period pain, endometriosis) that one might want to discuss with a medical doctor and the guidance specifies that they should do, especially if they are currently trying to conceive. Finally, the ‘red’ category represents those factors that one would most definitely need to see a doctor about if trying to conceive (e.g. absence of periods, class A drug use) and that is the guidance provided. For example, a woman may tick off that she has severe period pain (orange risk) and that she smokes more than 10 cigarettes a day (yellow risk). Therefore, her guidance would be that she should stop smoking because her lifestyle is compromising her fertility (yellow guidance) and she should consider seeking medical attention when she starts trying to conceive as this factor (severe period pain) may be important to her fertility and may require further action (orange guidance).

FertiSTAT is aimed at women only but includes instructions for male risks in a separate section. If women tick that their partner has either had mumps after puberty or undescended testicles then they are advised that he needs to go and see his doctor for further investigation about his situation when they start trying to get pregnant. Women who tick that their partner engages in any of the lifestyle factors (except weight) are advised to follow the same guidance as for women.

Phase II: validity of fertility indicators

The 22 factors identified in Phase I were shown to have significant association with at least one aspect of fertility potential in empirical work and were deemed to have clinical relevance by a panel of fertility experts. In this sense, they are valid indicators of fertility. However, few of the empirical studies tested indicators in multifactorial analyses of risk (note for exceptions Augood et al., 1998; Jensen et al., 1998; Greenlee et al., 2003; Hassan and Killick, 2004; Khadem and Mazlouman, 2004; Axmon et al., 2006) on single reproductive outcomes and none on as many as 22 indicators. Therefore, an important step in the development of FertiSTAT was to replicate association between risk and fertility status at an individual level and also to demonstrate that the 22 indicators as a group were significant in their association with a single reproductive outcome.

In Phase II, the reproductive outcome was the current fertility status because the FertiSTAT risk indicators should discriminate between those of proven fertility (i.e. currently pregnant) and those meeting the clinical criteria for referral for infertility diagnostic investigation (i.e. currently infertile). The validation sample completed the Fertility Risk Factors Survey (FRFS) that comprised the 22 FertiSTAT indicators as well as demographic and medical characteristics. Associations between these items and current fertility status were examined using univariate (logistic regression) and multivariate (discriminant) analyses. We also examined whether the wording of items/critical thresholds (e.g. consuming >14 units of alcohol per week) would lead to prevalence rates for risk indicators that were in line with population values.

Materials and methods

Participants

During an 8-month period, 1073 women completed the FRFS. The sample was pooled from two waves of data collection using (i) an online version of the study survey (four websites and the Cardiff University electronic notice board, n = 603) and (ii) a paper version administered to consecutive patients attending fertility, antenatal or abortion clinics (n = 470) (participation rates in clinics about 35%). Only women who were of reproductive age (18–44) were recruited. Women were assigned to the pregnant group if they were currently pregnant and to the currently infertile if they had been trying to conceive for >12 months (or 6 months if women >34 years of age) (NICE, 2004).

Table II shows demographic and recruitment information for the total sample. On average women were 29.6 (SD = 5.8) years of age, with the majority educated to the university level and from the UK.

Table II

Demographic and recruitment information for total sample (N = 1073).

	Total	%
Country of origina
UK	730	77.00
USA	128	13.50
Canada	43	4.54
Australia	18	1.90
Other	29	3.06
Highest educational levelb
University	386	48.37
Post-secondary/college	285	35.71
Secondary	119	14.91
Primary	8	1.00
Age (SD)c	29.6 (5.8)
Age range
18–25	250	24.20
26–30	349	33.79
31–34	219	21.20
35–39	155	15.00
40–44	60	5.81
Recruitment source
Online (n = 603)
Askbaby	172	16.03
Myspace	115	10.72
Facebook	158	14.73
Verity	26	2.42
University	132	12.30
Clinic (n = 470)
Antenatal	326	30.38
Fertility	103	9.60
Termination	41	3.82

^aOwing to missing data, n = 948.

^bOwing to missing data, n = 816.

^cOwing to missing data, n = 1033.

Materials

The FRFS was developed for this study and comprised of questions derived from the 22 specific risks identified in Phase I and their subtypes (e.g. ‘I am a smoker who regularly smokes 10 or more cigarettes a day’). The response scale for all indicators was either ‘yes’ for the presence of the factor (coded 1) or ‘no’ for the absence of the factor (coded 0). Where relevant we also asked more specific details about the risk to allow comparisons with population values (e.g. total consumption, type of drug used, type of STD etc.). Wording was adapted for the recruitment method and target sample. For the pregnant women, all questions were presented in the past tense asking them to recall their lifestyle habits and reproductive history prior to their current pregnancy. Three additional questions were included to ascertain whether the participant was currently trying to get pregnant/pregnancy was planned, educational status and country of origin.

Procedure

Websites and groups on social networking sites aimed at women just starting out in the process of trying to get pregnant and those aimed at women already pregnant were contacted via e-mail to ask whether they would post the FRFS on their site. Three websites (Facebook, Askbaby.com and pregnancy groups on Myspace.com) and the Cardiff university-wide electronic notice board agreed to post the link. For facebook.com, the study was promoted through their advertisement scheme, whereby adverts pop-up by the side of individual users homepage targeted to women of reproductive age (18–44). The online FRFS was developed using SurveyTracker (Survey Tracker for Windows, Training Technologies Inc., Cincinnati, OH, USA, 2007). In addition, women were recruited from two antenatal clinics (all women ≥12 weeks pregnant), one pregnancy termination clinic and one infertility clinic. These participants completed a paper version of the FRFS. Paper surveys were returned anonymously in collection boxes in the clinic. The Ethics Committee of Cardiff University and the South Wales Ethics Research Committee provided ethical approval for the study.

Data analysis

A minimum of 1008 participants was required to detect low frequency events (e.g. drug use, calculated using G*Power computer program; Faul and Erdfelder, 1992). Preliminary data screening excluded two participants due to extreme outliers for the variable months trying to conceive.

Three types of analyses were carried out (see Fig. 1 for breakdown of sample according to analysis type). First, prevalence of the risk indicators in the total sample [N = 1073, see (1) in Fig. 1] was compared with population values to demonstrate the validity of the wording and critical thresholds used in FertiSTAT items. Absolute ratio between study sample and population data is presented. Second, logistic regressions were conducted to determine univariate association between risk factor and fertility status (coded 1 for currently pregnant and coded 0 for currently infertile). For these analyses, we excluded women who were not currently trying to get pregnant (n = 325) who did not respond to this item (n = 14) or women who were trying to conceive but had not yet met the definition for infertility (n = 34) because we could not assign them to either fertility status group. The odds ratio (±95% CI) is presented for each risk factor for the remaining sample [n = 700, see (2) in Fig. 1]. Finally, discriminant analysis was conducted to determine whether the set of 22 indicators was associated with fertility status. The analysis computes a linear composite of all risk indicators and the degree of correlation between the composite and the grouping factor (in this case, fertility status, pregnant or infertile) determined via several statistics. The canonical correlation shows the degree of association between the linear composite and fertility status, the group centroids show mean scores for each group on the linear composite and the classification table shows whether the linear composite could correctly classify people into their group (pregnant, infertile). As this was a multivariate analysis, only women who had data for all risk indicators (including time trying to conceive) and for whom fertility status could be ascertained were included in the analysis [n = 380, see (3) in Fig. 1]. The difference in sample size is mainly because many pregnant women who were not trying to conceive when they got pregnant omitted a reply to ‘time trying to conceive’. In the total sample of pregnant women (n = 532), 63% stated that they were not trying to get pregnant which is consistent with epidemiological population data on starting families (Ray et al., 2004; Lakha and Glasier, 2006; Mohllajee et al., 2007). See Fig. 1 for other reasons. The item anabolic steroid use was not included in the logistic regression or discriminant analysis due to low frequency (n = 5) in the sample. Male risk indicators (mumps after puberty and undescended testicles) were not included in the analysis due to low frequencies (0.9% for mumps and 1.6% for undescended testicles). All analyses were performed with the software Statistical Package for the Social Sciences (SPSS). For all analyses, a value of P < 0.05 was regarded as statistically significant.

Figure 1

Flowchart detailing participate breakdown according to analysis type. Sample for (1) descriptive statistics on prevalence, (2) logistic regression on univariate risk associations and (3) discriminant analysis on multivariate associations. Single asterisk indicates infertile referred to trying >12 months or >6 months if >34 years of age. Double asterisks indicate missing items referred mainly to the variables time to pregnancy and illegal drug use. No other item accounted for more than 5% of the missing data. TTP = time to pregnancy.

Results

Frequency of risk indicators compared with population values

Table III shows frequencies of risk indicators in total study sample (N = 1073) and population (and their ratio). The sample frequencies were similar to those reported in the population values but the sample reported a higher frequency for menstrual irregularities, menstrual cycle <21 days and alcohol consumption per week (any amount) compared with the population. However, the latter reported more excessive alcohol consumption (e.g. >14 units a week; Department of Health, 1992), and more marijuana use than the study sample.

Table III

Frequency of risk factors in the sample (N = 1073) and the population.

Factors	Population (%) (a)	Sample (%) (b)	Ratio score (a/b)	Population source
Demographic
Education (university level)	31.20	48.37	0.65	Office for National Statistics (2008)
Reproductive
Period pains	46.83	32.92	1.42	Zondervan et al. (1998, p. 95, Table I)
Endometriosis	6.00–10.00	5.48	1.55	Giudice and Kao (2004)
Pelvic inflammatory disease	2.00	2.19	0.91	NHS Choices website (1 in 50 women per year develops the disease)
Menstrual cycle less than 21 days	3.20	8.54	0.37	WHO (1983)
Menstrual cycle more than 35 days	8.05	13.19	0.61	Harlow and Ephross (1995)
Menstrual cycle irregular	30.00	34.03	0.88	Harlow and Ephross (1995)
Period	3.10	5.84	0.53	Harlow and Ephross (1995)
Pelvic surgery	—	11.891	—	—
Sexually transmitted disease	12.60	11.57	1.09	Fenton et al. (2001)
Lifestyle
Overweight	33.00	23.40	1.41	Office of National Statistics (2001)
Unprotected sexual intercourse with multiple partners	32.002	23.96	1.34	Fontes and Roach (2007). Percentage based on those reporting having had up to five sexual partners
Stress	11.00	16.12	0.68	Office of National Statistics (2006)
Class A drug ever used	10.00	13.43	0.74	Roe and Man (2006, p.24, Table 4.6)
Last 12 months	2.10	3.96	0.53	Roe and Man (2006, p. 24, Table 4.6)
Anabolic steroid	0.602	0.85	0.71	Roe and Man (2006, p. 45, TableA2.1)
Alcohol	56.50	69.253	0.82	Goddard (2006, p. 63, Table II.3)
≥14 units a week	23.50	10.00	2.35	Goddard (2006, p. 62, Table II.2)
Smoke	26.67	23.583	1.13	Goddard (2006, p. 15, Table I.1)
Caffeine	97.102	91.593	1.06	Heatherley et al. (2006)
Marijuana	9.702	4.563	2.13	Roe and Man (2006, p. 45, Table A2.1)

¹No population data could be obtained for comparison.

²Percentage includes men.

³Number based on participants reporting of any consumption of the variable.

Univariate analysis of risk factors

Table IV shows the risk indicators significantly associated with current infertility (univariate analysis) which were age, period pain, endometriosis, PID, a long menstrual cycle (>35 days), an irregular menstrual cycle, not having a period, previous pelvic surgery, being overweight, having unprotected sexual intercourse and stress. A prior STD, use of class A drugs and caffeine consumption were in the predicted direction but the CI included unity. Risk factors significantly associated with being in the pregnant group were short menstrual cycle (<21 days), smoking more than 10 cigarettes a day and use of marijuana. The odds for consuming >14 units of alcohol per week were in the opposite direction for pregnancy than predicted but the CI included unity.

Table IV

Frequencies and odds ratios (95% CI) for association between FertiSTAT risk factor and fertility status in univariate analysis (n = 700^†).^a

FertiSTAT risk factors	Currently pregnant (N = 532)	Currently infertile (N = 168)	Odds ratio	95% CIb
Time to pregnancy [mean (SD)]c	9.16 (18.47)d	48.14 (40.61)
Time to pregnancy range (months)	0–144	14–204
Reduced chance of pregnancy
Age [mean (SD)]	29.7 (5.7)	30.8 (5.4)	0.94***	0.91, 0.97
	n (%)	n (%)
Period pains	147 (27.95)	62 (37.13)	0.66*	0.46, 0.95
Endometriosis	12 (2.31)	17 (10.56)	0.20***	0.09, 0.43
Pelvic inflammatory disease (PID)	6 (1.15)	7 (4.19)	0.27**	0.09, 0.80
Menstrual cycle >35 days	66 (13.07)	32 (19.63)	0.62*	0.39, 0.98
Menstrual cycle irregular	156 (29.89)	70 (42.17)	0.59**	0.41, 0.84
Amenorrhoea	23 (4.47)	15 (9.32)	0.46*	0.23, 0.90
Pelvic surgery	30 (5.75)	36 (21.95)	0.22***	0.13, 0.37
Overweight (>13 kg/28 pounds/2 stone)	89 (17.45)	57 (34.34)	0.39***	0.26, 0.57
Stress one cannot cope with	63 (12.40)	32 (19.51)	0.58*	0.37, 0.93
Unprotected sexual intercourse with multiple partners	66 (12.67)	53 (37.95)	0.31***	0.20, 0.47
No effect	n (%)	n (%)
Class A drug	38 (9.07)	18 (10.71)	0.83	0.46, 1.50
Caffeine (≥7 units per day)	39 (7.44)	13 (7.78)	0.95	0.5, 1.83
Sexually transmitted disease (STD)	54 (10.27)	21 (12.73)	0.79	0.50, 1.34
Alcohol (>14 units per week)	55 (10.48)	9 (5.45)	2.03	0.98, 4.20
Increased chance of pregnancy	n (%)	n (%)
Menstrual cycle <21 days	55 (10.68)	5 (2.99)	3.87**	1.52, 9.84
Smoke (>10 cigarettes per day)	92 (17.66)	13 (7.88)	2.51**	1.36, 4.61
Marijuana (>4 times a week)	24 (4.68)	1 (0.61)	8.05*	1.08, 59.96
Anabolic steroide	4 (0.76)	1 (0.50)

^†n varies between variables. Range for pregnant sample was 505–526. Range for infertile sample was 161–166.

^aDV = 0 (infertile) and 1 (pregnant).

^bCI, confidence interval.

^cFor not pregnant, time to pregnancy = months trying to conceive.

^dTime to pregnancy only available for 267 pregnant women.

^eAnabolic steroid was excluded from univariate and multivariate analyses due to low frequency.

*P < 0.05.

**P < 0.01.

***P < 0.001.

Unexpectedly, the results of the logistic regressions showed that higher levels of negative lifestyle habits (alcohol, smoking tobacco and smoking marijuana) were associated with increased odds of being in the pregnant group. To assess whether this association was due to women modifying their behaviour when trying to conceive, possibly resulting in the impression of relatively less smoking and drinking in women who are infertile, the frequencies of negative lifestyle habits were assessed in both groups. Table V shows successively lower frequencies on the negative health habits from the general population, study participants and unplanned pregnancies followed by those with planned pregnancies and/or currently trying to get pregnant. Therefore, to test for the expected association between these negative lifestyle factors and fertility, we examined TTP instead, and these results were in the expected direction. Table VI shows that those who took longest to conceive (in the pregnant group) reported more smoking (cigarettes, marijuana) and drinking than did those who conceived relatively more quickly.

Table V

Percentage of woman engaging in lifestyle risk for women according to intention to conceive compared with population values.

Factors	Population (Table III)	Unplanned pregnancy (n = 335)	Planned pregnancy or trying to conceive (n = 399)1
Alcohol (>14 units/week)	23.50	10.7	9.52
Smoke (smokers in reproductive age)	26.67	18.5	12.28
Marijuana (frequent use)	9.70	3.9	3.26

¹Planned pregnancy = 197 and trying to conceive = 202 (168 infertile women and 34 women currently trying <12 months or <6 months if >34 years).

Table VI

Percentage of woman engaging in lifestyle risk factors according to time trying to conceive (n = 197 women with planned pregnancy).

Factors	Time to pregnancy
Fertility category	<12 months (n = 138)	>12 months (n = 59)
	n (%)	n (%)
Alcohol (>14 units per week)	12.10 (8.76)	7.12 (12.07)
Smoke (>10 cigarettes per day)	19.72 (14.29)	11.00 (18.64)
Marijuana (>4 times a week)	5.10 (3.68)	6.00 (10.17)

Discrimination of fertility status groups

The overall discriminant analysis was significant [χ²(19) = 204.21, P< 0 .001, n = 380, eigenvalue 0.74, canonical correlation = 0.65]. The group centroids (means) on the linear function for the set of indicators showed overall significantly less risk for those in the currently pregnant group (centroid = −0.56) than in the currently infertile group (centroid = 1.32). Using all FertiSTAT items, the correct classification rate for the overall sample was 85.8% (326/380), 91.0% (n = 243/267) in the currently pregnant subgroup and 73.5% (n = 83/113) in the currently infertile group. As the infertile group is partly confounded with months infertile, we re-computed the discriminant analysis excluding the variable years trying to conceive (TTP in pregnant group). The discriminant analysis remained significant [χ²(18) = 125.08, P< 0 .001, n = 446, eigenvalue 0.33, canonical correlation = 0.50] and the overall correct classification was 76.2%. The correct classification rate for the currently pregnant subgroup was 79.6% and for the currently infertile subgroup 66.4%, which was significantly above chance levels (χ²(1) = 6.58, P = 0.01). Sample size for this final analysis includes an additional 66 women who were previously omitted because they did not provide TTP data.

Discussion

The main finding of the current study is the generation of a self-administered, multifactorial tool that can enable women to get fertility guidance based on their own lifestyle and reproductive profile. FertiSTAT comprises 22 indicators of reduced fertility extracted from the empirical literature, confirmed as relevant to fertility potential by a panel of experts in fertility health and validated in univariate and multivariate analyses. The FertiSTAT is a tailored communication tool that provides guidance to help women make informed decisions about their lifestyle and/or seek timely medical advice when needed (and if desired).

The results replicate past research in demonstrating support for risk factors associated with reduced fertility, and as a group, these proved informative of fertility status. The classification rate overall and for subgroups (currently pregnant or infertile) was statistically above chance (with or without TTP) and it was achieved from a self-administered tool without input from medical test results, knowledge of parity, male fertility status or other fertility variables and was not dissimilar to the range reported for other fertility tests. For example, Verhagen et al. (2008) did a comprehensive review of current ovarian prediction tests/models with the best test (antral follicle count) having a prediction of rate 88% and 61% for good and bad ovarian response, respectively. In addition, these indicators are about fertility potential at a specific point in time and not about the capacity to conceive. The latter is often what people want to know about (or want to safeguard) and what is reflected in FertiSTAT.

Other online fertility awareness tools exist but are not as comprehensive in their coverage of risk as the FertiSTAT. They do not take into account specific critical thresholds for risks (number of cigarettes smoked), their differing importance/weight in predicting fertility (smoking versus amenorrhea) or the multiplicative relationships that occur (age, years infertile), which together lead to different guidance. Importantly, the FertiSTAT risk indicators have been validated, personalized guidance developed in line with clinical recommendations and wording tested to be clear and understandable to most women likely to use the tool. Third, the multifactorial nature of the FertiSTAT means that overall fertility status would be more reliably assessed than would be the case with, for example, fertility tests, carried out in a single point in time (e.g. basal body temperature, ovulation detection kit) or time trying to conceive, in isolation because the value of single indicators can often be compromised by other unmeasured problems. Finally, many online tools are sponsored by industry or private fertility clinics and guidance may not have been developed without commercial motive. Developmental work continues on the FertiSTAT but results thus far suggest it to be a promising addition to existing fertility awareness tools.

The FertiSTAT risk indicators showed the expected pattern of association with current fertility status except for drinking alcohol and smoking (tobacco and marijuana) which showed an intriguing association that we believe is worth pursuing in future research. Specifically, these showed an increased chance of pregnancy, whereas they are typically associated with a lack of pregnancy (Hassan and Killick, 2004). We can only speculate about the cause of this unexpected result. One possibility is that pregnant women in this sample were risk takers in general (consumed more alcohol, smoked more) and this risk taking extended to their sexual life (indeed, 63% of women had unplanned pregnancy). However, this explanation seems unlikely because risk seeking was not mirrored in the rest of their profile (e.g. unprotected sexual intercourse, higher incidence of STDs) and the level of smoking and drinking in pregnant women, though higher than the infertile group, was still lower than in the general population. A more plausible explanation might be that the currently infertile women were risk averse. Women were assigned to the infertile group if they had been trying to conceive for >12 months (or 6 months if >34 years of age) and it is possible that a lack of success over this extended period led women to actively try to modify lifestyle habits known to affect health to improve their chances of pregnancy (e.g. smoking, drinking: Bunting and Boivin, 2008). Secondary analyses supported this contention in showing an association between risk exposure (i.e. smoking and drinking) and intention to conceive as well as TTP (in the expected direction). In order to adequately test this hypothesis, one would have to conduct a prospective study to follow women from the moment they start trying to conceive, to see whether lifestyle habits do change over time and, if so, at what point this change occurs.

The overall aim of our research programme was to develop a tool that would enable women to safeguard their fertility and/or seek timely medical advice (where required and desired). The tool is therefore appropriate for all women whether or not they are trying to conceive. There are some negative aspects to raising fertility awareness, e.g. causing unnecessary fear, which we feel has been minimized by providing the guidance to help people address fear, as recommended by health practitioners. However, there is also concern that people may be given a false sense of security. However, FertiSTAT guidance was designed to minimize the chance of women having the impression that they are ‘just fine’. Women who do not check any risks are informed that they should continue to monitor their situation (and alerted to the effect of age on fertility) rather than told they do not have a problem. Importantly, as soon as risk-free women have been trying for >12 months (or 6 months for older women), their FertiSTAT profile would change and the new guidance would recommend they consult a medical doctor. This recommendation is in keeping with clinical practice and the fact that 25–30% of women have unexplained infertility (Evers, 2002; Smith et al., 2003) with the only risk being time trying to conceive.

The next phase of this project is to investigate the predictive utility of FertiSTAT in prospective research with women currently trying to get pregnant. Cross-sectional designs make it impossible to disentangle cause and effect, and we cannot know whether risk indicators caused women's current fertility status or whether causation was reversed and current fertility status caused profile risk, or both. A further objective is to examine the predictive ability of FertiSTAT on behavioural intentions. If FertiSTAT does as we intend, then women should be optimizing fertility behaviour in line with their own personal goals of eventual parenthood, that is, they should, for example, smoke less and be timelier about seeking medical advice when they suspect a problem (and desire medical care). This so-called ‘nudging in the right direction’ (Thaler and Sunstein, 2003) and, more generally, the FertiSTAT tool raises questions about the ethics (and economics) of presymptomatic fertility monitoring and individual autonomy/freedom that need to be deliberated.

It is worth considering potential sources of bias in the present study. First, although risk indicators were derived from the empirical literature, the final set relied on the opinions of the 10 medical fertility experts. Experts were not randomly selected and their views not necessarily representative of their profession. However, experts were asked to achieve a consensus in their decision-making; therefore, we can at least be confident that factors were not reflective of idiosyncratic judgements. Second, the recruitment method was successful (N > 1000) but not all women could be included in analyses because we could not assign them to a fertility status or data were missing/non-applicable on one or more indicators. Future studies should take into account this ratio in recruitment in order to maximize the sample size in analyses.

Third, the recruitment methods employed (e.g. Internet) have a number of methodological strengths and limitations. Methodological research comparing the Internet versus traditional data collection methods shows high consistency between methods (Lieberman, 2008) but access often leads to a bias towards higher socioeconomic users (Weissman et al., 2000). In the present study, we additionally recruited from medical clinics (abortion, antenatal, infertility) to broaden representation and achieve heterogeneity in some sample characteristics (age, time trying to conceive). This bias due to our methods should not affect the nature of association between risk and fertility status, but may have affected prevalence.

A final issue arising from the current research programme is the emphasis on female fertility. The goal of this research programme is to raise awareness about fertility health issues and this requires more attention to be devoted to male fertility. Current evidence for men is, according to the medical experts, probably too inconsistent (except for mumps and undescended testicles) to develop a tool at this time. The present study did include a limited number of male risk indicators but their incidence was too low to include in the analyses possibly because almost 20% of women stated they did not know if their partner had had either risk. Nevertheless, the FertiSTAT does contain two risk indicators for male partners and associated guidance. Future research needs to focus much more attention on developing a valid tool for men. In all likelihood, the FertiSTAT will not be one tool but a collection of tools that reflect female, male and joint contributions to fertility health to better raise awareness of fertility health issues.

This research comes at time when the importance of fertility health has been prioritized on the public health and social policy agendas of the European Union (Evans, 2007). Our research programme proposes that the future of fertility health care should be centred on providing people with information leading to informed choice about all aspects of their own fertility, in essence promoting fertility health literacy. The present study and FertiSTAT provide the foundational work for public health campaigns to increase awareness about fertility health.

Funding

This work was developed during a PhD CASE studentship (awarded to L.B.) funded by the Biotechnology and Biological Sciences Research Council (BBSRC) with Merck Serono S.A. as the industrial partner.

Acknowledgement

We thank all the members of the Assisted Conception Taskforce for their contribution to the developmental phase of the project. For data collection, we thank Hannah Maundrell at Askbaby.com, Janet Evans (IVF Wales), Mary James and Susan Jose (Cardiff and Vale Antenatal Unit), Caroline Scherf and Carolyn Alport (Cardiff and Vale Gynaecology Unit) and the staff at each clinic. For technical development and support, we thank Lorraine Woods and Andrew Thomas.

Appendix