Objectives
The primary objective of the study is to evaluate changes in clinical
pain levels among subjects with fibromyalgia using the Defense Veterans
Pain Rating Scale (DVPRS) at baseline, 6 weeks, and 12 weeks following
treatment in both the SHAM and TRUE groups.
The secondary objective is to assess the device’s effectiveness in
improving sleep quality, sleep depth, and sleep restoration in both
groups.
Additionally, the study aims to examine the impact of alcohol use,
medication use, and major life events on pain scores or sleep
quality.
Method
In this study, we plan to present both the results of an
intent-to-treat analysis and a per-protocol analysis.
- Per-protocol Analysis is used for only including participants who
followed the study protocol.
- CES13 and CES42 were ineligible upon initial screening
- Intention-to-treat analysis involves linear mixed-effects model,
accounting for missing data under a missing at random assumption. Linear
mixed-effects model allows us to use all available data without
excluding participants who drop out at later time points. In this way,
we can handle missing follow-up data well when at least some outcome
data is available. Although mixed model analysis is appropriate for
longitudinal data with missing outcomes, cases with only baseline values
are excluded from the analysis since they lack follow-up outcome
measurements.
For CES13, CES25 and CES42 with notes “Ineligibility”, they were not
randomized to either groups since they were screened out. For CES45, the
subject should be classified as a dropout at baseline because of lost to
follow-up. For CES17, the subject was also classified as a baseline
dropout with the reason noted as “Withdrew.”
Inclusion Criteria Flowchart
## Linking to librsvg 2.56.3
Data Quality Check
To validate data integrity and identify potential issues, we
conducted an initial inspection of the dataset CES_2 using
various diagnostic techniques:
Structural Examination: The dataset’s column names and structure
were reviewed to verify the presence and correctness of variables by
using colnames(CES_2) and str(CES_2).
Summary Statistics: Basic descriptive statistics were computed to
check for potential anomalies, outliers, and data distribution by using
summary(CES_2).
Missing Values Diagnostics: The count of missing values per
column was assessed to determine the extent of missingness and inform
imputation strategies.
Data Cleaning and Preprocessing
To prepare the dataset for analysis, we need to implement data
cleaning and preprocessing steps to ensure accuracy and consistency.
Variable Labeling and Transformation: We converted
Group into a numeric variable (SHAM = 0, TRUE = 1).
Additionally, we extracted numeric IDs from the ID variable and
converted them into integers (CES01 -> 1).
Handling Missing Values: We used forward fill to populate missing
values in ID, Age, Sex (1-f, 2-m)
and Ethnicity. We also ensured DVPRS was numeric by
replacing “NA” values (character string) with NA and converting the
column to numeric.
Filtering Dropout Subjects: We retained only participants who
completed the study at three time points. For per-protocol Analysis, we
excluded participants who dropped out before receiving any treatment.
For intention-to-treatment analysis, we conducted linear mixed-effects
model to use all available data without excluding participants who drop
out at later time points. However, the cases with only baseline values
are excluded from the analysis since they lack follow-up outcome
measurements.
Usage Time Conversion: We transformed the “HH:MM” format in
Usage Time into a numeric variable (Usage_Time_hrs), and
forward-filled missing values within Usage_Time_hrs for
consistency. For the participant CES48, we replaced “NO
POWER” with “37:00” before converting usage time.
Computing DVPRS Differences: We created two new variables, DVPRS6
and DVPRS12. DVPRS6 represents the change in DVPRS pain scores between
baseline and 6 weeks, and DVPRS12 represents the change in DVPRS pain
scores between baseline and 12 weeks.
Table 1: Baseline Demographic Table for Intention-to-Treat
Analysis
SHAM (N=23) vs. TRUE (N=24)
|
SHAM
(N=23) |
TRUE
(N=24) |
P-value |
| Age |
|
|
|
| Mean (SD) |
49.4 (7.92) |
49.4 (7.07) |
0.978 |
| Median [Min, Max] |
49.0 [33.0, 60.0] |
51.5 [34.0, 57.0] |
|
| Ethnicity |
|
|
|
| AA |
15 (65.2%) |
16 (66.7%) |
0.513 |
| W |
8 (34.8%) |
6 (25.0%) |
|
| AA (mixed) |
0 (0%) |
1 (4.2%) |
|
| declined to answer |
0 (0%) |
1 (4.2%) |
|
| Sex (1-f, 2-m) |
|
|
|
| 1 |
18 (78.3%) |
15 (62.5%) |
0.389 |
| 2 |
5 (21.7%) |
9 (37.5%) |
|
| Defense and Veterans Pain Rating Scale (Baseline) |
|
|
|
| Mean (SD) |
6.85 (1.67) |
6.60 (1.31) |
0.582 |
| Median [Min, Max] |
7.00 [2.00, 10.0] |
6.50 [4.00, 10.0] |
|
| Widespread Pain Index |
|
|
|
| Mean (SD) |
13.0 (3.27) |
11.8 (3.19) |
0.192 |
| Median [Min, Max] |
13.0 [7.00, 20.0] |
11.0 [7.00, 19.0] |
|
| Missing |
0 (0%) |
1 (4.2%) |
|
| Symptom Severity Score |
|
|
|
| Mean (SD) |
8.96 (1.89) |
8.48 (1.65) |
0.366 |
| Median [Min, Max] |
9.00 [5.00, 12.0] |
8.00 [5.00, 11.0] |
|
| Missing |
0 (0%) |
1 (4.2%) |
|
| Alcohol Screener(1-yes, 2-no) |
|
|
|
| 1 |
11 (47.8%) |
10 (41.7%) |
0.896 |
| 2 |
12 (52.2%) |
14 (58.3%) |
|
| Baseline Bicep Curls - Left |
|
|
|
| Mean (SD) |
13.6 (6.37) |
15.0 (7.14) |
0.458 |
| Median [Min, Max] |
13.0 [4.00, 27.0] |
14.5 [0, 34.0] |
|
| Baseline Bicep Curls - Right |
|
|
|
| Mean (SD) |
16.7 (6.58) |
16.4 (8.26) |
0.899 |
| Median [Min, Max] |
17.0 [5.00, 29.0] |
15.0 [3.00, 39.0] |
|
| 30s Chair Stand Test |
|
|
|
| Mean (SD) |
7.91 (3.38) |
7.75 (4.26) |
0.885 |
| Median [Min, Max] |
7.00 [3.00, 15.0] |
7.00 [2.00, 18.0] |
|
| Baseline Average Handgrip Strength - Left |
|
|
|
| Mean (SD) |
16.3 (12.5) |
18.8 (9.43) |
0.453 |
| Median [Min, Max] |
19.7 [0, 46.7] |
18.5 [1.00, 34.7] |
|
| Baseline Average Handgrip Strength - Right |
|
|
|
| Mean (SD) |
21.6 (13.3) |
18.7 (10.6) |
0.403 |
| Median [Min, Max] |
22.7 [0.333, 52.0] |
18.0 [1.33, 38.0] |
|
| Requested Device After Study |
|
|
|
| no |
9 (39.1%) |
8 (33.3%) |
1 |
| yes |
12 (52.2%) |
12 (50.0%) |
|
| Missing |
2 (8.7%) |
4 (16.7%) |
|
| Device Perception |
|
|
|
| Real |
8 (34.8%) |
16 (66.7%) |
0.044 |
| Fake |
6 (26.1%) |
1 (4.2%) |
|
| Unsure |
5 (21.7%) |
4 (16.7%) |
|
| Missing |
4 (17.4%) |
3 (12.5%) |
|
| Usage Time (hours) |
|
|
|
| Mean (SD) |
47.3 (23.8) |
47.1 (33.1) |
0.982 |
| Median [Min, Max] |
42.4 [7.23, 86.6] |
64.2 [0, 83.7] |
|
| Sleep T-score |
|
|
|
| Mean (SD) |
62.7 (7.03) |
61.4 (5.37) |
0.541 |
| Median [Min, Max] |
62.0 [51.3, 77.5] |
61.0 [50.2, 72.0] |
|
| Missing |
7 (30.4%) |
7 (29.2%) |
|
Table 2: Dropout Timepoints and Ineligible Subjects
| 1 |
SHAM |
17 |
| 1 |
TRUE |
45 |
| 2 |
SHAM |
10, 30 |
| 2 |
TRUE |
22, 27, 31, 34, 37 |
| 3 |
SHAM |
3, 9, 26, 28 |
| 3 |
TRUE |
1 |
| Ineligible subjects |
NA |
13, 25, 42 |
Table 3: Patient Disposition After Dropout – ITT and PP
Populations
Table 3: Patient Disposition After Dropout – ITT and PP Populations
|
|
Dropout Summary
|
|
Analysis
|
Group
|
Final.Patient.Count
|
Ineligible.Subject.ID
|
IDs.Dropout.at.Baseline.T1
|
IDs.Dropout.at.Week.6.T2
|
IDs.Dropout.at.Week.12.T3
|
|
ITT
|
SHAM
|
23
|
13, 42
|
|
|
|
|
ITT
|
TRUE
|
24
|
25
|
|
|
|
|
ITT
|
Total
|
47
|
|
|
|
|
|
PP
|
SHAM
|
16
|
|
17
|
10, 30
|
3, 9, 26, 28
|
|
PP
|
TRUE
|
17
|
|
45
|
22, 27, 31, 34, 37
|
1
|
|
PP
|
Total
|
33
|
|
|
|
|
|
Note:
|
ITT: Intent-to-Treat analysis; PP: Per-Protocol analysis.
Dropout numbers reflect patients lost to follow-up, adverse events,
or withdrawal.
CES16 remains in the ITT analysis as TRUE, and
reassigns to the SHAM group in the PP analysis.
|
Results
Primary Outcome: Pain Reduction
To conduct the model selection, we did backward-selection which
includes all potential predictor variables and eliminate one variable at
a time until only variables with statistically significant p-value
remain. Since age, sex, ethnicity, usage time, interaction for timepoint
and usage time, alcohol screener and medication did not show the
significant effect on DVPRS, we excluded them from the model and only
major life event and timepoint remained in the final model.
The model results indicate a significant decrease in DVPRS scores
over time. Compared to Timepoint 1, the DVPRS score is, on average, 1.47
points lower at Timepoint 2 (p = 3.85 × 10⁻⁵) and 1.88 points lower at
Timepoint 3 (p = 5.59 × 10⁻⁷), suggesting a consistent reduction in
reported pain levels over time. Additionally, experiencing a major life
event is associated with a 1.28-point increase in the DVPRS score, with
a statistically significant effect (p = 0.0482), indicating a potential
association between major life events and higher pain perception.
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: DVPRS ~ factor(Timepoint) + major_life_event + (1 | ID)
## Data: complete_data
##
## REML criterion at convergence: 376.3
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.81361 -0.57307 0.09169 0.56529 2.15433
##
## Random effects:
## Groups Name Variance Std.Dev.
## ID (Intercept) 1.520 1.233
## Residual 1.763 1.328
## Number of obs: 99, groups: ID, 33
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 6.9091 0.3154 66.7065 21.905 < 2e-16 ***
## factor(Timepoint)2 -1.4059 0.3360 65.6993 -4.184 8.67e-05 ***
## factor(Timepoint)3 -1.8105 0.3411 66.8325 -5.308 1.36e-06 ***
## major_life_event 1.3490 0.6432 94.8874 2.097 0.0386 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) fc(T)2 fc(T)3
## fctr(Tmpn)2 -0.504
## fctr(Tmpn)3 -0.497 0.532
## majr_lf_vnt 0.000 -0.232 -0.286
## 2.5 % 97.5 %
## .sig01 0.83 1.69
## .sigma 1.11 1.56
## (Intercept) 6.29 7.52
## factor(Timepoint)2 -2.06 -0.75
## factor(Timepoint)3 -2.47 -1.15
## major_life_event 0.10 2.60
Table 4: Linear Mixed Model Results
|
term
|
estimate
|
std.error
|
statistic
|
p.value
|
confidence_interval
|
|
(Intercept)
|
6.91
|
0.32
|
21.90
|
<0.001
|
[6.28, 7.54]
|
|
factor(Timepoint)2
|
-1.41
|
0.34
|
-4.18
|
<0.001
|
[-2.08, -0.74]
|
|
factor(Timepoint)3
|
-1.81
|
0.34
|
-5.31
|
<0.001
|
[-2.49, -1.13]
|
|
major_life_event
|
1.35
|
0.64
|
2.10
|
0.04
|
[0.07, 2.63]
|
|
sd__(Intercept)
|
1.23
|
NA
|
NA
|
NA
|
[NA, NA]
|
|
sd__Observation
|
1.33
|
NA
|
NA
|
NA
|
[NA, NA]
|
# DVPRS
# To determine whether the treatment is better than the sham
model_treatment <- lmer(DVPRS ~ factor(Timepoint) * Group + (1|ID), data=complete_data)
summary(model_treatment)
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: DVPRS ~ factor(Timepoint) * Group + (1 | ID)
## Data: complete_data
##
## REML criterion at convergence: 377.8
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.76498 -0.53282 0.03342 0.54329 2.17019
##
## Random effects:
## Groups Name Variance Std.Dev.
## ID (Intercept) 1.651 1.285
## Residual 1.836 1.355
## Number of obs: 99, groups: ID, 33
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 7.1250 0.4668 64.2142 15.262 < 2e-16 ***
## factor(Timepoint)2 -1.6250 0.4791 62.0000 -3.392 0.001213 **
## factor(Timepoint)3 -1.9375 0.4791 62.0000 -4.044 0.000148 ***
## GroupTRUE -0.4191 0.6504 64.2142 -0.644 0.521638
## factor(Timepoint)2:GroupTRUE 0.7426 0.6675 62.0000 1.113 0.270183
## factor(Timepoint)3:GroupTRUE 0.6434 0.6675 62.0000 0.964 0.338855
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) fc(T)2 fc(T)3 GrTRUE f(T)2:
## fctr(Tmpn)2 -0.513
## fctr(Tmpn)3 -0.513 0.500
## GroupTRUE -0.718 0.368 0.368
## f(T)2:GTRUE 0.368 -0.718 -0.359 -0.513
## f(T)3:GTRUE 0.368 -0.359 -0.718 -0.513 0.500
Based on the linear mixed model results, the treatment (TRUE) does
not appear to be significantly better than the SHAM. At Timepoint 2,
pain is 1.62 points lower compared to Timepoint 1 (p=0.0012). At
Timepoint 3, pain is 1.94 points lower compared to Timepoint 1
(p=0.0001). This suggests that there is a statistically significant
reduction in DVPRS (pain scores) over time. However, the main effect of
Group (Treatment vs. SHAM) is not statistically significant (Estimate =
-0.42, p=0.51). This suggests that being in the treatment group did not
significantly impact pain levels compare to the SHAM group. To check
whether the treatment had a different effect over time compared to the
SHAM, we included interaction between time and treatment as one of the
predictors. Since the interactions are not significant, there is no
evidence that the treatment provided greater pain relief over time
compared to the SHAM.
Intention-to-treatment Analysis
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: DVPRS ~ factor(Timepoint) + major_life_event + (1 | ID)
## Data: ces_itt_data
##
## REML criterion at convergence: 401.4
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.76813 -0.49526 0.06557 0.54886 2.13569
##
## Random effects:
## Groups Name Variance Std.Dev.
## ID (Intercept) 1.437 1.199
## Residual 1.825 1.351
## Number of obs: 105, groups: ID, 37
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 7.0958 0.3037 76.9402 23.368 < 2e-16 ***
## factor(Timepoint)2 -1.5621 0.3283 70.1735 -4.758 1.02e-05 ***
## factor(Timepoint)3 -1.9381 0.3430 70.5445 -5.650 3.16e-07 ***
## major_life_event1 1.3137 0.6451 100.9441 2.036 0.0443 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) fc(T)2 fc(T)3
## fctr(Tmpn)2 -0.537
## fctr(Tmpn)3 -0.502 0.524
## mjr_lf_vnt1 0.006 -0.218 -0.281
Table 5: Linear Mixed Model Results
|
term
|
estimate
|
std.error
|
statistic
|
p.value
|
confidence_interval
|
|
(Intercept)
|
7.10
|
0.30
|
23.37
|
<0.001
|
[6.49, 7.7]
|
|
factor(Timepoint)2
|
-1.56
|
0.33
|
-4.76
|
<0.001
|
[-2.22, -0.91]
|
|
factor(Timepoint)3
|
-1.94
|
0.34
|
-5.65
|
<0.001
|
[-2.62, -1.25]
|
|
major_life_event1
|
1.31
|
0.65
|
2.04
|
0.04
|
[0.03, 2.59]
|
|
sd__(Intercept)
|
1.20
|
NA
|
NA
|
NA
|
[NA, NA]
|
|
sd__Observation
|
1.35
|
NA
|
NA
|
NA
|
[NA, NA]
|
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: DVPRS ~ medication + factor(Timepoint) + (1 | ID)
## Data: sham_itt_data
##
## REML criterion at convergence: 190.7
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.37660 -0.51828 0.02078 0.41021 1.97274
##
## Random effects:
## Groups Name Variance Std.Dev.
## ID (Intercept) 1.951 1.397
## Residual 2.327 1.525
## Number of obs: 48, groups: ID, 16
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 7.2188 0.5170 31.1209 13.962 6.06e-15 ***
## medication 0.4258 0.7214 43.6429 0.590 0.55808
## factor(Timepoint)2 -1.8472 0.6033 32.7795 -3.062 0.00437 **
## factor(Timepoint)3 -2.2129 0.6488 34.6863 -3.411 0.00166 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) medctn fc(T)2
## medication 0.000
## fctr(Tmpn)2 -0.466 -0.448
## fctr(Tmpn)3 -0.433 -0.556 0.621
## 2.5 % 97.5 %
## .sig01 0.72 2.19
## .sigma 1.16 1.91
## (Intercept) 6.22 8.22
## medication -0.97 1.83
## factor(Timepoint)2 -3.02 -0.69
## factor(Timepoint)3 -3.48 -0.97
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: DVPRS ~ medication + factor(Timepoint) + (1 | ID)
## Data: ces_itt_data
##
## REML criterion at convergence: 380
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.80423 -0.59928 0.05418 0.55394 2.09784
##
## Random effects:
## Groups Name Variance Std.Dev.
## ID (Intercept) 1.576 1.256
## Residual 1.819 1.349
## Number of obs: 99, groups: ID, 33
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 7.0152 0.3207 65.9998 21.871 < 2e-16 ***
## medication 0.2813 0.4765 94.3932 0.590 0.556369
## factor(Timepoint)2 -1.3949 0.3565 67.9376 -3.913 0.000214 ***
## factor(Timepoint)3 -1.7841 0.3745 71.1148 -4.764 9.73e-06 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) medctn fc(T)2
## medication 0.000
## fctr(Tmpn)2 -0.482 -0.365
## fctr(Tmpn)3 -0.459 -0.463 0.581
## 2.5 % 97.5 %
## .sig01 0.84 1.73
## .sigma 1.12 1.58
## (Intercept) 6.39 7.64
## medication -0.66 1.22
## factor(Timepoint)2 -2.09 -0.70
## factor(Timepoint)3 -2.52 -1.06
## Warning: Using size for a discrete variable is not advised.
Secondary Outcome: Sleep Quality
Due to missing scores in the Neural Quality of Sleep (NQSLP) columns,
we implemented an adjustment strategy. We considered two approaches:
Primary method (implemented): For participants with missing item
scores, we summed the available valid NQSLP items and applied a
proportional adjustment factor (8/7) to estimate the complete score.
These adjusted total scores were then converted to standardized T-scores
using the PROMIS (Patient-Reported Outcomes Measurement Information
System) Scoring Manual.
Alternative method (not used): We considered imputing missing
values using last observation carried forward (e.g., for CES08 with
missing values at time points 1 and 3, we considered using the score of
3 from time point 2). We ultimately rejected this approach in favor of
the proportional adjustment method, which provides more robust estimates
in our context.
Specific adjustments made:
- CES08: One missing score at baseline and another at 12 weeks
- Baseline adjusted total: (3+3+3+4+4+4+4)×(8/7) = 28.57
[T-score=61]
- 12-week adjusted total: (3+2+1+3+4+3+4)×(8/7) = 22.86
[T-score=55.3]
- CES15: One missing score at baseline
- Baseline adjusted total: (5+4+5+4+5+4+4)×(8/7) = 35.43
[T-score=68.7]
- CES32: Entire row of scores missing at 6 weeks
- No adjustment needed as our linear mixed-effects model can handle
missing follow-up data, effectively ignoring the entire missing
row.
Since there are some missing values in Usage_Time_hrs,
the linear mixed-effects model automatically exclude the observations
with NA values.
## Warning: NAs introduced by coercion
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula:
## t_score ~ Age + Sex + Ethnicity + Usage_Time_hrs + factor(Timepoint) +
## major_life_event.x + Alcohol_screener.y + `Using Drug (1 = Yes/ = 0 No)` +
## (1 | ID)
## Data: final_slp_data
##
## REML criterion at convergence: 545.5
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.95790 -0.53129 0.00281 0.56527 2.11445
##
## Random effects:
## Groups Name Variance Std.Dev.
## ID (Intercept) 21.302 4.615
## Residual 9.534 3.088
## Number of obs: 98, groups: ID, 33
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 60.57649 6.54339 29.93546 9.258 2.73e-10
## Age 0.18481 0.12665 26.76047 1.459 0.156156
## Sex2 2.89532 1.99107 28.23893 1.454 0.156931
## EthnicityW -1.13620 1.97364 26.87566 -0.576 0.569615
## Usage_Time_hrs -0.12483 0.03867 27.29649 -3.228 0.003234
## factor(Timepoint)2 -2.97429 0.87652 65.48256 -3.393 0.001176
## factor(Timepoint)3 -3.35644 0.92736 67.40500 -3.619 0.000566
## major_life_event.x1 2.56828 1.67412 79.65732 1.534 0.128965
## Alcohol_screener.y -0.88036 1.49009 62.53633 -0.591 0.556780
## `Using Drug (1 = Yes/ = 0 No)` -0.39333 1.20574 80.95597 -0.326 0.745107
##
## (Intercept) ***
## Age
## Sex2
## EthnicityW
## Usage_Time_hrs **
## factor(Timepoint)2 **
## factor(Timepoint)3 ***
## major_life_event.x1
## Alcohol_screener.y
## `Using Drug (1 = Yes/ = 0 No)`
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) Age Sex2 EthncW Usg_T_ fc(T)2 fc(T)3 mj__.1 Alch_.
## Age -0.874
## Sex2 -0.004 0.009
## EthnicityW -0.218 0.076 -0.320
## Usag_Tm_hrs 0.034 -0.354 0.008 0.124
## fctr(Tmpn)2 -0.084 -0.014 -0.006 0.023 -0.012
## fctr(Tmpn)3 -0.071 -0.021 -0.004 0.018 0.004 0.630
## mjr_lf_vn.1 0.043 0.022 -0.085 -0.021 -0.039 -0.280 -0.322
## Alchl_scrn. -0.362 0.057 -0.217 0.101 -0.102 0.140 0.116 -0.116
## `UD(1=Y/=0N -0.086 0.048 -0.008 0.006 -0.017 -0.375 -0.472 0.066 0.130
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: t_score ~ Usage_Time_hrs + factor(Timepoint) + (1 | ID)
## Data: final_slp_data
##
## REML criterion at convergence: 564.2
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.95218 -0.51697 -0.00116 0.54646 1.99753
##
## Random effects:
## Groups Name Variance Std.Dev.
## ID (Intercept) 21.367 4.622
## Residual 9.683 3.112
## Number of obs: 98, groups: ID, 33
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 67.7572 2.2026 34.0792 30.763 < 2e-16 ***
## Usage_Time_hrs -0.1029 0.0355 31.5122 -2.898 0.006786 **
## factor(Timepoint)2 -2.6908 0.7745 63.2180 -3.474 0.000931 ***
## factor(Timepoint)3 -3.0303 0.7661 63.0732 -3.956 0.000196 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) Usg_T_ fc(T)2
## Usag_Tm_hrs -0.898
## fctr(Tmpn)2 -0.159 -0.014
## fctr(Tmpn)3 -0.174 0.000 0.495
## 2.5 % 97.5 %
## .sig01 3.40 6.00
## .sigma 2.60 3.68
## (Intercept) 63.46 72.06
## Usage_Time_hrs -0.17 -0.03
## factor(Timepoint)2 -4.21 -1.18
## factor(Timepoint)3 -4.53 -1.53
Table 6: Linear Mixed Model Results
|
term
|
estimate
|
std.error
|
statistic
|
p.value
|
confidence_interval
|
|
(Intercept)
|
67.76
|
2.20
|
30.76
|
<0.001
|
[63.28, 72.23]
|
|
Usage_Time_hrs
|
-0.10
|
0.04
|
-2.90
|
0.01
|
[-0.18, -0.03]
|
|
factor(Timepoint)2
|
-2.69
|
0.77
|
-3.47
|
<0.001
|
[-4.24, -1.14]
|
|
factor(Timepoint)3
|
-3.03
|
0.77
|
-3.96
|
<0.001
|
[-4.56, -1.5]
|
|
sd__(Intercept)
|
4.62
|
NA
|
NA
|
NA
|
[NA, NA]
|
|
sd__Observation
|
3.11
|
NA
|
NA
|
NA
|
[NA, NA]
|
Based on the model results, we refined our analysis by including only
covariates with a statistically significant effect on sleep t-score. The
final model indicates that increased usage time is significantly
associated with a lower t-score (p = 0.006), suggesting better sleep
quality with higher usage time. Since experiencing a major life event
did not show a statistically significant effect on sleep t-score (p =
0.10), it was removed from the model to improve model interpretability
and efficiency.
Conclusion
While the results in this study did not demonstrate statistically
significant superiority of active CES over SHAM treatment, it presents a
comprehensive exploration of CES as a potential treatment for
fibromyalgia in Veterans. This study has several limitations and
proposes important future research directions.
Limitations: A primary limitation of this study is the insufficient
sample size of 50 participants, which significantly constrains the
statistical power and generalizability of the findings. Additionally,
Future investigations should conduct long-term follow-up studies to
assess the prolonged effects of CES on pain and sleep management
comprehensively.
Strengths: Despite no significant differences between groups, the
association between increased CES usage time and improved sleep quality
suggests potential benefits with longer exposure. We decided to conduct
ITT analysis. Given that the study faced limitations due to a small
sample size (N = 50), ITT would help preserve statistical power while
ensuring robust conclusions.
The boxplot displays the distribution of 