It is important for clinicians to be able to record both objective and subjective outcome measures and demonstrate the efficacy of their treatment. Simple visual analogue scales (VAS) have been used to assess patients' perception of severity of urinary symptoms (Frazer et al 1987) and urinary loss (Frazer et al 1989), but these have correlated poorly with objective measures such as cystometry or the two hour pad test. Simons and colleagues (1999) noted that a simple VAS measuring the impact of leakage on lifestyle correlated poorly with two disease specific Quality of Life tests ie IIQ (Kendall's rank correlation, tau=0.278, p=0.001) and UDI (tau =0.252, p=0.004) and with a single one hour pad test (tau =0.154, p=0.03). One reason for this inadequate correlation could be that some patients, particularly the elderly, find the VAS difficult to understand. Moreover, urinary incontinence is such a multi faceted problem that a single VAS may not be sufficiently sensitive. The reliability of the VAS may be improved by increasing the number of items on the scale (Streiner and Norman 1995). Current urinary incontinence scoring systems which employ Likert type scales, such as the Lago-Janssen, tend to be restrictive in their nature. We wished to devise a robust assessment system which would allow patients latitude in expressing the severity of their incontinence. We have designed a composite VAS, comprising four 10 cms items (assessing the frequency of episodes of stress and urge leaks, the number of pads used and the impact on lifestyle disturbance) based on a faecal incontinence scoring system (Jagro and Wexner 1996). We sought to investigate the following : 1) the test retest reliability of the composite VAS 2) the construct validity of the composite VAS by correlating it to the one hour pad test (mean and largest).

Methods
New referrals to the Pelvic Floor Unit were requested to fill in the VAS questionnaires at their initial visit and one week thereafter. All questionnaires were completed prior to commencement of treatment. Women complaining of significant incontinence also had two one hour pad tests performed one week apart. The questionnaire consisted of four 10 cms VAS assessing the patients' perception of the frequency of episodes of stress and urge incontinence, the number of pads used and impact of incontinence upon lifestyle. The total of the individual four components was added and the composite VAS was expressed as a percentage. Construct validity of the VAS was assessed by correlating the composite scores to the mean of two one hour pad tests and to the larger of the two one hour pad tests. The test retest reliability of the VAS was assessed by the method as described by Bland and Altman (1986), where to achieve statistical repeatability, 95% of the difference between composite scores have to lie within 2 standard deviations (sd) of the mean difference. Spearman's correlation coefficient was used to examine the relationship between the composite VAS and the pad tests. Wilcoxon's signed rank test was used to compare data from the first and second composite VAS.

Results
Forty six women completed two consecutive VAS questionnaires one week apart and thirty two of them had two one hour pad tests performed. The mean difference between the first and second composite VAS was -4.8% with s.d. of 12.03 (median -2, IQR -10-6.5). As per Bland and Altman, 95% of the differences lay within 2 s.d.'s of the mean difference between the composite VAS (figure 1). The limits of agreement was -28.9% to 19.3%, ie the second composite VAS could be 28.9% lower or 19.3% higher than the first. The test retest reliability of the four individual VAS components (/10 cms) was also assessed : a) the mean difference in stress leaks was 0.054/10 (s.d. 1.81), limits of agreement -3.55 to 3.654. b) the mean difference in urge leaks was 0.03/10 (s.d. 3.07), limits of agreement -6.11 to +6.17. c) the mean difference in pad usage was -1.7/10 (s.d. 4.9), limits of agreement -11.5 to +8.1. d) the mean difference in impact of incontinence on lifestyle was 0.44/10 (s.d. 2.09), limits of agreement -3.74 to +4.62. The scores of the individual components (/10 cms) and composite VAS (%) did not differ significantly between the first and second VAS questionnaires, Wilcoxon signed rank test, p<0.05. Construct validity for the composite VAS was good as evidenced by the significantly positive correlation between the composite VAS and the mean of the two pad tests (Spearman's rank correlation, r=0.74, p<0.0001) and the larger of the two pad tests (r=0.74, p<0.0001).

Figure 1 : Bland Altman test for reliability

Conclusion
The composite VAS has statistically satisfactory test retest reliability, but is clinically unacceptable as the second VAS may differ from the first by a factor of up to 30%. The individual components of the VAS also had poor clinical reproducibility from the first to the second test (particularly pad usage, which may vary by 100%). The construct validity of the composite VAS with regards to the mean of two pad tests and the larger of the two pad tests (r=0.74 and 0.73 respectively, p<0.0001) was good. Previous studies have shown poor correlation with a single one hour pad test and the fact that we demonstrated good correlation when using data from two pad tests highlights the inadequacy of using a single one hour test to assess the severity of incontinence. The composite VAS has good construct validity and would be useful as a baseline measure of patients' perception of severity of urinary incontinence. However, its' sensitivity is insufficient for it to used as an outcome measure for treatment.

References
Frazer MI, Sutherst JR and Holland EFN (1987) B Med J 35: 582
Frazer MI, Haylen BT and Sutherst JR (1989) Br J Obstet Gynaecol 63 : 14 -15
Jorge JMN, Wexner SD (1993) Dis Colon Rectum 36: 77-97
Simons AM et al Proc ICS Neurourol Urodyn 1999 (in press)
Streiner DL, Norman GR. Health Measurement Scales. A practical guide to their development and use. Second edition 1995. Oxford Medical Publications. Chapter 4 : Scaling Responses; pp 28-53.