Does This Woman Have Adnexal Torsion?
Does This Woman Have Adnexal Torsion?
Our team has developed a SAQ-GE (Huchon et al., 2012). The SAQ-GE is an 89-item questionnaire divided into six themes: (i) qualitative description of pain, (ii) intensity of pain, (iii) location and (iv) evolution of pain, (v) vaginal bleeding and (vi) associated signs.
This SAQ-GE was constructed using semi-structured interviews in 2003 with 39 patients at the Poissy-Saint Germain en Laye Hospital, who were subsequently diagnosed with gynecologic emergencies, including 5 AT (Huchon et al., 2012). An adaptation of Colaizzi's method (Colaizzi, 1978) was used for a qualitative assessment of the semi-structured interview.
The SAQ-GE was finally built with items identified from the semi-structured interviews and that were: (i) often associated with potentially life-threatening gynecologic emergencies; and/or (ii) specific of potentially life-threatening gynecologic emergencies; and/or (iii) given high ratings by French experts identified by searching Medline for publications on gynecologic emergencies. To make the questionnaire easy to understand by the patients, we worded the items using the phrases and sentences collected from the semi-structured interviews.
Participants From September 2006 to April 2008, we asked all patients presenting at the study-center's gynecologic EDs for acute pelvic pain to complete the SAQ-GE on a voluntary basis. In France, the diagnosis of acute pelvic pain may take place in general EDs, followed by referral to a gynecologic ED, or directly in gynecologic EDs to which all patients have free access. So, all patients presenting gynecologic emergencies are seen in gynecologic EDs.
The patients were enrolled at five gynecology departments in the Greater Paris metropolitan region, France. Four departments were in teaching hospitals (Poissy-Saint Germain en Laye, Créteil, Port-Royal and Louis Mourier) and one was in a general hospital (Versailles). All these gynecologic EDs have a resident and a senior gynecologist on duty 24 h a day with night and day access to ultrasonography and operative rooms. They also have appropriate written pain management guidelines and a specific interest in gynecologic emergencies.
The SAQ-GE was completed by patients themselves after appropriate initial pain management and before any surgery. The nurses then collected the completed questionnaires.
Exclusion criteria were a history of chronic pelvic pain, neurologic or psychiatric disease, hemodynamic instability and no knowledge of French. Patients with a verbal numeric pain rating scale (NRS) <4 on an 11-point rating scale and with bartholinitis were excluded from the study.
The study was approved by the French Department of Higher Education and Research (no. 06.336) and by the French National Committee of Information Technology and Individual Liberties (no. 906253).
Diagnosis of AT The reference standard for the diagnosis of AT was the laparoscopic diagnosis. AT was defined as the twisting of the adnexa, ovary or tube alone by at least one complete turn, around a center-line consisting of the infundibulopelvic ligament and tubo-ovarian ligament.
Patients who did not have a laparoscopy were classed as not having AT. Diagnoses of other diseases were obtained by the various investigations performed in EDs or during hospitalization: clinical examination and/or ultrasonography and/or computed tomographic scan. Some diagnoses were made by surgery: laparoscopy, dilatation and curettage or diagnostic hysteroscopy. Ectopic pregnancy was diagnosed by either surgery or an algorithm (Mol et al., 1999). Diagnosis of pelvic infectious disease was also made by surgery, if needed, or by a non-invasive prediction rule (Kahn et al., 1991). Uncertain initial diagnoses not requiring immediate emergency care, such as early pregnancy of uncertain location, were reconvened in the early pregnancy clinic until diagnostic certainty (Bignardi et al., 2010). When a patient was discharged from a gynecologic ED, she was advised to refer again if the pain continued or started again.
Statistical Methods The required sample size was estimated as follows. Based on the appropriate selection of several items of the SAQ-GE, we expected to be able to derive a prediction rule for preoperative AT screening. To rule out AT, our analysis had to focus on sensitivity (Se) (Loong, 2003). To be of clinical interest, the rule has to achieve at least 95% Se (Jaeschke et al., 1994). Inversely, we would conclude that the questionnaire is inefficient if we were not able to derive any model with a Se over 80%. On a pragmatic basis, type III errors (Schwartz and Lellouch, 2009) were taken into consideration in order to determine the required sample size. The first type III error is defined here as the probability that the observed Se exceeds 95% while the true Se is equal or below 80%. The second type III error is defined as the probability that the observed Se is below 80% while the true Se is equal or above 95%. If the observed Se for the diagnosis of AT by the SAQ-GE is between 80 and 95%, further investigations would be necessary to assess its exact value for the preoperative diagnosis of AT. Based on the binomial distribution, we calculated that the inclusion of 30 patients with AT would guarantee that both type III errors are below 0.025 (one-sided).
The prediction rule was derived as follows: first we compared the patients with and without AT according to their answers to questionnaire items, using Pearson's χ test or Fisher's exact test. For each variable that was significantly associated with AT at a threshold of P < 0.05, we computed Se, specificity (Sp), positive likelihood ratio (Lr+), negative likelihood ratio (Lr−) and crude diagnostic odds ratio.
We then used multiple logistic regression analysis to select the best combination of variables for predicting AT among variables with P-values below 0.05 in the bivariate analyses. We identified the best combinations of variables independently associated with AT at a threshold of P< 0.05. Adjusted diagnostic odds ratios (aDOR) were computed with their 95% confidence intervals (95% CI).
Finally, the SAQ-GE torsion score was based on items found to be significant by multivariate analysis. Missing data were considered as absent. The number of torsion score points contributed by each score item was obtained by rounding up the β coefficients of the logistic regression to generate a simple scale. The area under the receiver operating characteristic curve (ROC-AUC) of the SAQ-GE torsion score was then compared with the ROC-AUC of the logistic regression to check that the two values were not significantly different. The probability of AT, Se, Sp, Lr+, Lr−, positive predictive value (PPV) and negative predictive value (NPV) were calculated for different SAQ-GE torsion score values.
Risk groups for AT were then constructed by cut-offs for the SAQ-GE torsion score, selected using the ROC curve, in order to maximize classification rates using the expected value of Se > 95%. Diagnostic values (Se, Sp, Lr+, Lr−, PPV and NPV) were also computed for non-pregnant patients.
The predictive ability of the derived model was then tested by cross-validation using the jackknife procedure (Efron and Gong, 1983). This method is especially useful for estimating the unbiased diagnostic performance of a model when no external validation population exists. The method was applied as follow: (i) the study population (N) was randomly stratified into 10 equivalent subgroups of N/10 (±1) women; (ii) for each subgroup (i), a new prediction model (Mi) was constructed with the entire study sample except those in subgroup (i); (iii) model Mi was then used to predict the outcome of the women omitted in constructing it. These steps were repeated for each subgroup.
Analyses were carried out using Stata® version 11.0 (Stata Corp., College Station, TX, USA.).
Materials and Methods
Self Assessment Questionnaire for Gynecologic Emergencies
Our team has developed a SAQ-GE (Huchon et al., 2012). The SAQ-GE is an 89-item questionnaire divided into six themes: (i) qualitative description of pain, (ii) intensity of pain, (iii) location and (iv) evolution of pain, (v) vaginal bleeding and (vi) associated signs.
This SAQ-GE was constructed using semi-structured interviews in 2003 with 39 patients at the Poissy-Saint Germain en Laye Hospital, who were subsequently diagnosed with gynecologic emergencies, including 5 AT (Huchon et al., 2012). An adaptation of Colaizzi's method (Colaizzi, 1978) was used for a qualitative assessment of the semi-structured interview.
The SAQ-GE was finally built with items identified from the semi-structured interviews and that were: (i) often associated with potentially life-threatening gynecologic emergencies; and/or (ii) specific of potentially life-threatening gynecologic emergencies; and/or (iii) given high ratings by French experts identified by searching Medline for publications on gynecologic emergencies. To make the questionnaire easy to understand by the patients, we worded the items using the phrases and sentences collected from the semi-structured interviews.
Derivation of the SAQ-GE Torsion Score
Participants From September 2006 to April 2008, we asked all patients presenting at the study-center's gynecologic EDs for acute pelvic pain to complete the SAQ-GE on a voluntary basis. In France, the diagnosis of acute pelvic pain may take place in general EDs, followed by referral to a gynecologic ED, or directly in gynecologic EDs to which all patients have free access. So, all patients presenting gynecologic emergencies are seen in gynecologic EDs.
The patients were enrolled at five gynecology departments in the Greater Paris metropolitan region, France. Four departments were in teaching hospitals (Poissy-Saint Germain en Laye, Créteil, Port-Royal and Louis Mourier) and one was in a general hospital (Versailles). All these gynecologic EDs have a resident and a senior gynecologist on duty 24 h a day with night and day access to ultrasonography and operative rooms. They also have appropriate written pain management guidelines and a specific interest in gynecologic emergencies.
The SAQ-GE was completed by patients themselves after appropriate initial pain management and before any surgery. The nurses then collected the completed questionnaires.
Exclusion criteria were a history of chronic pelvic pain, neurologic or psychiatric disease, hemodynamic instability and no knowledge of French. Patients with a verbal numeric pain rating scale (NRS) <4 on an 11-point rating scale and with bartholinitis were excluded from the study.
The study was approved by the French Department of Higher Education and Research (no. 06.336) and by the French National Committee of Information Technology and Individual Liberties (no. 906253).
Diagnosis of AT The reference standard for the diagnosis of AT was the laparoscopic diagnosis. AT was defined as the twisting of the adnexa, ovary or tube alone by at least one complete turn, around a center-line consisting of the infundibulopelvic ligament and tubo-ovarian ligament.
Patients who did not have a laparoscopy were classed as not having AT. Diagnoses of other diseases were obtained by the various investigations performed in EDs or during hospitalization: clinical examination and/or ultrasonography and/or computed tomographic scan. Some diagnoses were made by surgery: laparoscopy, dilatation and curettage or diagnostic hysteroscopy. Ectopic pregnancy was diagnosed by either surgery or an algorithm (Mol et al., 1999). Diagnosis of pelvic infectious disease was also made by surgery, if needed, or by a non-invasive prediction rule (Kahn et al., 1991). Uncertain initial diagnoses not requiring immediate emergency care, such as early pregnancy of uncertain location, were reconvened in the early pregnancy clinic until diagnostic certainty (Bignardi et al., 2010). When a patient was discharged from a gynecologic ED, she was advised to refer again if the pain continued or started again.
Statistical Methods The required sample size was estimated as follows. Based on the appropriate selection of several items of the SAQ-GE, we expected to be able to derive a prediction rule for preoperative AT screening. To rule out AT, our analysis had to focus on sensitivity (Se) (Loong, 2003). To be of clinical interest, the rule has to achieve at least 95% Se (Jaeschke et al., 1994). Inversely, we would conclude that the questionnaire is inefficient if we were not able to derive any model with a Se over 80%. On a pragmatic basis, type III errors (Schwartz and Lellouch, 2009) were taken into consideration in order to determine the required sample size. The first type III error is defined here as the probability that the observed Se exceeds 95% while the true Se is equal or below 80%. The second type III error is defined as the probability that the observed Se is below 80% while the true Se is equal or above 95%. If the observed Se for the diagnosis of AT by the SAQ-GE is between 80 and 95%, further investigations would be necessary to assess its exact value for the preoperative diagnosis of AT. Based on the binomial distribution, we calculated that the inclusion of 30 patients with AT would guarantee that both type III errors are below 0.025 (one-sided).
The prediction rule was derived as follows: first we compared the patients with and without AT according to their answers to questionnaire items, using Pearson's χ test or Fisher's exact test. For each variable that was significantly associated with AT at a threshold of P < 0.05, we computed Se, specificity (Sp), positive likelihood ratio (Lr+), negative likelihood ratio (Lr−) and crude diagnostic odds ratio.
We then used multiple logistic regression analysis to select the best combination of variables for predicting AT among variables with P-values below 0.05 in the bivariate analyses. We identified the best combinations of variables independently associated with AT at a threshold of P< 0.05. Adjusted diagnostic odds ratios (aDOR) were computed with their 95% confidence intervals (95% CI).
Finally, the SAQ-GE torsion score was based on items found to be significant by multivariate analysis. Missing data were considered as absent. The number of torsion score points contributed by each score item was obtained by rounding up the β coefficients of the logistic regression to generate a simple scale. The area under the receiver operating characteristic curve (ROC-AUC) of the SAQ-GE torsion score was then compared with the ROC-AUC of the logistic regression to check that the two values were not significantly different. The probability of AT, Se, Sp, Lr+, Lr−, positive predictive value (PPV) and negative predictive value (NPV) were calculated for different SAQ-GE torsion score values.
Risk groups for AT were then constructed by cut-offs for the SAQ-GE torsion score, selected using the ROC curve, in order to maximize classification rates using the expected value of Se > 95%. Diagnostic values (Se, Sp, Lr+, Lr−, PPV and NPV) were also computed for non-pregnant patients.
The predictive ability of the derived model was then tested by cross-validation using the jackknife procedure (Efron and Gong, 1983). This method is especially useful for estimating the unbiased diagnostic performance of a model when no external validation population exists. The method was applied as follow: (i) the study population (N) was randomly stratified into 10 equivalent subgroups of N/10 (±1) women; (ii) for each subgroup (i), a new prediction model (Mi) was constructed with the entire study sample except those in subgroup (i); (iii) model Mi was then used to predict the outcome of the women omitted in constructing it. These steps were repeated for each subgroup.
Analyses were carried out using Stata® version 11.0 (Stata Corp., College Station, TX, USA.).
Source...