A Prospective Assessment of Bowel Habit in Irritable Bowel Syndrome in Women: Defining an Alternator

Transcription

1 GASTROENTEROLOGY 2005;128: A Prospective Assessment of Bowel Habit in Irritable Bowel Syndrome in Women: Defining an Alternator DOUGLAS A. DROSSMAN,* CAROLYN B. MORRIS,* YUMING HU,* BRENDA B. TONER, NICHOLAS DIAMANT, JANE LESERMAN,* MICHAEL SHETZLINE, CHRISTINE DALTON,* and SHRIKANT I. BANGDIWALA* *UNC Center for Functional GI and Motility Disorders, University of North Carolina, Chapel Hill, North Carolina; Centre for Addiction and Mental Health, University of Toronto, Toronto, Ontario, Canada; and Novartis Pharmaceuticals, East Hanover, New Jersey Background & Aims: Irritable bowel syndrome (IBS) is subtyped as IBS with diarrhea (IBS-D) or IBS with constipation (IBS-C) based on Rome II guidelines. The remaining group is considered as having mixed IBS (IBS- M). There is no standard definition of an alternator (IBS-A), in which bowel habit changes over time. Our aim was to use Rome II criteria to prospectively assess change in bowel habit for more than 1 year to understand IBS-A. Methods: Female patients (n 317) with IBS entering a National Institutes of Health treatment trial were studied at baseline with questionnaires and 2-week daily diary cards of pain and stool frequency and consistency. Studies were repeated at the end of treatment (3 months) and at four 3-month intervals for one more year. Algorithms to classify subjects into IBS-D, IBS-C, and IBS-M groups used diary card information and modified Rome II definitions. Changes in bowel habit at 3-month intervals were then assessed using these surrogate diary card measures. Results: At baseline, 36% had IBS-D, 31% IBS-M, and 34% IBS-C. Except for stool frequency, there were no differences between groups. While the proportion of subjects in each subgroup remained the same over the year, most individuals (more than 75%) changed to either of the other 2 subtypes at least once. IBS-M was the least stable (50% changed out by 12 weeks). Patients were more likely to transition between IBS-M and IBS-C than between IBS-D and IBS-M. Notably, only 29% switched between the IBS-D and IBS-C subtypes over the year. Conclusions: While the proportion of subjects in each of the IBS subtypes stays the same, individuals commonly transition between subtypes, particularly between IBS-M and IBS-C. We recommend that IBS-A be defined as at least one change between IBS-D and IBS-C by Rome II criteria over a 1-year period. Irritable bowel syndrome (IBS) is defined as abdominal pain or discomfort associated with a change in bowel habit. 1 3 Because of the recent introduction of new drugs targeted at specific predominant stool subtypes, IBS has been further subclassified by Rome II consensus criteria 2 as diarrhea predominant (IBS-D) or constipation predominant (IBS-C). There is also growing interest to understand those with IBS who shift from diarrhea to constipation and vice versa (alternators), particularly with regard to their clinical response to new pharmaceutical agents. Some have designated the group with IBS who do not fulfill Rome criteria for IBS-D or IBS-C as having alternating IBS (IBS-A). However, there are no specific criteria for IBS-A, and not fulfilling criteria for either IBS-C or IBS-D at one point in time does not necessarily mean it is an alternating stool pattern. This group is more appropriately called mixed IBS (IBS-M). To identify a person with true IBS-A, changes in bowel habit over time are needed, preferably prospectively. In doing so, some patients initially characterized as having IBS-D, IBS-M, or IBS-C may be classified as having IBS-A if their stool patterns change during the period of observation. This type of assessment may be difficult to implement because (1) it is best accomplished with repeated prospective assessments, (2) the Rome criteria are not sufficient because they rely on patient recall at one point in time, and (3) patient recall may be faulty. 4 Furthermore, there are differences of opinion as to how an alternating bowel habit should be defined: as diarrhea alternating with constipation or as diarrhea or constipation alternating with normal bowel habit. This confusion in definition for IBS-A has led to considerable variation in its estimated prevalence based on the definition used. Therefore, to obtain more meaningful data and ultimately develop specific criteria for IBS-A, we studied individuals with Rome II defined IBS and their subtypes (IBS-D, IBS-M, and IBS-C) by tracking the bowel Abbreviations used in this paper: IBS, irritable bowel syndrome; IBS-A, alternating irritable bowel syndrome; IBS-C, irritable bowel syndrome with constipation; IBS-D, irritable bowel syndrome with diarrhea; IBS-M, mixed irritable bowel syndrome by the American Gastroenterological Association /05/$30.00 doi: /j.gastro

2 March 2005 ASSESSMENT OF BOWEL HABIT IN IBS IN WOMEN 581 pattern of these subtypes prospectively. Using a cohort of patients with IBS entering a treatment trial, we recorded their stool pattern with diary cards at baseline and every 3 months for 1 year after completion of the treatment trial. The aims of this study were to (1) identify and compare at baseline the clinical, physiologic, and health status features of IBS-D, IBS-M, and IBS-C; (2) using 2-week diary cards, determine pain scores and stool frequency and consistency for each group during 6 periods (a total of 15 months) of observation; (3) develop surrogate diary card criteria that approximate Rome II criteria; and (4) using these surrogate measures, evaluate stool patterns prospectively for IBS-D, IBS-M, and IBS-C to determine the frequency and rate of changes between subtypes. From this prospective information, we were able to characterize changes in stool patterns by group and then propose a recommendation for the definition of IBS-A. Patients and Methods Study Population and Protocol Between 1996 and 2001, female patients between the ages of 18 and 70 years with a functional bowel disorder were enrolled at the University of North Carolina and the University of Toronto to participate in a treatment trial. 4 Before randomization and at the end of the 3-month treatment period either with desipramine or pill placebo or with cognitivebehavioral treatment or education, patients filled out 2 weeks of daily diary cards and then underwent clinical and physiologic assessments. After the treatment period, patients received by mail additional questionnaires and diary cards at 6, 9, 12, and 15 months. The study was limited to women between the ages of 18 and 70 years with moderate to severe abdominal pain for at least 2 days per week for 6 months or more. Although patients included in the clinical trial could have IBS, constipation, or functional abdominal pain, for the purpose of this study, only patients who met Rome criteria for IBS were included (about 80% of the study population). Patients were not permitted to take antidepressants during the follow-up period. All patients signed an informed consent form that was approved by our committee for the protection of human subjects. A total of 317 women with IBS were studied. For details on recruitment methods, inclusion and exclusion criteria, and randomization, see Drossman et al. 4 Assessments Before randomization, all subjects filled out daily diary cards for 2 weeks that contained information on abdominal pain or discomfort (McGill Pain Questionnaire 5 and a visual analogue scale), stool frequency, and stool consistency using the Bristol Stool Scale form. 6 At the time of randomization and at study completion (at 12 weeks), patients received a comprehensive set of clinical and health care utilization questions and underwent testing for visceral sensitivity using a barostat. At 3, 6, 9, and 12 months after completion of treatment, patients filled out additional sets of daily diary cards and questionnaires. Detailed information on the assessment measures and their reliability and validity are presented elsewhere. 4 Data Analysis Data from the University of North Carolina and the University of Toronto were combined because there were no relevant variables that were different between sites at baseline or study completion. Descriptive statistics were obtained for each bowel category. Comparisons between bowel subtypes were performed by 2 tests for homogeneity for categorical variables and by t tests/analysis of variance for continuous variables. Because 1 of the 4 treatment arms (desipramine) could produce altered bowel habit, comparisons were made by individual treatment arm to determine whether desipramine itself accounted for any significant differences, and we found that it did not. Further, the desipramine-treated patients were randomly distributed in the 3 groups (IBS-D, 34.5%; IBS-M, 35.9%; IBS-C, 32.9%); thus, any differences would be equalized across groups. To track stool patterns over time, we classified subjects as having IBS-D, IBS-M, and IBS-C from their daily diary cards filled out for 2-week periods at 6 points in time, rather than the Rome II questions, which were asked only at baseline. Based on consensus of the authors, each Rome II question that asked about bowel habit was varied in its definition (most conservative to least conservative/most stringent) to elicit at least 4 different subpopulations. For example, for the Rome II question that asked whether the subject had hard and lumpy stools for at least 25% of the time over the past 3 months, we approximated this criterion with surrogate items identified from the diary cards and determined sensitivity and specificity for hard and lumpy stools for 7% (most conservative), 14%, 21%, 25%, and 28% (least conservative) of the 2-week period. This process was repeated for each of the Rome II bowel questions ( 3 bowel movements per week for 25% of the time, 3 bowel movements per day for 25% of the time, and loose or watery stools for 25% of the time in the past 3 months). Each grouping was then evaluated and displayed with receiver operating curve plots to see which version had the greatest sensitivity, the greatest specificity, the greatest product of sensitivity and specificity, and the most direct clinical comparison. We then grouped question sets by combining the versions per question with similar criteria (ie, greatest sensitivity). We then followed the identical Rome II algorithm (for the bowel habit questions) with each of the question sets to classify the individuals as having IBS-D, IBS-M, or IBS-C. Weighted statistics were used to determine which set best approximated the Rome II criteria by calculating the level of agreement between the classifications using Rome II criteria and the diary cards. Fleiss Cohen weights were chosen because a switch from IBS-M to either IBS-C or IBS-D (and vice versa) was seen as more plausible and

3 582 DROSSMAN ET AL GASTROENTEROLOGY Vol. 128, No. 3 thus expected to exhibit covariances with greater weights than a switch between IBS-D and IBS-C. This allowed us to classify subjects as having IBS-D, IBS-C, or IBS-M using diary cards prospectively that were based on Rome II definitions at one point in time. Results Study Population At baseline, there were 317 women with IBS. These patients were subcategorized as having IBS-D (35.6%), IBS-M (neither IBS-D nor IBS-C; 30.6%), or IBS-C (33.8%) using Rome II definitions. The mean age was 38.6 years, 84.9% were white, the mean education level was 14.9 years, and 49.2% were married. There were no demographic differences between subtypes of IBS, except that women with IBS-M were about 2 years younger than women with IBS-D or IBS-C (P.04). Baseline Comparisons of Clinical Variables Table 1 compares the 3 IBS subtypes at baseline in terms of clinical, physiologic, health care utilization, and quality-of-life features. Overall, except for stool frequency and possibly stool consistency, there were no differences between groups. After adjusting for multiple comparisons using the Bonferroni method, the only significant difference was as expected; stool frequency was greatest for IBS-D ( per day), intermediate for IBS-M ( per day), and least for IBS-C ( per day) (P.0001). Thus, after accounting for multiple comparisons, there were no meaningful differences at baseline across subtypes other than stool habit. Prospective Evaluation of Pain Scores and Stool Frequency/Consistency Data were obtained using 2-week daily diary cards at baseline, week 12 (end of treatment), and at 3, 6, 9, and 12 months after treatment. Pain scores. Using 14-day averaged visual analogue scale scores (scale of 0 100), there was no difference by group with regard to pain scores at baseline (IBS-D, ; IBS-M, ; IBS-C, ; P NS). However, there was a significant reduction in visual analogue scale pain scores from baseline to week 12 during the treatment period for all 3 groups (IBS-D, ; IBS-M, ; IBS-C, ; P.0001 from baseline to week 12 for all subtypes), and this low level was maintained over the next year. There were no differences between subtypes at any of the 6 periods of observation. Similar results were found using the McGill Pain Questionnaire. Stool frequency and consistency. Table 2 shows stool frequency (stools per day averaged over 14 days) and consistency (Bristol Stool Scale form where 7 indicates watery and 1 indicates hard pellets, averaged over 14 days) by subgroup for the 6 observation periods. Significant differences were seen across the 3 groups at baseline and between IBS-D and IBS-C and between IBS-D and IBS-M for all other periods (P.0001). Notably, the difference between IBS-C and IBS-M after baseline disappeared and was not significant for stool frequency and consistency. These findings suggest that the stool pattern in IBS-M is more similar to IBS-C than to IBS-D. Change from baseline to subsequent time points was measured within each treatment group for both stool frequency and consistency. Within the IBS-D group, the changes between baseline and 12 weeks were significantly different for both stool consistency (toward a normal stool) and frequency (decrease), and these initial changes were not significantly different from each other throughout the study, indicating that these initial changes were maintained. Stool consistency did show a significant change between 12 weeks and 3 months; however, from that point on, it remained constant (stool frequency: change from baseline was , and subsequent changes ranged between 0.25 and 0.11; stool consistency: change from baseline was , and subsequent changes ranged between 0.26 and 0.09). For the IBS-M group, the differences between baseline and 12 weeks were significant for stool frequency and consistency and then were not significantly different from these initial changes at all subsequent time points. No changes were significant throughout time within the IBS-C group. Treatment was evaluated in each model, and it was not responsible for any of these effects. Development of Subtypes Using Diary Cards Surrogate end points using daily diary cards were constructed to best fit the Rome II criteria as discussed in Patients and Methods and are shown in Table 3. Notably, the last 2 Rome II items (ie, straining and urgency) were not questions on stool pattern and were not included in the questionnaire. The best fit for the 4 remaining items on bowel habit was determined to be the most direct clinical comparisons with the Rome II questions (weighted, 0.60; P.001). Using these criteria (column 3 of Table 3), we were able to track subjects belonging to the IBS-D, IBS-M, or IBS-C groups and determine if they stayed within their category or shifted to either of the other 2 subtypes for the prospective periods of observation. Table 4 shows the response rate of subjects filling out the sets of questionnaires for each time period, for the entire group, and by subgroup. Because the number of

4 March 2005 ASSESSMENT OF BOWEL HABIT IN IBS IN WOMEN 583 Table 1. Comparison of Baseline Data Category Clinical Variable IBS-D (n 110) IBS-M (n 128) IBS-C (n 79) Overall (3-way) IBS-D vs IBS-M P value IBS-C vs IBS-M IBS-D vs IBS-C Visual analogue scale (2-wk average 36.2 (17.2) 38.7 (20.5) 35.7 (19.9).46 daily score) a McGill Pain Questionnaire (PRI) 12.8 (10.0) 13.4 (9.9) 12.1 (7.9).61 (2-wk average daily score) a Visual analogue scale pain value 33.8 (22.1) 38.1 (26.4) 36.0 (25.4).40 (single measure) a Stool frequency (2-wk average no. 3.1 (2.0) 2.1 (1.5) 1.2 (0.7) per day) a Stool consistency (2-wk average 2.5 (0.6) 2.3 (0.5) 2.3 (0.6) daily score as absolute difference from normal on Bristol Stool Scale) a Global well-being (2-wk average 2.9 (0.7) 2.7 (0.7) 2.8 (0.7).17 daily score) a Urgency one fourth of the time in 87.3 (96.0) 86.7 (111.0) 15.2 (12.0).0001 the past 3 mo b Straining one fourth of the time in the past 3 mo b 32.7 (36) 78.1 (100) 86.1 (68).0001 Physiology Tracking volume at urge threshold a (81.5) (79.7) (81.2).41 Tracking pressure at urge threshld a 23.5 (8.6) 24.5 (9.9) 23.8 (9.0).78 Volume at first report of pain a (80.7) (81.5) (94.9) Pressure at first report of pain a 25.5 (9.5) 28.0 (10.7) 28.3 (11.5).20 Baseline motility: motility index a 5.1 (2.2) 5.6 (4.9) 4.9 (2.4).48 Health care No. of treatments in the past 3 mo a 3.2 (3.9) 3.3 (5.9) 3.2 (3.4) (3.9) 2.1 (4.6) 2.0 (3.1).95 utilization No. of gastrointestinal treatments in the past 3 mo a No. of times telephoned physician 1.1 (2.0) 0.9 (1.7) 1.1 (2.9).77 in the past 3 mo a No. of times telephoned physician 0.8 (2.0) 0.5 (1.0) 0.7 (2.8).48 for gastrointestinal symptoms in the past 3 mo a No. of hospitalizations in the past (1.0) 0.3 (1.1) 0.4 (0.8).86 years a No. of days in bed for 4.6 (11.1) 3.3 (7.0) 2.9 (5.3).31 gastrointestinal symptoms in the past 3 mo a No. of days cut down on things 14.2 (20.6) 14.5 (23.0) 10.1 (18.5).30 because of gastrointestinal symptoms in the past 3 mo a No. of gastrointestinal procedures 0.7 (1.0) 0.7 (1.2) 0.5 (0.8).25 in the past 3 mo a No. of lifetime surgeries a 2.9 (3.4) 2.3 (2.6) 2.1 (2.1).15 No. of prescription gastrointestinal 0.7 (0.9) 0.8 (1.1) 1.0 (1.2).21 medications currently taking a Telephoned physician for a 50.9 (56.0) 34.4 (44.0) 38.0 (30.0) prescription or advice in the past 3mo b Have medical insurance b 80.0 (88.0) 77.3 (99.0) 77.2 (61.0) Seen counselor/social worker/ 9.1 (10.0) 7.8 (10.0) 10.1 (8.0) psychologist in the past 3 mo b Seen chiropractor/holistic or 26.4 (29.0) 19.5 (25.0) 32.9 (26.0) alternative practitioner in the past 3mo b Quality of Quality of life sum score a 65.1 (19.2) 62.5 (21.0) 70.3 (18.5) life SIP total score a 5.8 (6.5) 6.9 (7.8) 5.1 (5.1).17 PRI, Pain Rating Index; SIP, Sickness Impact Profile. a Mean (SD). b Percentage (n).

5 584 DROSSMAN ET AL GASTROENTEROLOGY Vol. 128, No. 3 Table 2. Comparison of Stool Frequency and Stool Consistency Over Time Mean (SD) P value IBS-D IBS-M IBS-C IBS-D vs IBS-M IBS-C vs IBS-M IBS-D vs IBS-C Stool frequency Baseline 3.1 (2.0) 2.1 (1.5) 1.2 (0.7) mo 2.4 (1.6) 1.7 (1.0) 1.1 (0.6) mo 2.7 (1.8) 1.7 (1.0) 1.2 (0.8) mo 2.9 (1.9) 1.6 (0.8) 1.3 (0.6) mo 2.7 (1.9) 1.7 (1.1) 1.2 (0.7) mo 2.7 (1.5) 1.6 (1.0) 1.4 (0.9) Stool consistency Baseline 5.0 (0.9) 3.9 (1.1) 3.2 (1.1) mo 4.5 (1.1) 3.5 (1.2) 3.1 (1.2) mo 4.8 (1.0) 3.6 (1.2) 3.3 (1.1) mo 4.6 (1.1) 3.6 (1.1) 3.1 (1.1) mo 4.6 (1.0) 3.8 (1.3) 3.2 (1.0) mo 4.7 (1.0) 3.6 (1.2) 3.2 (1.2) responses over time varied, we analyzed the subtypes 4 times, using each of the following 4 situations: women with 2, 3, 4, and 6 observations. There was little difference in the results of the analyses using different numbers of observations over time. Therefore, we performed the analyses using women who had at least 4 observations to retain a sufficient number to evaluate differences among subtypes. Of the 317 women evaluated at baseline, there were 163 retained at 12 months (51%). Of the 317 women completing the baseline questionnaires, 109 (34.4%) had data for all 6 periods, 190 (59.9%) had data for at least 4 periods, 218 (68.8%) had data for at least 3 periods, and 272 (85.8%) had data for at least 2 periods. Assessment of Subtypes Over Time Figure 1 shows a static representation of the number of subjects for each group at baseline and the number evaluated at the end of treatment (12 weeks) and at 3, 6, 9, and 12 months after treatment. Notably, there is little difference in the proportion of each subgroup for the periods of observation. Note that this analysis in Figure 1 does not account for the moving in and out from one subgroup to another, because it just presents the marginal distribution of subtypes at each static time period. Table 3. Prospective Tracking of Stool Pattern Rome II question 1. Fewer than 3 bowel movements per week for 25% of the time in the past 3 months 2. More than 3 bowel movements per day for 25% of the time in the past 3 months 3. Hard or lumpy stools for 25% of the time in the past 3 months 4. Loose (mushy) or watery stools for 25% of the time in the past 3 months 5. Straining during a bowel movement 25% time in the past 3 months 6. Urgency (having to rush to have a bowel movement) 25% of the time in the past 3 months Applicable diary card question 1. Total number of bowel movements in last 24 hours 1. Total number of bowel movements in last 24 hours 2. Please fill in the oval that best describes your bowel movements today (Bristol Stool Scale) 2. Please fill in the oval that best describes your bowel movements today (Bristol Stool Scale) None None Derived diary card question (selected was most direct clinical comparison ) Sensitivity Specificity At least 25% of days at 0 bowel movements per day At least 25% of days at 3 bowel movements per day At least 25% of days at Bristol Stool Scale score 1or2 At least 25% of days at Bristol Stool Scale score 6or7 None None

6 March 2005 ASSESSMENT OF BOWEL HABIT IN IBS IN WOMEN 585 Table 4. Response Rate by Time, Overall, and by Subgroup Baseline (n) 3 Months 6 Months 9 Months 12 Months 15 Months Total (253) 58.4 (185) 55.2 (175) 54.9 (174) 51.4 (163) IBS-D (62) 50.4 (57) 46.0 (52) 45.1 (51) 45.1 (51) IBS-M (92) 58.8 (57) 49.5 (48) 59.8 (58) 49.5 (48) IBS-C (99) 66.4 (71) 70.1 (75) 60.7 (65) 59.8 (64) NOTE. Results are expressed as percentage (n) unless otherwise indicated. Changes in Subgroup Classification Despite the similarity in subgroup proportions at each period of observation (Figure 1), there was in fact considerable variation and movement of individuals between subtypes over time. Of the 190 subjects classified by Rome II criteria and with at least 4 data points of observation, only 46 (24.2%) retained their baseline classification throughout the study period (Table 5). Of the remaining 75.8% of subjects who switched to another subgroup (ie, did not retain their baseline classification), 55.6% switched between IBS-D and IBS-M, and a larger proportion (68.8%) switched between IBS-C and IBS-M. Finally, of the 190 subjects, only 28.9% switched between IBS-D and IBS-C. Note that these are overlapping subtypes (ie, a subject may have switched between IBS-D and IBS-M and then between IBS-M and IBS-C, so a subject could then be counted in both subtypes). Using bubble plots, Figure 2 shows the degree of variation between groups at 2 points in time, from baseline to 3, 6, 9, and 12 months. Overall, with 2 points of observation, there was little change between subtypes. Most individuals stayed within their same subgroup, the next largest group involved a transition between IBS-D or IBS-C and IBS-M or vice versa, and the smallest group transitioned between IBS-C and IBS-D. These findings held in general for the 5 comparisons made between baseline and a later time period. Therefore, when comparing subtypes across 2 points in time, there was less change noted than when compared for the whole year with at least 4 points of observation (see Table 5). Between-Group Comparisons of Time to Change Figure 3 shows the survival curve for time to first change in subgroup for the 3 subtypes, each including women with at least 4 observations over time. The IBS-M subgroup was the least stable, with 50% changing within 3 months (12 weeks) after the end of treatment; the IBS-C subgroup had the longest time to the first change, with 50% changing at 9 months (36 weeks); and the IBS-D subgroup was intermediate, with 50% changing by 6 months (24 weeks). The change with IBS-M, the greatest over time, was significantly greater than IBS-C and IBS-D (log-rank , P.0001). There was no difference in change between IBS-D and IBS-C (P.24). We also note that the retention rate was the least for IBS-M, with only 5% retaining their category by the end of the study period, compared with retention of about 30% 35% for IBS-C and IBS-D (see Table 5). Individual Changes in Stool Pattern Over Time by Subgroup The ribbon plots in Figures 4 6 track changes in stool habit for each subgroup over time. The subjects were those in the placebo groups in the National Institutes of Health study who had at least 4 observation points, and missing periods of observation are shown as breaks in the ribbon. There is considerable individual variation in stool pattern over time. IBS-D at baseline. Of the 12 women starting with IBS-D, (1) 8, the largest proportion, ended in the Table 5. Those Who Remained in Original Classifications Figure 1. Proportion of cases of IBS-D, IBS-M, and IBS-C by 2-week diary cards. Data are shown at 6 time periods: at baseline before entering the treatment trial, at 12 weeks after completing treatment, and at 3, 6, 9, and 12 months after treatment. The proportions for all 3 subtypes remain approximately the same at all periods of assessment. n n (stayed in original classification) Percentage IBS-D IBS-M IBS-C Total

7 586 DROSSMAN ET AL GASTROENTEROLOGY Vol. 128, No. 3 Figure 2. Bubble plots showing change in stool subgroup at 2 points in time starting at baseline with 6 follow-up periods. For all plots, the y-axis shows the values at baseline and the x-axis shows the follow-up period (12 weeks after completing treatment and at 3, 6, 9, and 12 months). The circles are proportionate to the number for the particular category. The largest circles were seen for retention of the subgroup (IBS-D, IBS-M, and IBS-C) at both time points. The next largest groups were changes between IBS-D or IBS-C and IBS-M. Finally, the smallest group was transition between IBS-D and IBS-M. The data suggest that at 2 points of observation, regardless of how far the period is from baseline, there is only a small amount of transition. same category (IBS-D) and only 2 transitioned through IBS-M during the intermediate observation periods before going back to IBS-D; (2) another 3 transitioned through and ended with IBS-M; and (3) only one ended with IBS-C (Figure 4). IBS-M at baseline. Of the 18 women starting with IBS-M, (1) 10, the largest proportion, ended with Figure 4. Ribbon plot showing change in bowel subtypes for all subjects with IBS-D at baseline who were in the placebo groups of the National Institutes of Health treatment study. Most of the time was spent with IBS-D during the period of observation, and transitions were primarily to IBS-M and then back to IBS-D. On only 1 occasion did a subject transition to IBS-C. IBS-C and transitioned between IBS-C and IBS-M but only one transitioned through IBS-D; (2) 5 ended in the same category (IBS-M) and transitioned through both IBS-M and IBS-C during the intermediate observation periods; and (3) only 3 ended with IBS-D (Figure 5). IBS-C at baseline. Of the 23 women starting with IBS-C, (1) 13, the largest proportion, ended in the same category (IBS-C) and most of these stayed as IBS-C and a few transitioned to IBS-M and IBS-D during the intermediate observation periods; (2) 8 ended with IBS-M, with most transitions through IBS-M and only 3 transition points through IBS-D; and (3) only 2 ended Figure 3. Time to change classification. This is a life table analysis showing the time to change from IBS-D, IBS-M, and IBS-C to another category. It is noted that the earliest change occurs with IBS-M (50% change at 3 months). The longest subgroup to change is IBS-C (50% change at 9 months), and the IBS-D group is intermediate (50% change at 6 months). The change with IBS-M was greater than with IBS-C and IBS-D (log-rank ; P.0001), and there was no difference in change between IBS-D and IBS-C. Furthermore, the retention rate was also the least for IBS-M (5% retaining their category by study end), compared with retention of about 30% 35% for IBS-C and IBS-D. Figure 5. Ribbon plot showing change in bowel subtypes for all subjects with IBS-M at baseline who were in the placebo groups of the National Institutes of Health treatment study. Most subjects transitioned to IBS-C. Only 3 of 19 subjects transitioned to IBS-D.

8 March 2005 ASSESSMENT OF BOWEL HABIT IN IBS IN WOMEN 587 Figure 6. Ribbon plot showing change in bowel subtypes for all subjects with IBS-C at baseline who were in the placebo groups of the National Institutes of Health treatment study. Most subjects stayed at IBS-C or transitioned to IBS-M. Only 2 of 25 ended with IBS-D. with IBS-D, with transitions between IBS-M and IBS-C (Figure 6). Overall, tracking individual subjects over time shows that transition periods between IBS-C and IBS-M are more frequent than IBS-D and IBS-M, and it is more common for IBS-M to transition through IBS-C. Discussion IBS is defined as abdominal pain or discomfort associated with altered bowel habit (ie, diarrhea, constipation, or both). 1 3 In recent years, IBS has been subcategorized into either IBS-D or IBS-C using definitions from the Rome II criteria. 2,3 The remaining group with Rome II IBS (ie, non IBS-D, non IBS-C) is best considered as IBS-M, because this is a heterogeneous group consisting of individuals having symptoms of diarrhea and constipation but not meeting Rome II criteria for either diarrhea or constipation. Many clinicians have inappropriately classified these patients with IBS-M as having IBS-A. However, IBS-A is a unique group that can include individuals meeting criteria for IBS-D, IBS-M, or IBS-C (because these subtypes are defined at one point in time) if their bowel habit changes to the degree that it meets Rome criteria for a different group. Because IBS-A has not been defined in a consistent manner, it is difficult to compare studies in the literature. In some studies, the definitions for IBS-A are not reported 7 or, when they are, definitions vary from being IBS-M, to having a combination of one diarrhea and one constipation symptom at one point in time, 8,9 or preferentially to reports of symptoms changing from diarrhea to constipation or vice versa over a retrospectively defined time period, 10 although retrospective assessment is subject to bias. 11 Currently, no Rome II criteria exist for IBS-A, and no prospective assessment has been done for IBS-D, IBS-M, or IBS-C to determine the pattern and frequency of bowel habit change. This type of information is important to understand whether the clinical features of these subtypes differ from each other and whether the efficacy and safety for IBS drugs targeted specifically for patients with predominant diarrhea or constipation can be applied to IBS-A. This study is the first to prospectively assess change in bowel habit over a 1-year period using criteria consistent with Rome II definitions. We evaluated women with IBS enrolled in a multicenter treatment trial to clinically characterize the subtypes (IBS-D, IBS-M, and IBS-C). We developed surrogate definitions from diary cards derived from Rome II criteria and then tracked their bowel patterns prospectively every 3 months over 6 periods of assessment. Using this information, we propose a reasonable criterion for IBS-A (see below). There are several observations to be noted. First, at study baseline, except for stool frequency, after adjusting for multiple comparisons there are no significant differences among patients with IBS-D, IBS-M, and IBS-C in terms of clinical, physiologic, health care use, or quality-of-life data (Table 1). These results held when evaluating stool habit among the 3 subtypes over the 6 observation periods. Specifically, pain scores at all points in time were similar among the 3 subtypes and differences remained in stool pattern across each time period. However, after the baseline period, the significant differences in stool frequency and stool consistency distinguished IBS-D from IBS-C and IBS-M, but not between IBS-C and IBS-M. Second, the proportion of individuals in each of the subtypes does not change with multiple assessments. When looking at these subtypes over time (Figure 1), each group represents about one third of the total sample, a value that is consistent with the literature. 7,12 14 However, when tracking individual subjects over time, we note that most move in and out of the subtypes (Table 5 and Figures 3 6). Thus, for the first time, we have shown that individuals frequently transition to other subtypes but at rates that do not change the overall proportion for each subgroup over time. Furthermore, when the observation period is short and subjects are studied at only 2 points in time, subjects tend to retain their classification (Figure 2). However, when the analysis is for 4 periods of observation or more, more than three fourths (75.8%) transition out of their subtype after 1 year. This may explain why Mearin et al, 9 in the

9 588 DROSSMAN ET AL GASTROENTEROLOGY Vol. 128, No. 3 only other prospective assessment of bowel habit, found very little transition between subtypes. However, there were only 2 observation points over just a 1-month period. Third, we also provide evidence that the bowel patterns of IBS-C and IBS-M are more similar to each other than either to IBS-D. This is supported by evidence that mean stool frequency and consistency ratings are more similar between IBS-C and IBS-M, and both are different from IBS-D at various points in time (Table 2), and by evidence that subjects transition more frequently between IBS-C and IBS-M (68.8%) than between IBS-D and IBS-M (55.6%) (see also Figures 4 6). Additional evidence comes from a study 8 in which subjects with IBS-A (defined as having one diarrhea and one constipation symptom) considered themselves to be constipated. This observation may, however, relate primarily to women, because one survey of 662 patients with IBS in China found that IBS-C and IBS-A were more common in women, while IBS-D was more common in men. 15 Fourth, we note that the IBS-M group is the least stable of the subtypes, with a 50% rate of change to another subcategory at 12 weeks (Figure 3) and a retention rate in their original subgroup at the end of the study of only 4.9% (Table 5). This is significantly greater (P.0001 and P.0001, respectively) compared with IBS-D (50% change rate at 24 weeks and retention of 30.8%) and IBS-C (50% change rate at 36 weeks and retention of 35.9%). Finally, these data support a criterion for IBS-A that would include at least one change between IBS-D and IBS-C by Rome II criteria over a 1-year period. This recommended criterion is simple in concept, has face validity, is easy to place into question form, and is supported by our study where the proportion of IBS-A is 29% (ie, those who switched between IBS-D and IBS-C during this time frame). Also, for the purposes of defining an alternator, a transition between IBS-D and IBS-C is more practical than between either IBS-D and IBS-C with IBS-M, because the latter occur very frequently (about 60% of the time) and are complicated by the closer association between IBS-C and IBS-M than for IBS-D and IBS-M. Further work will be needed to determine the reliability and validity of this criterion and its understandability by patients. There are several considerations to note when interpreting the results of our study. First, because the Rome II criteria are based on retrospective assessment of stool habit, we could not directly apply these criteria to diary cards. Therefore, we needed to develop surrogate measures for change in bowel habit to prospectively classify subjects as having IBS-D, IBS-M, and IBS-C. Using diary card symptoms of stool frequency and consistency over 2-week periods, we were able to replicate the Rome II criteria with high reliability. Thus, for example, a stool frequency of at least 25% of days at 0 bowel movements per day by 2-week diary would match the Rome II criterion of 3 bowel movements per week for 25% of the time in the past 3 months (see Table 3 for all criteria). We believe that other investigators can use these diary criteria in future studies. Second, these findings may not necessarily apply to men, who were not evaluated in this study. Further data are needed to determine if these findings are sex specific or not. Third, because only a little more than one half of the subjects responded after 1 year, there may be some risk of dropout bias. However, with respect to stool habit, we do not believe there is an a priori reason that those who dropped out are any different than those who did not. In addition, we found little difference in the results of the analyses using different numbers of observations over time, even though there were more subjects in the earlier analyses. Therefore, we do not have any evidence for bias. To compensate for missing data at each observation period, we performed the analyses with 4 or more of the 6 observations to increase the evaluable number of subjects. Finally, because subjects were enrolled in a treatment trial, the first 2 time periods between baseline and end of treatment (12 weeks) may be confounded by the therapeutic effect of treatment, and indeed during this period there were significant improvements in pain scores and stool frequency. We believe that this effect relates to a regression to the mean with more symptomatic subjects entering the trial at baseline. Nevertheless, these changes occurred for all 3 bowel subtypes, and the values within each time period were not statistically different. It may be more appropriate to compare the actual values for the year after the first 3 months. Furthermore, the desipramine treatment arm was distributed equally over bowel subtypes, and additional analyses comparing the treatment arms did not show differences. Finally, 4 of the 6 observation periods occurred after treatment. In conclusion, in this long-term assessment of bowel habit in women with IBS, there are few clinical differences between IBS-D, IBS-M, and IBS-C except for stool habit, and IBS-M and IBS-C are more closely related than IBS-D in their stool patterns over time. While the proportions of subjects in each of these 3 subtypes remain the same over multiple periods of observation, most individuals (about three fourths) switch out of their baseline subtype. However, a full switch between IBS-D and IBS-C occurred only 29% of the time. Based on these data, we recommend that the criterion for Rome II

10 March 2005 ASSESSMENT OF BOWEL HABIT IN IBS IN WOMEN 589 IBS-A be at least one change between IBS-D and IBS-C by Rome II criteria over a 1-year period. This can be asked by retrospective questionnaire by having the criteria for IBS-D and IBS-C fulfilled in the preceding year. Further studies will be needed to characterize IBS-A relative to those with IBS-C and IBS-D who retain their symptom features. References 1. Drossman DA, Camilleri M, Mayer EA, Whitehead WE. AGA technical review on irritable bowel syndrome. Gastroenterology 2002; 123: Thompson WG, Longstreth GF, Drossman DA, Heaton KW, Irvine EJ, Mueller-Lissner SA. C. Functional bowel disorders and D. functional abdominal pain. In: Drossman DA, Talley NJ, Thompson WG, Whitehead WE, Corazziari E, eds. Rome II: functional gastrointestinal disorders: diagnosis, pathophysiology, and treatment. 2nd ed. McLean, VA: Degnon Associates, Inc, 2000: Thompson WG, Longstreth GF, Drossman DA, Heaton KW, Irvine EJ, Mueller-Lissner SA. Functional bowel disorders and functional abdominal pain. Gut 1999;45:II43 II Drossman DA, Toner BB, Whitehead WE, Diamant NE, Dalton CB, Duncan S, Emmott S, Proffitt V, Akman D, Frusciante K, Le T, Meyer K, Bradshaw B, Mikula K, Morris CB, Blackman CJ, Hu Y, Jia H, Li JZ, Koch GG, Bangdiwala SI. Cognitive-behavioral therapy versus education and desipramine versus placebo for moderate to severe functional bowel disorders. Gastroenterology 2003; 125: Melzack R. The McGill Pain Questionnaire: major properties and scoring methods. Pain 1975;1: Heaton KW, Gosh S. Relation between stool form on a seven point scale and symptoms of urgency, straining, and incomplete evacuation: a new way of looking at irritable bowel syndrome (abstr). Gut 1989;30:A Coffin B, Dapoigny M, Cloarec D, Comet D, Dyard F. Relationship between severity of symptoms and quality of life in 858 patients with irritable bowel syndrome. Gastroenterol Clin Biol 2004; 28: Mearin F, Balboa A, Badia X, Baro E, Caldwell E, Cucala M, Diaz-Rubio M, Fueyo A, Ponce J, Roset M, Talley NJ. Irritable bowel syndrome subtypes according to bowel habit: revisiting the alternating subtype. Eur J Gastroenterol Hepatol 2003;15: Mearin F, Baro E, Roset M, Badia X, Zarate N, Perez I. Clinical patterns over time in irritable bowel syndrome: symptom instability and severity variability. Am J Gastroenterol 2004;99: Whitehead WE, Palsson OS, Levy RL, Feld AD, Von Korff M, Turner MJ, et al. Identification of irritable bowel (IBS) patients with alternating bowel habits. Gastroenterology 2004;126(Suppl 2): A Manning AP, Wyman JB, Heaton KW. How trustworthy are bowel histories? Comparison of recalled and recorded information. Br Med J 1976;2: Wilson S, Roberts L, Roalfe A, Bridge P, Singh S. Prevalence of irritable bowel syndrome: a community survey. Br J Gen Pract 2004;54: Lembo T, Fullerton S, Diehl D, Raeen H, Munakata J, Naliboff B, Mayer EA. Symptom duration in patients with irritable bowel syndrome. Am J Gastroenterol 1996;91: Talley NJ, Zinsmeister AR, Melton LJ III. Irritable bowel syndrome in a community: symptom subgroups, risk factors, and health care utilization. Am J Epidemiol 1995;142: Si JM, Wang LJ, Chen SJ, Sun LM, Dai N. Irritable bowel syndrome consulters in Zhejiang province: the symptoms pattern, predominant bowel habit subgroups and quality of life. World J Gastroenterol 2004;10: Received October 7, Accepted December 2, Address requests for reprints to: Douglas A. Drossman, MD, UNC Center for Functional GI and Motility Disorders, University of North Carolina at Chapel Hill, Room 1110 Burnett-Womack CB#7080, Chapel Hill, North Carolina Fax: (919) Supported by National Institutes of Health grant RO1DK49334 and Novartis Pharmaceuticals.