Author's response to reviews Title: Who does not participate in a follow-up postal study? A survey of infertile couples treated by in vitro fertilization Authors: Pénélope Troude (penelope.troude@inserm.fr) Estelle Bailly (estelle.bailly@ined.fr) Juliette Guibert (juliette.guibert@imm.fr) Jean Bouyer (jean.bouyer@inserm.fr) Elise de La Rochebrochard (roche@ined.fr) Version: 2 Date: 28 May 2012 Author's response to reviews: see over
Pénélope Troude Inserm CESP U1018 - INED 94276 Le Kremlin-Bicêtre France tel : +33 145212150 fax : +33 145212075 E-mail : penelope.troude@inserm.fr May 28, 2012 Dear Editor, Please find enclosed our manuscript, entitled Who does not participate in a follow-up postal study? A survey of infertile couples treated by in vitro fertilization (MS: 1004901016686531), which we wish to resubmit for publication as an original article in BMC Medical Research Methodology. We would like to thank you and the reviewers for your helpful comments. We have modified the manuscript according to your remarks, as described in the point-by-point answer below. Sincerely, The Authors - 1 -
Reviewer's report Title: Who does not participate in a postal study? A survey of infertile couples treated by in vitro fertilization Version: 1 Date: 18 March 2012 Reviewer: Batool Rashidi Reviewer's report: Authors response: We thank the reviewer for these helpful comments. Please find below our point-by-point response. minor essential revisions 1-in result section you reported contact rate~62% but in the discussion it is 36%, which one is correct? Authors response: We recognize that this is confusing. 62% corresponds to the contact rate and 36% corresponds to the participation rate (including contact and response). The result section has been modified page 8 to clarify this point and so, the proportions are the same in the results and discussion sections: Among the 6,507 couples who began an IVF program in 2000-2002, the contact rate was 62% (n = 4,029). Among contacted couples, the response rate was 58% (n = 2,321). ). Therefore, 36% of the initial cohort participated in the postal study (Fig. 1). The proportion of couples who had a child during IVF treatment was higher among contacted (44%) than among non-contactable couples (38%) and it was higher among respondents (53%) than among non-respondents (31%) Moreover, Figure 1 has been modified and now presents all rates without decimals, to have exactly the same numbers in the figure and in the text. Definition of all rates reported in the manuscript is given in the analysis section, bottom of page 6 : Contact rate was defined as the number of couples contacted among the total number of included couples, response rate as the number of respondents among the contacted couples, and participation rate as the number of respondents among the total number of included couples. - 2 -
2-it is better to have a regression analysis in no birth couples between responders and non responders, in order to find out the effect of childless on response rate. Authors response: Participation analysis was conducted on all couples, whether or not they obtained a birth in their inclusion center, in order to include this variable in our analysis so that we could study its impact on participation. As expected, having a child during IVF was associated with a higher probability of response to the postal study. However, we also observed an association between contact rate and having a child during IVF, which was more surprising. Following the reviewer s suggestion, analyses have been conducted only among unsuccessfully treated couples (no-birth couples), and thus after having removed the variable result of IVF. Multivariate analyses of factors associated with contact (n=3,597) and of factors associated with response (n=2,274) among unsuccessfully treated couples are presented below in tables 1 and 2, respectively. Results regarding the different variables (other than result of IVF) among unsuccessfully treated couples are very close to those observed for the whole cohort for probability of contact as well as for probability of response. Table 1 Factors associated with probability of contact in the study among unsuccessfully treated couples (n=3,597) Multivariate analysis OR 95% CI P Woman s age (years) <0.001* < 30 1 30-34 1.46 1.21-1.77 35-39 1.50 1.23-1.83 40 1.69 1.33-2.15 Inclusion center <0.001 Marseille 1 Bois-Guillaume 0.79 0.63-0.97 Sèvres 1.06 0.83-1.37 Besançon 0.99 0.76-1.30 Caen 1.83 1.40-2.41 Cochin 1.07 0.82-1.39 Clermont-Ferrand 1.18 0.89-1.56 Montsouris 1.56 1.18-2.05 Year of 1 st oocyte retrieval <0.001 2000 1-3 -
2001 1.13 0.94-1.35 2002 1.43 1.19-1.72 Origin of infertility 0.206 Female 1 Male 1.03 0.97-1.23 Couple 0.90 0.74-1.09 Unexplained 0.83 0.67-1.04 Number of embryos obtained 0.030 at 1 st attempt 0-1 1 2-5 1.10 0.93-1.31 > 5 1.29 1.06-1.55 Number of attempts <0.001 1 1 2-4 1.59 1.37-1.84 >4 3.76 2.78-5.08 * P for trend - 4 -
Table 2 Factors associated with probability of response to the postal questionnaire among unsuccessfully treated couples (n = 2,152) Multivariate analysis OR 95% CI P Woman s age (years) 0.001 < 30 1 30-34 1.48 1.15-1.90 35-39 1.01 0.78-1.31 40 0.64 0.47-0.88 Inclusion center <0.001 Marseille 1 Bois-Guillaume 1.76 1.32-2.34 Sèvres 1.37 0.99-1.88 Besançon 1.33 0.93-1.91 Caen 2.18 1.58-3.00 Cochin 0.88 0.62-1.24 Clermont-Ferrand 1.50 1.06-2.12 Montsouris 0.96 0.70-1.33 Year of 1 st oocyte retrieval 0.003 2000 1 2001 1.45 1.14-1.83 2002 1.44 1.14-1.82 Origin of infertility 0.14 Female 1 Male 0.85 0.69-1.06 Couple 0.80 0.62-1.03 Unexplained 1.09 0.81-1.46 Number of embryos obtained 0.79 at 1 st attempt 0-1 1 2-5 1.06 0.85-1.32 > 5 1.08 0.85-1.38 Number of attempts 0.027 1 1 2-4 1.25 1.03-1.53 >4 1.44 1.06-1.97-5 -
This point has not yet been added to the manuscript, but it could be done if the editor considers it necessary. Level of interest: An article whose findings are important to those with closely related research interests Quality of written English: Acceptable Statistical review: Yes, but I do not feel adequately qualified to assess the statistics. Declaration of competing interests: I declare that I have no competing interest - 6 -
Reviewer's report Title: Who does not participate in a postal study? A survey of infertile couples treated by in vitro fertilization Version: 1 Date: 30 April 2012 Reviewer: Sakari Bertel Alfred Suominen Reviewer's report: Authors response: We thank the reviewer for these helpful comments. Please find below our point-by-point response. The ms is dealing with an important methodological topic, namely the drop-out related to a postal follow up study of infertile couples. The data are collected from several centers which increases the scientific significance, I am prepared to give my support for publishing the study but this would according to my opinion need a major type of revision with delivery of some additional data that I did not find from the present version of the ms. The principal data lacking is about how many couples could theoretically have been contacted or then the matter is not clearly enough described. Totally 6507 coupled were contacted but did they represent all couples that theoretically could have been reached? By what kind of mechanism did one end up to this number? Authors response: The 6,507 couples represent all couples that theoretically could have been reached. The Population section has been modified to clarify this point, page 6: This study is based on the DAIFI study (Devenir Après Initiation de la FIV, outcome after IVF initiation), a retrospective cohort exhaustively including all couples who began an IVF program between 2000 and 2002 (n = 6,507) in one of the eight participating French IVF centers (the centers at Besançon University Hospital, Cochin Hospital, Caen, Marseille, Sèvres, Bois-Guillaume, Clermont-Ferrand and Montsouris). Moreover, more data on potential differences between inclusion centers are needed and if necessary also discussed. Authors response: Between-center differences observed for contact and response probably reflect a complex reality. - 7 -
Regarding probability of contact, observed differences probably involved both probability of the likelihood of relocation in a given area and the resources devoted by the center to updating patients addresses. To develop this point, the discussion section has been modified as follows, page 10: The association between inclusion center and probability of contact may be linked to differences between centers in financial and human resources devoted to patient address update. It may also reflect the geographical location of the center as well as population dynamics, with mobility rates that can vary widely between regions. For instance, a change of address may be more likely in more urbanized areas [27]. Regarding probability of response to the postal questionnaire, differences observed between centers probably reflect differences in couples. Couples feelings about their treatment in IVF center may lead them to respond or not according to their perception of this experience, whether negative or positive. The socioeconomic level of couples may also partly explain differences observed between centers. Indeed, educational level is known to be associated with participation in epidemiological studies. To develop this point, the discussion section has been modified page 11, as follows: The inverse J-pattern between age and response suggests that age impacts as a medical factor on probability of response. Probability of response was also associated with inclusion center. Differences observed between centers may reflect in part couples feelings on their IVF treatment in the center, but probably also reflects sociodemographic characteristics of couples that may vary according to geographical localization. Indeed, socioeconomic and educational levels are known to be associated with response rates in epidemiological studies [1, 8]. The trend toward a higher response rate among couples with unexplained infertility than in couples with infertility of female origin also suggests that demographic and medical factors influence contact and response in different ways. - 8 -
The N of those that participated in the postal survey should be given already in the Abstract. Authors response: The abstract has been modified as follows: The DAIFI study is a retrospective cohort including 6,507 couples who began an IVF program in 2000-2002 in one of the eight participating French IVF centers. Medical data on all 6,507 couples were obtained from IVF center databases, and information on longterm outcome was available only for participants in the postal survey (n = 2,321). Finally, one has to keep in mind that the participants are highly selected, i.e. couples receiving treatment for infertility. Moreover, being able to have a child could for many become or primarily is a matter of uttermost importance and hence can also clearly influence the motivation to participate in a study. Results supporting this assumption were gained but I would still like to have a more thorough discussion about can any aspects of the results be generalized to health follow up surveys. Authors response: We agree with the reviewer that the participants are highly selected, but that is very often the case in health follow-up surveys (studying a specific treatment, illness or exposure). However, this study shows how selection bias can be much more complex and challenging than an a priori hypothesis would suggest (especially for contact). The generalized conclusion is on how careful one should be about selection bias and the importance of trying to collect information during the study in order to explore such questions and be able to use appropriate statistical tools in analysis afterwards. The conclusion has been modified (pages 11 and 12) to develop further the generalizability of our results: It is necessary to understand the mechanisms underlying contact and response in order to choose the appropriate methodology for analysis of the results of epidemiological surveys [35]. To take into account attrition and potential bias, new methods are being developed but most rely on hypotheses that require an understanding of attrition mechanisms [36]. Studies on attrition mechanisms are needed, especially as these may vary according to the study population. In our study based on infertile couples treated by IVF, we found that an a priori hypothesis on attrition may be too simplistic and may underestimate potential bias. Non-response as well as non-contact were linked to the outcome of interest. Attrition is a common issue in all health surveys and one that is rarely addressed in analysis. This study illustrates the importance of developing a study design that yields a minimum of - 9 -
information on the whole of the eligible population. In the context of growing use of analytical methods that take attrition into account (such as multiple imputation), we need to better understand the mechanisms that underlie attrition in order to choose the most appropriate method. I would still like to remind the authors about the fact that how well the participants represent some population is a different question from that how reliable are the findings about associations between the variables studied, i.e. the participants can be somewhat biased but the associations between the variables studied can in spite of this theoretically be of the same magnitude as would be the case with unbiased participants. Authors response: We agree with the reviewer s comment. Indeed, several studies that have assessed the impact of non-participation on measures of association have provided reassuring results (Nohr et al., 2006; Osler et al., 2008; Pizzi et al., 2010). However, understanding selection bias appears to be essential if we are concerned about association analysis. Moreover, when estimating a frequency, the issue is much more serious. Nohr EA, Frydenberg M, Henriksen TB, Olsen J. Does low participation in cohort studies induce bias? Epidemiology 2006, 17:413-418. Osler M, Kriegbaum M, Christensen U, Lund R. Nybo Andersen AM. Loss to follow up did not bias associations between early life factors and adult depression. J Clin Epidemiol 2008, 61:958-963. Pizzi C, De Stavola B, Merletti F, Bellocco R, dos Santos Silva I, Pearce N, Richiardi L. Sample selection and validity of exposure-disease association estimates in cohort studies. J Epidemiol Community Health 2011, 65:407-411. Epub 2010 Sep 29. Minor comments The title is somewhat misleading. in its present form and should rather be 'Who does not participate in a follow up postal study?' Authors response: The title has been modified as suggested by the reviewer and is now: Who does not participate in a follow-up postal study? A survey of infertile couples treated by in vitro fertilization - 10 -
The first sentence of the Abstract is probably wrongly written. Now it is: 'As a good response rate has been considered as a proof of a study s quality, decreasing participation and its potential impact on the internal validity of the study are of growing interest.' In my view the wording should be 'A good response rate has been considered as a proof of a study s quality, decreasing participation and its potential impact on the internal validity of the study are of growing interest.' Authors response: The first sentence of the Abstract has been modified as follows: A good response rate has been considered as a proof of a study s quality. Decreasing participation and its potential impact on the internal validity of the study are of growing interest. Moreover, the Introduction section has been also modified page 4, as follows: Participation rates in cohort studies have decreased during the last two decades[1]. A good response rate has been considered as a proof of a study s quality [2]. Therefore, decreasing participation and its potential impact on the internal validity of studies are of growing interest [3-5]. Level of interest: An article of importance in its field Quality of written English: Acceptable Statistical review: Yes, but I do not feel adequately qualified to assess the statistics. Declaration of competing interests: No competing interests - 11 -