Hum. Reprod. Advance Access originally published online on August 4, 2007
Human Reproduction 2007 22(10):2685-2692; doi:10.1093/humrep/dem251
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Identifying subfertile ovulatory women for timely tubal patency testing: a clinical decision rule based on medical history
1 Centre for Reproductive Medicine, Department of Obstetrics and Gynaecology, Academic Medical Centre, Amsterdam, The Netherlands 2 Department of Clinical Epidemiology and Biostatistics, Academic Medical Centre, Amsterdam, The Netherlands 3 Department of Obstetrics and Gynaecology, Máxima Medical Centre, Veldhoven, The Netherlands 4 Department of Obstetrics and Gynaecology, Onze Lieve Vrouwe Gasthuis, Amsterdam, The Netherlands 5 Department of Public Health, Erasmus MC, University Medical Centre, Rotterdam, The Netherlands 6 Department of Obstetrics and Gynaecology, Vrije Universiteit Medical Centre, Amsterdam, The Netherlands
7 Correspondence address. Department of Clinical Epidemiology and Biostatistics, Academic Medical Centre, Amsterdam, The Netherlands. Tel: +31 205667002; Fax: +31 206912683; E-mail: s.f.coppus{at}amc.uva.nl
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionFundingAppendixAcknowledgementsReferences |
|---|
BACKGROUND: The aim of tubal testing is to identify women with bilateral tubal pathology in a timely manner, so they can be treated with IVF or tubal surgery. At present, it is unclear for which women early tubal testing is indicated, and in whom it can be deferred.
METHODS: Data on 3716 women who underwent tubal patency testing as a part of their routine fertility workup were used to relate elements in their medical history to the presence of tubal pathology. With multivariable logistic regression, we constructed two diagnostic models. One in which tubal disease was defined as occlusion and/or severe adhesions of at least one tube, whereas in a second model, tubal disease was defined as the presence of bilateral abnormalities.
RESULTS: Both models discriminated moderately well between women with and women without tubal disease with an area under the receiver-operating characteristic curve (AUC) of 0.65 (95% CI: 0.63–0.68) for any tubal pathology and 0.68 (95% CI: 0.65–0.71) for bilateral tubal pathology, respectively. However, the models could make an almost perfect distinction between women with a high and a low probability of tubal pathology. A decision rule in the form of a simple diagnostic score chart was developed for application of the models in clinical practice.
CONCLUSIONS: In conclusion, the present study provides two easy to use decision rules that can accurately express a woman's probability of (severe) tubal pathology at the couple's first consultation. They could be used to select women for tubal testing more efficiently.
Key words: tubal pathology/medical history/tubal patency testing/clinical decision rule
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionFundingAppendixAcknowledgementsReferences |
|---|
Tubal pathology ranks among the most frequent causes of subfertility, next to ovulatory disorders and sperm defects (Evers, 2002
). Therefore, in British and Dutch guidelines, tubal patency testing is a part of the fertility workup (National Institute of Clinical Excellence, 2004; Nederlandse Vereniging voor Obstetrie en Gynaecologie, 2004). The most frequently used and widely accepted tests for this are hysterosalpingography (HSG) and diagnostic laparoscopy with chromopertubation. Both tests have several drawbacks however. Women consider HSG a painful test and HSG has a considerable risk of infection (Forsey et al., 1990; Liberty et al., 2007). Laparoscopy is an invasive surgical procedure that requires general anaesthesia with associated risks, and its availability is restricted as it is an expensive investigation requiring operating time and dedicated personnel.
For these reasons, tubal testing is usually performed in a selected group of subfertile women. At present, guidelines advocate medical history taking as a tool to select women for tubal testing (National Institute of Clinical Excellence, 2004). It is not entirely clear how elements from the medical history should be combined to identify women that should undergo early tubal patency testing. So far, only a few studies with relatively small sample sizes have tried to summarize this information in a quantitative manner (Hubacher et al., 2004; Coppus et al., 2007).
The aim of the present study was to form clinical decision rules based on medical history that can help to identify women at high and low risk for (severe) tubal disease. These decision rules, presented as simple quantitative score charts, were developed in a large cohort of women who underwent routine tubal patency testing.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionFundingAppendixAcknowledgementsReferences |
|---|
Patients
Between January 2002 and February 2004, consecutive couples presenting at the fertility clinic of 38 centres in The Netherlands were invited to join a prospective cohort study. The local ethics committee of each participating centre gave Institutional Review Board approval for this study. All couples underwent a basic fertility workup according to the guidelines of the Dutch Society of Obstetrics and Gynaecology. The details of this workup have been described in detail elsewhere (Van der Steeg et al., 2007
). The present analysis was limited to couples with a regular ovulatory cycle, defined as a cycle length between 23 and 35 days with a within cycle variation of less than 8 days. Ovulation was detected by a basal body temperature chart, mid-luteal serum progesterone, or by ultrasonographic monitoring of the cycle. Couples with a history of reversal of sterilization, tubal surgery, IVF or previous tubal patency testing were excluded. Couples in whom semen analysis showed a severe impairment of semen quality requiring IVF–ICSI (defined as a post-wash total motile count <1 x 106) were also excluded from the present analysis.
For each couple, we registered referral status (by a general practitioner or gynaecologist), duration of child wish, female and male age, dysmenorrhoea, female and male smoking habits, history of pelvic inflammatory disease (PID), history of Chlamydia infection, previous pelvic surgery other than Caesarean section, previous treatment with intrauterine insemination and previous pregnancies from the female and male partner in the present and/or prior partnerships, including whether these resulted in an ectopic pregnancy, legally induced abortion, spontaneous abortion or term delivery. Moreover, the total motile sperm count (TMC) was recorded.
Duration of child wish was defined as the period between the date the couple had started unprotected intercourse and the date of first consultation. Female and male age was calculated at the time of this first visit. Previous PID had either been confirmed laparoscopically or was documented if a woman had been treated with antibiotics for a period of intense abdominal-pelvic pain and fever, for which no other focus had been found. Infection with Chlamydia trachomatis was self-reported by the patient, and could have been symptomatic or asymptomatic. Missing data in patients' history were imputed, as restricting the analysis to complete cases leads to loss of statistical power and potentially biased results in multivariable analyses (Van der Heijden et al., 2006). For this purpose the ‘aRegImpute’ multiple imputation function in S-plus 2000 (MathSoft Inc.) was used.
Tubal patency testing was performed with either HSG, diagnostic laparoscopy (DLS) or transvaginal hydrolaparoscopy (THL), depending on the local protocol of participating hospitals. In case, tubal patency was evaluated both with HSG and diagnostic laparoscopy or THL, the result of the most invasive test was taken as conclusive. Tubal pathology was defined as uni- or bilaterally impaired or absent flow of contrast medium on HSG, and as uni- or bilateral occlusion and/or hydrosalpinx and/or severe adhesions disturbing ovum pick-up if the tubes were evaluated with laparoscopy or THL. This definition included endometriosis related tubal disease.
Statistical analyses
We developed two multivariable logistic regression models. In the first model, all women with either uni- or bilateral abnormalities were coded as diseased, whereas only those women with bilaterally normal tubes were considered as non-diseased. This first model will be referred to as the ‘any pathology model’. In the second model, only women with bilateral tubal abnormalities were classified as diseased, whereas women with uni- or bilaterally normal tubes were considered to be non-diseased. The latter model will be referred to as the ‘bilateral pathology model’.
For the development of the multivariable models, we first checked the assumption of linearity between the continuous variables female age, male age, TMC and duration of child wish on the one hand and both definitions of tubal pathology on the other hand using smoothed piecewise polynomials (splines) (Harrell et al., 1988). Non-linear associations were redefined, based on these spline functions. Subclasses within categorical variables were dichotomized. We then performed univariable and multivariable logistic regression analysis to estimate odds ratios (ORs), 95% confidence intervals (95% CI) and corresponding P-values for both dichotomous and continuous variables.
As the use of too stringent P-values for variable selection is more deleterious for a model than including too many factors (Steyerberg et al., 2000), all variables that showed a significance level of <30% in univariable analyses were entered in the multivariable logistic regression model. To make parsimonious models, we used a stepwise backwards selection procedure, using a predefined significance level of >10% for removing variables from the models.
To adjust for overfitting, we performed internal validation with bootstrapping (Steyerberg et al., 2001). We generated 500 bootstrap datasets in which the same multivariable logistic regression model was estimated. By analysing the difference between the model based on the original dataset and the bootstrap estimates, a shrinkage factor was calculated, and used to correct the original model coefficients and associated probability estimates for overfit.
The discriminative capacity of the models was evaluated by calculating the area under the receiver-operating curve (AUC). Calibration of the models was assessed by comparing the calculated probabilities across five risk groups (quintiles) with the observed proportion of tubal disease in each of these groups. The goodness-of-fit was evaluated graphically and tested formally with the Hosmer and Lemeshow test statistic.
Finally, a paper score worksheet was developed, for which the shrunken regression coefficients were multiplied by 10 and rounded to simplify the computation for clinical practice. Values for continuous variables were obtained by linear interpolation. This diagnostic rule was then transformed into a score chart to simplify computations in clinical practice. Data were analysed using SPSS 12.0.1 (SPSS Inc., Chicago, IL, USA) and S-PLUS 2000 (MathSoft Inc.).
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionFundingAppendixAcknowledgementsReferences |
|---|
During the study period, data on 7860 couples were collected. Of these, 938 women did not have a regular ovulatory cycle, 636 men had severely impaired sperm quality and 196 couples had previously undergone fertility surgery or IVF. In another 568 women, HSG or DLS had been performed in a previous episode of subfertility. In total, these factors left 5522 couples for inclusion.
In total, 3716 women underwent 2752 hysterosalpingographies, 1527 diagnostic laparoscopies and 160 transvaginal THLs (Fig. 1). About 2039 women underwent only HSG, 808 women underwent only laparoscopy and 147 women underwent only THL. The remaining 722 women underwent multiple tests: 709 women had HSG and laparoscopy, 3 women had HSG and THL, 9 women had THL and laparoscopy and 1 woman underwent all three procedures.
|
In 1806 women (32%), tubal patency was not assessed invasively. Women that were tested were significantly less often referred by a gynaecologist (7.4% versus 10.4%, P <0.001), had a longer duration of subfertility (median 1.5 versus 1.4 years, P < 0.001), were less often treated previously with IUI (3.1% versus 6.7%), had conceived less often previously in the present relationship (30.3% versus 36.5%, P < 0.001) and had more often a history of pelvic surgery (14.2% versus 11.3%, P = 0.008) than women that did not undergo tubal testing.
The median time between the couple's first appointment and tubal patency testing with HSG was 3.1 months. For laparoscopy and for THL, these intervals were 6.5 and 4.9 months, respectively. For all women, the median time from first visit to a definitive diagnosis regarding tubal patency was 4.1 months (5th–95th percentile: 1.1–15.8 months). Severe bilateral tubal pathology was diagnosed in 312 women, corresponding with a prevalence of 8.4%. In 369 women, unilateral pathology was diagnosed, which results in an overall prevalence for combined uni- or bilateral tubal disease of 18%.
The baseline characteristics of the 3716 couples are shown in Table 1. In total, 15% of data points were missing and subsequently imputed. The spline analyses showed that the relation between most continuous variables and the probability of both outcomes was approximately linear. A non-linear relationship was observed for female age, with a marked increase in risk with age over 32 years. We therefore redefined female age as two linear variables, one for female age below and one for age above 32 years.
|
Results of the univariable and multivariable analyses are summarized in Table 2. For any tubal pathology, the multivariable analysis showed that prolonged duration of subfertility, a pregnancy of both partners or of only the male partner in a previous relationship, male smoking habits, a history of PID, history of Chlamydia infection, history of ectopic pregnancy and previous pelvic surgery were all factors that increased the likelihood of tubal pathology. A history of a legally induced abortion also contributed to a higher probability of tubal pathology, although this parameter was not statistically significant at the 5% level (P = 0.07).
|
For bilateral tubal pathology, multivariable analysis showed that referral by a gynaecologist, female age above 32 years, dysmenorrhoea and smoking of the woman increased the likelihood of disease, as did a longer duration of subfertility, a pregnancy of both partners in a previous relationship, male smoking habits, history of PID, history of Chlamydia infection, ectopic pregnancy and previous pelvic surgery. In contrast, a previous pregnancy in the current relationship and a higher male age lowered the chances of bilateral tubal pathology.
The ‘any pathology’ model had a modest overfit of 4%, whereas the ‘bilateral pathology’ model showed 9% overfit. To improve calibration of the probabilities in future patients, the regression coefficients of the models were shrunk with these factors. Both models discriminated modestly between diseased and non-diseased with an AUC of 0.65 (95% CI: 0.63–0.68) for the ‘any pathology’ model and 0.68 (95% CI: 0.65–0.71) for the ‘bilateral pathology’ model, respectively.
The calibration of the model for any tubal pathology is shown in Fig. 2A, and that for bilateral tubal pathology in Fig. 2B. The calculated probabilities of bilateral tubal pathology correlated well with the observed proportions of affected women, as all points were located close to the line of equality (x = y), indicating good calibration of the model. This was confirmed by the Hosmer and Lemeshow test statistic (P = 0.83). Because of the overestimation of the lowest probabilities for women with any tubal pathology, the calibration of this model was less optimal (P = 0.27). However, the women with the highest likelihood of tubal disease could still be distinguished adequately from women with an average probability.
|
A summary score can be calculated based on the respective elements in a couple's medical history. By plotting this score, the probability of tubal pathology can subsequently be read from a graph. The score charts for both models are presented in Fig. 3.
|
The consequences of the use of the clinical decision rules are shown in Table 3. For example, deferring tubal testing in case of a probability of bilateral tubal pathology of less than 5% results in a 25% reduction in the number of invasive tests procedures, whereas 90% of women with bilateral tubal disease are still detected. At this cut-off level, 10 tests have to be performed to detect one case of bilateral tubal disease. Increasing the cut-off to a higher pre-test probability of tubal pathology at which tubal patency testing is performed will result in fewer procedures, but also in a higher number of women for whom tubal testing is incorrectly postponed. By planning tubal testing in women at highest risk for tubal pathology early, the detection rate will increase while delaying testing in the majority of women correctly.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionFundingAppendixAcknowledgementsReferences |
|---|
We developed two clinical decision rules to calculate the probability of any tubal pathology as well as the probability of severe bilateral tubal pathology in subfertile women. These rules were derived from the clinical characteristics of 3716 couples who had been referred for evaluation of their subfertility in a large prospective multicentre study in the Netherlands. Both models could discriminate moderately well between women with and women without tubal pathology. Calibration of the models was good, allowing distinction between women with a high-risk profile and women with a low-risk profile.
The relative importance of discrimination and calibration depends on the clinical application of a model (Mol et al., 2005). As our models are intended to counsel women about the probability that tubal testing would reveal a cause for their subfertility, the accuracy of the numeric probability (calibration) is relevant, less so the fact of having tubal disease or not (discrimination).
A strength of this study is the national multicentre design, which enabled us to include a large number of women. A possible disadvantage of the multicentre nature is the potential for heterogeneity, due to differences in tubal testing protocols between hospitals or because of variability in tubal pathology prevalence between different areas of the country. Therefore, we explored whether the centre as such acted as a confounder for variables that were included in the models. To do so, we added centre as a covariate to the multivariate regression models. This analysis did not alter the models, nor did it affect the strength of associations.
Data from medical history might suffer from a lack of reliability, and one cannot exclude that certain items have been reported incorrectly. For example, due to the design of the study, not all cases that reported a history of PID or Chlamydia infection were verified by laparoscopy or culture. On the other hand, some women may have reported incorrectly not having suffered an episode of PID. This will have caused ‘false-positive’ or ‘false-negative’ histories. One should realise however that these reports do also occur in daily clinical practice, where the information from medical history is collected during the fertility workup on a routine basis, a situation that is reflected by our prospective study.
From a study design perspective, ideally all women would have undergone visual assessment of their Fallopian tubes. In the present study, 32% of eligible women did not undergo tubal testing, as they conceived prior to tubal testing, or were referred for treatment with IVF, thereby surpassing the need for tubal testing. As our purpose was to examine the relation between medical history and outcome of tubal assessment, this mechanism could have led to biased estimates of associations. To examine the impact of this partial verification, we explored whether there were any systematic differences between women that did and did not undergo tubal testing. This was indeed the case, although the absolute differences were small. However, studies in which the diagnosis is immediately verified in all women are in our opinion hardly feasible, leaving the present approach as the most practical and possible one.
We chose to develop models for two distinct definitions of tubal pathology; one in which any type of tubal pathology was considered abnormal, and one in which only bilateral abnormalities were considered to be abnormal. The latter is in our opinion clinically the one most relevant, as these women have virtually no chance of conceiving either spontaneously or after intrauterine insemination. From a clinical point of view, it is undesirable that these women are exposed to a long period of expectant management and they should be identified early in the diagnostic workup and counselled for IVF. At present, the detection of unilateral tubal pathology is unlikely to imply a major change in clinical management, unless the woman already has a poor chance of achieving a treatment independent pregnancy (Van der Steeg et al., 2007).
We limited the study to those women that underwent their first tubal patency test. Couples with a history of previous tubal testing, previous reproductive surgery or previous IVF are less likely to undergo tubal testing than couples without such a history. We also excluded couples in whom the woman was anovulatory or in whom the man had severe oligoasthenozoospermia requiring IVF–ICSI. The rationale for excluding these couples was to prevent biased associations as in these conditions tubal patency testing will not always be performed routinely.
To our surprise, we found male characteristics to be associated with the probability of tubal subfertility. Although male age, male smoking habits and a pregnancy in a previous relationship cannot causally be related to female tubal damage, it is possible that these male characteristics act as indicators for some unrecorded variables, such as number of sexual partners or socio-economic status, or reflect lifestyle and selection mechanisms. We performed sub-analyses in which only female factors were evaluated, but these models had a significantly lower predictive performance than the models in which data of the couple were included (data not shown). Furthermore, the model was developed from data collected in over 30 centres, and we corrected for possible confounding cluster effects of male characteristics. However, it might be possible that these associations do not hold in a different population.
There are two ways in which our decision rule can be used. A clinician can use the paper score chart, which transforms the total sum score into a probability of tubal disease. Alternatively, the clinician can use the regression formula, which can be incorporated into software, i.e. used in an electronic patient file, or a handheld device such as a PDA.
Preferences of subfertile women regarding the trade-off between the likelihood of detecting tubal pathology and the risks and discomforts associated with tubal testing are currently unknown; therefore, a clear cut-off level cannot be advised at present. The decision rules presented here cannot be used to exclude subfertile women from tubal patency testing overall, but allow one to estimate a probability of disease in order to time tubal patency tests adequately. This means that in clinical practice, tubal testing can be performed directly in case of a high probability of tubal disease or postponed for several months in women with a low probability.
In conclusion, the present study provides two easy-to-use decision rules that can accurately express the women's probability of (severe) tubal pathology at the couple's first consultation. They could be used to select women for tubal testing more efficiently.
|
Funding |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionFundingAppendixAcknowledgementsReferences |
|---|
VIDI-programme of ZonMW (grant 917.46.364), The Hague, The Netherlands; MMC Academy of the Máxima Medical Centre, Veldhoven, The Netherlands. Funding for open access charge: National Institutes of Health (CB5453961).
|
Appendix |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionFundingAppendixAcknowledgementsReferences |
|---|
The chance of tubal pathology can be calculated from the multivariable models with the formula: probability = 1/[1 + exp(–beta)], where the betas for the different models are: ‘any tubal pathology’ model: beta = –2.086 + (0.085 x duration of child wish) + (0.445 x conceiving of only male partner prior relation) + (0.672 x both partners conceived prior relation) + (0.193 x male smoker) + (0.680 x PID) + (0.494 x Chlamydia infection) + (1.729 x ectopic pregnancy) + (0.280 x legally induced abortion) + (0.771 x pelvic surgery); ‘bilateral tubal pathology’ model: beta = –2.201 + (0.351 x referral gynaecologist) + (0.054 x female age above 32 years) + (0.129 x duration of child wish) + (–0.235 x conceived in present relation) + (0.792 x both partners conceived prior relation) + (–0.034 x male age) + (0.230 x dysmenorrhoea) + (0.272 x female smoker) + (0.259 x male smoker) + (0.889 x PID) + (0.583 x Chlamydia infection) + (0.961 x ectopic pregnancy) + (0.305 x pelvic surgery).
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionFundingAppendixAcknowledgementsReferences |
|---|
The following centres participated in this study: Medisch Centrum Alkmaar, Dr YM van Kasteren; Twenteborg Ziekenhuis Almelo, Dr PFM van der Heijden; Meander Medisch Centrum, Drs WA Schöls; Academisch Medisch Centrum Amsterdam, Drs MH Mochtar; BovenIJ Ziekenhuis Amsterdam, Drs GLM Lips; Sint Lucas Andreas Ziekenhuis, Drs J Dawson; Onze Lieve Vrouwe Gasthuis Amsterdam, Drs HR Verhoeve; Slotervaart Ziekenhuis Amsterdam, Drs S Milosavljevic; Vrije Universiteit MC Amsterdam, Dr PGA Hompes; Gelre Ziekenhuis Apeldoorn, Drs LJ van Dam; Wilhelmina Ziekenhuis Assen, Dr AV Sluijmer; Rode Kruis Ziekenhuis Beverwijk, Dr HE Bobeck; Ziekenhuis Gooi-Noord Blaricum, Dr RE Bernardus; Amphia Ziekenhuis Breda, Drs MCS Vermeer; Westeinde Ziekenhuis Den Haag, Dr JP Dörr; Ziekenhuis Deventer, Dr PJQ van der Linden; Albert Schweitzer Ziekenhuis Dordrecht, Drs HJM Roelofs; Scheper Ziekenhuis Emmen, Drs JM Burggraaff; Medisch Spectrum Twente, Dr GJE Oosterhuis; St Anna Ziekenhuis Geldrop, Drs MH Schouwink; Spaarne Ziekenhuis Haarlem, Dr MH Emanuel; Atrium Medisch Centrum Heerlen, Dr PXJM Bouckaert; Elkerliek Ziekenhuis Helmond, Dr FMC Delemarre; Jeroen Bosch Ziekenhuis ‘s-Hertogenbosch, Dr CJCM Hamilton; Ziekenhuis Hilversum, Drs M van Hoven; Westfries Gasthuis Hoorn, Dr CM Renckens; Academisch Ziekenhuis Maastricht, Dr JA Land; Sint Antonius Ziekenhuis Nieuwegein, Dr JH Schagen-van Leeuwen; Universitair Medisch Centrum Nijmegen, Dr JAM Kremer; Waterland Ziekenhuis Purmerend, Drs C van Katwijk; Sint Franciscus Gasthuis Rotterdam, Dr MHA van Hooff; Twee Steden Ziekenhuis Tilburg/Waalwijk, Dr HJHM van Dessel; Universitair Medisch Centrum Utrecht, Dr FJM Broekmans; Ziekenhuis Bernhoven Veghel, Dr HJLA Ruis; Máxima Medisch Centrum Veldhoven, Dr CAM Koks; Vie Curi Medisch Centrum Venlo, Drs P Bourdrez; De Heel Zaandam, Drs WWJ Riedijk; Isala Klinieken Zwolle, Dr BJ Cohlen.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionFundingAppendixAcknowledgementsReferences |
|---|
Coppus SF, Opmeer BC, Logan S, van der Veen F, Bhattacharya S, Mol BW. The predictive value of medical history taking and Chlamydia IgG ELISA antibody testing (CAT) in the selection of subfertile women for diagnostic laparoscopy: a clinical prediction model approach. Hum Reprod (2007) 22:1353–1358.
Evers JL. Female subfertility. Lancet (2002) 360:151–159.[CrossRef][Web of Science][Medline]
Forsey JP, Caul EO, Paul ID, Hull MG. Chlamydia trachomatis, tubal disease and the incidence of symptomatic and asymptomatic infection following hysterosalpingography. Hum Reprod (1990) 5:444–447.
Harrell FE Jr, Lee KL, Pollock BG. Regression models in clinical studies: determining relationships between predictors and response. J Natl Cancer Inst (1988) 80:1198–1202.
Hubacher D, Grimes D, Lara-Ricalde R, de la Jarra J, Garcia-Luna A. The limited clinical usefulness of taking a history in the evaluation of women with tubal factor subfertility. Fertil Steril (2004) 81:6–10.[CrossRef][Web of Science][Medline]
Liberty G, Gal M, Halevy-Shalem T, Michaelson-Cohen R, Galoyan N, Hyman J, Eldar-Geva T, Vatashky E, Margalioth E. Lidocaine-Prilocaine (EMLA) cream as analgesia for hysterosalpingography: a prospective, randomized, controlled, double blinded study. Hum Reprod (2007) 22:1335–1339.
Mol BW, Coppus SF, Van der Veen F, Bossuyt PM. Evaluating predictors for the outcome of assisted reproductive technology: ROC-curves are misleading; calibration is not! Fertil Steril (2005) 84:s253–s254. (Abstract).
National Institute for Clinical Excellence. Guideline CG11: Fertility: Assessment and Treatment for People with Fertility Problems. (2004) (http://www.nice.org.uk/download.aspx?o=CG011fullguideline).
Nederlandse Vereniging voor Obstetrie en Gynaecologie. Orienterend fertiliteitsonderzoek [NVOG guideline diagnostic fertility workup]. (2004).
Steyerberg EW, Eijkemans MJ, Harrell FE Jr, Habbema JD. Prognostic modelling with logistic regression analysis: a comparison of selection and estimation methods in small data sets. Stat Med (2000) 19:1059–1079.[CrossRef][Web of Science][Medline]
Steyerberg EW, Harrell FE Jr, Borsboom GJ, Eijkemans MJ, Vergouwe Y, Habbema JD. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis. J Clin Epidemiol (2001) 54:774–781.[CrossRef][Web of Science][Medline]
Van der Heijden GJ, Donders AR, Stijnen T, Moons KG. Imputation of missing values is superior to complete case analysis and the missing-indicator method in multivariable diagnostic research: a clinical example. J Clin Epidemiol (2006) 59:1102–1109.[CrossRef][Web of Science][Medline]
Van der Steeg JW, Steures P, Eijkemans MJ, Habbema JD, Hompes PG, Broekmans FJ, van Dessel HJ, Bossuyt PM, van der Veen F, Mol BW. Pregnancy is predictable: a large-scale prospective external validation of the prediction of spontaneous pregnancy in subfertile. Hum Reprod (2007) 22:536–542.
Submitted on April 19, 2007; resubmitted on May 25, 2007; accepted on June 26, 2007.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||