Post a new comment on this article
Post Your Discussion Comment
Please follow our guidelines for comments and review our competing interests policy. Comments that do not conform to our guidelines will be promptly removed and the user account disabled. The following must be avoided:
- Remarks that could be interpreted as allegations of misconduct
- Unsupported assertions or statements
- Inflammatory or insulting language
Why should this posting be reviewed?
See also Guidelines for Comments and Corrections.
Thank you for taking the time to flag this posting; we review flagged postings on a regular basis.close
Could the authors give the raw (non-normalized) SF-36 subscale scores
Posted by tkindlon on 05 Nov 2011 at 14:49 GMT
The eight SF-36 subscale scores in the Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) field have generally been given in the past as raw scores (the PCS and MCS are given as normalised scores). The figures in Table 4 look to me like the eight subscales have been normalised. This makes comparisons with most of the studies in the ME/CFS field difficult especially when different population norms exist for different countries. Also, it means many may interpret them as raw scores when as I say they appear to me to be normalised scores.
Assuming I am correct and these are normalised scores, could the authors give the eight subscale scores as raw (i.e. unadjusted scores). Alternatively, could the authors give the data on which population norms were used e.g. was it a Norwegian sample or a US sample (for example).
The figures look much more impressive if the figures in Table 4 are normalised scores e.g. if one looks at the physical function subscale, with a baseline mean of 34 and a mean max change of 39%. A mean max change of 39% is 47.26 (although this may not be exactly correct as there is rounding). This is very close to 50 i.e. the population norm so looks much more impressive than if it was 47 as a raw score.
RE: Could the authors give the raw (non-normalized) SF-36 subscale scores
We thank dr. Kindlon for his remark to Table 4 in our article describing SF-36 scores, and for the opportunity to clarify aspects of the analyses to help the readers interpret the data.
The patients recorded the SF-36 short form scheme (Norwegian translation) at baseline (pre-treatment) and every month until 10 months follow-up.
The analyses of SF-36 short forms, with physical health summary score, mental health summary score, and scores for the eight SF-36 subdimensions, were generated from a SPSS syntax file.
Dr. Kindlon is correct that the values for the SF-36 subdimentions are norm-based, and the population norm from US 1998 was used.
The table presents data on SF-36 scores for physical health summary score, mental health summary score, and scores for the eight SF-36 subdimensions, and shows baseline levels (mean with SD) for both the Rituximab and Placebo groups. Because the time-frame for responses vary among patients, we chose also to present the mean values of maximum changes in SF-scores during follow-up, as compared to baseline.
We think the SF-36 data supports the main findings of the study, showing a significant difference in maximum changes as compared to baseline, in favour of the Rituximab group, for physical health summary score, and the subdimentions physical function and bodily pain. We should take caution in comparing these data to other CFS/ME studies, and we think the most important aspect is the comparison between the two intervention groups in our study.
RE: RE: Could the authors give the raw (non-normalized) SF-36 subscale scores
Thank you for clarifying that Table 4 uses SF-36 norm-based scores. Could the authors also confirm if the maximum change percentages are per cent changes to the norm-scores or to the raw (0-100) scores?
Ideally, could the authors also provide mean "maximum" SF-36 scores (I don't think they can be calculated from the data given in Table 4) for both Rituximab and placebo groups? This aditional information would give readers the fullest picture of the size of the effect from Rituximab.
RE: RE: RE: Could the authors give the raw (non-normalized) SF-36 subscale scores
The analyses of the SF-36 short forms were generated from a SPSS syntax file, with values norm-based using the population norm from US 1998.
For all dimensions (physical health summary score, mental health summary score, and the eight subdimensions), the baseline values (before intervention) and values generated from SF-36 recorded every month until 10 months after intervention, were calculated the same way using the same syntax file.
Due to different time frames for clinical responses, we chose to calculate in each patient the maximum changes during follow-up (for each dimension), as compared to the baseline level for the actual dimension.
In addition to the baseline levels for each dimension presented as mean with standard deviation in each group, we also the present the maximum changes during follow-up as compared to baseline (in percent), and demonstrated as mean with standard deviation for each group (Rituximab and Placebo). These maximum changes during follow-up therefore are not representative for the complete follow-up period, but reflect the status at the time the patients are feeling at the best (or worst) i.e. with maximum difference from baseline.
As an answer to the question, the maximum changes presented are percent changes from the baseline scores, which were norm-based.
For patient ID4 in the Rituximab group, the scores for mental health summary score during the study were: baseline: 52.9, month 1: 52.9, month 2: 54.0, month 3: 56.9, month 4: 57.5, month 5: 56.9, month 6: 52.4, month 7: 53.4, month 8: 52.9, month 9: 52.9, month 10: 51.2.
Thus in this patient the maximum change in mental health summary score during follow-up was 9%. (Score 57.5 (at 8 months) – score 52.9 (baseline)/ score 52.9 (baseline)=0.09).
In Table 4, the mean values for mental health summary score at baseline were 46 in the rituximab group, and 46 in the placebo group.
When calculating the maximum difference from baseline during follow-up for all patients, for mental health summary score, the mean values were 9% (SD 54) in the rituximab group, and 5% (SD 32) in the placebo group, which was not a significant difference between the two groups.
For patient ID19 in the rituximab group, the scores for the subdimension bodily pain during follow-up were: baseline: 28.7, month 1: 28.7, month 2: 36.6, month 3:36.6, month 4: 50.3, month 5: 50.3, month 6: 36.6, month 7: 36.6, month 8: 40.7, month 9: 32.8, month 10: 28.7.
Thus in this patient the maximum change in bodily pain score during follow-up was 75%. (Score 50.3 (at 8 months) – score 28.7 (baseline)/ score 28.7 (baseline)=0.75)
In Table 4, the mean values for bodily pain score at baseline were 32 in the rituximab group, and 34 in the placebo group.
When calculating the maximum difference from baseline during follow-up for all patients, for bodily pain, the mean values (for maximum difference) were 40% (SD 31) in the rituximab group, and 8% (SD 24) in the placebo group, which was a significant difference between the two groups.
RE: RE: RE: RE: Could the authors give the raw (non-normalized) SF-36 subscale scores
Thank you for the comprehensive explanation of how maximum change percentages were calculated from the norm-scores for individual patients.
For those not familiar with norm-based scoring, it's worth pointing out that the percentage increase in the underlying raw (0-100) scores will be much bigger than the increase for norm-scores.
Take the SF-36 Physical Function (PF) results as an example. The norm-score Rituximab group baseline was 34 with a mean maximum increase of 39%, giving an estimated mean maximum norm-score of 47.3. However, the equivalent raw scores for these figures are: baseline 44.9 , maximum 76.5 - an increase in the raw score of 70% or 31.6 points (i.e. much higher than the 39% increase in norm-scores).
The 31.6 points raw score gain for the Rituximab group in this example compares with a gain of 9.2 points (19%) in the Placebo group on the same basis. This represents an exceptional relative peak gain in physical function for the Rituximab group.
SF-36 Physical Function scores are of particular interest as substantial impairment is a defining feature of CFS. Although in this study SF-36 PF scores were one of a number of secondary outcomes, other studies - including the two largest clinical trials to date (1,2) - have used them as a primary outcome, albeit using endpoint rather than peak scores.
1. Wearden AJ, Dowrick C, Chew-Graham C, et al on behalf of the PACE trial management group Fatigue Intervention by Nurses Evaluation (FINE) trial writing group and the FINE trial group. Nurse led, home based self help treatment for patients in primary care with chronic fatigue syndrome: randomised controlled trial. BMJ. 2010;340:c1777.
2. White PD, Goldsmith KA, Johnson AL, et al on behalf of the PACE trial management group. Comparison of adaptive pacing therapy, cognitive behaviour therapy, graded exercise therapy, and specialist medical care for chronic fatigue syndrome (PACE): a randomised trial. Lancet. 2011;377(9768):823-36.
SF-36 PF estimate if non-responders only scored the same as the placebo group
Thanks to SMcGrath for posting that information.
Note that 76.5 would be a lower bound* for the SF-36 PF scores for the responders.
We only have SF-36 PF scores for 13 of the CFS patients on Rituximab, 9 of whom were responders.
If the non-responders only increased at peak by the same amount as the placebo group i.e. 9.2, then the 9 responders would have reached, on average, a very impressive 86.5** on the SF-36 PF.
*Approximately as these calculations depend on average percentage increases - what a particular percentage increase translates to depends on the initial value. But we don't have added information to make any assumptions other than the individual differences would balance out when the mean was obtained.
**Calculation (using SMcGrath's figures): Total increase among the 13 on Rituximab: 31.6*13=410.8 points. Amount of this obtained by the four non-responders, if the response the same as the placebo group: 9.2*4=36.8.
Therefore, improvement in responders compared to baseline: (410.8-36.8)/9=41.6.
Or mean score at peak = baseline score + improvement = 44.9+41.6=86.5
It took me a little while but I have now found a place giving the population norms in raw figures: http://www.sf-36.org/nbsc... (and type in a figure). There are probably other places but this does the job:
Non normalised USA 1998 population means (SDs):
PF: 83.0 (23.8)
RP: 77.9 (35.3)
BP: 70.2 (23.4)
GH: 70.1 (21.4)
VT: 57.0 (21.1)
SF: 83.6 (23.0)
RE: 83.1 (31.6)
MH: 75.2 (17.6)