Opioid Treatment for Chronic Low Back Pain and the Impact of Mood Symptoms: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Opioid Treatment for Chronic Low Back Pain and the Impact of Mood Symptoms (NCT01502644), a non-randomized phase 4 study. The primary outcome is Percent Change in Average Daily Pain Score.

2. Clinical Question

3. Trial Design

This completed phase 4 study enrolled 81 participants and used a non-randomized, double-masked, parallel design according to the registry. The study concerned chronic low back pain and included people with low, moderate, or high negative-affect symptoms. The record describes these categories using Hospital Anxiety and Depression Scale subscale thresholds: low negative affect was a score at most 5 on each subscale, moderate was from 6 through 8 on each subscale, and high was at least 9 on each subscale.

Within each negative-affect category, the arm descriptions state that participants received placebo or active opioid drug, with immediate-release morphine or oxycodone given as needed up to three times a day for 1 week each in random order. The descriptions then refer to morphine or oxycodone titrated to a maximum allowable daily dose, but the entries are truncated before the full titration information. Exact schedules beyond the reported periods and doses are not reported in the registry record.

The record’s overall non-randomized allocation matters for interpretation. The negative-affect categories are not randomized groups, so differences between them can reflect characteristics other than symptom category. The descriptions of random order for drug periods do not turn the overall comparison between negative-affect groups into a randomized comparison. The registry itself cautions that the arms are not a like-for-like control comparison.

The primary outcome was assessed from baseline and over weekly averages for up to 20 weeks. The primary completion date was 2013-01; the study status is completed. The registry does not report a detailed protocol, a full account of masking procedures, or a complete schedule for the later titration phase.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Before assignment: NA level determined by Hospital Anxiety and Depression Scale (HADS) score: Anxiety and Depression subscales with 7 questions each, each question scored 0 (least) to 3 (greatest) amount of anxiety/depression, total score 0-21. High NA=score ≥9 on each subscale, Moderate NA=score ≥6 to ≤8 on each subscale, Low NA=HADS score ≤5 on each subscale.

Enrolled at Visit 1 After Pre-screening

Reasons for not completing this period:

Assigned to NA Groups

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Modified Intent-to-Treat Population, all participants who completed at least 50% of the opioid treatment period (at least 10 weeks of treatment).

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

Population: Modified Intent-to-Treat Population, all participants who completed at least 50% of the opioid treatment period (at least 10 weeks of treatment).

6. Primary Endpoint

The registry lists the primary outcome as Percent Change in Average Daily Pain Score, measured over Baseline and Week 20.

The primary endpoint was percent change in average daily pain score, a continuous measure. Participants rated their average lower back pain over the preceding 24 hours on an 11-point scale, from 0 for no pain to 10 for the worst possible pain, and recorded ratings in an electronic diary. Weekly averages were used to calculate change from baseline over up to 20 weeks. The registry states that a positive change from baseline indicates improvement.

Using a continuous pain measure retains more information about the degree of change than reducing each participant’s experience to a binary category such as improved or not improved. Repeated weekly ratings can also describe the course of pain over the treatment period. These are general reasons to use continuous, repeated outcomes; they do not establish that every participant completed every diary entry or that the registry reports the resulting trajectory.

The registry identifies baseline and Week 20 as the timeframe and describes linear mixed modeling to incorporate the weekly observations. However, no mean percent change, standard deviation, confidence interval, p-value, or group contrast is reported in the registry record. The endpoint definition therefore explains what the analysis was intended to measure, but the record does not permit a quantitative assessment of the observed pain changes or their precision.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does the model description mean?

The registry reports a linear mixed model for the repeated weekly pain measurements. In general, a mixed model can represent both average patterns across participants and variation among individual participants. This can be useful when a continuous outcome is measured repeatedly, because measurements from the same participant are usually related rather than independent. The registry says the model was used to allow most participants with any missing data to be included, but does not report the exact number included or the estimates produced.

The listed fixed effects were group, group by week, average baseline pain, and opioid use at baseline, recorded as yes or no. In general, fixed effects describe modeled average associations. A group-by-week term allows the modeled pattern over time to differ among groups. Including average baseline pain and baseline opioid use makes those variables part of the model; it does not show that all relevant differences between groups were controlled.

The record specifies an autoregressive covariance structure for the fixed-effects portion and a compound symmetry covariance structure for participant, intercept, and week as random effects. In general, covariance structures describe assumptions about how repeated observations or random effects relate to one another. An autoregressive structure represents a pattern in which measurements nearer in time are more closely related than measurements farther apart. Compound symmetry represents a common correlation pattern across the observations it describes. These choices affect model estimation and uncertainty, but the record does not provide diagnostics or sensitivity analyses that would show how well the assumptions fit these data.

Why does the non-randomized allocation matter?

Random allocation is one way to balance, on average, both known and unknown participant characteristics across treatment groups. Here, the registry records the overall allocation as non-randomized and describes groups defined by negative-affect levels. Thus, contrasts between those groups are observational with respect to negative affect. They may reflect the symptom grouping and other differences that accompany it. A statistical model can adjust for variables included in the model, but cannot guarantee comparability when relevant variables are unmeasured or unmodeled.

The arm descriptions also state that placebo and active opioid drug periods occurred in random order. This is a separate design detail from assignment to negative-affect groups. The registry’s own caution that the arms are not a like-for-like control comparison should be kept in view when interpreting contrasts. Full information about sequencing, period effects, carryover, and the analysis of treatment periods is not reported in the registry record.

What does the endpoint represent?

The outcome is a mean of a continuous measure: percent change in average daily pain score. Participants recorded pain on an 11-point scale, with weekly averages used to derive change from baseline. The registry states that positive change indicates improvement. In general, expressing change relative to baseline can describe how a participant’s score shifted from their starting level, while analyzing repeated weekly averages can use information across the treatment period. However, the registry does not provide the calculation details for percent change beyond that description, nor does it provide the outcome estimates.

A confidence interval, if reported, would describe the range of values reasonably compatible with the data and model under the method’s assumptions. It would communicate precision as well as the estimated effect. Since no confidence interval or point estimate is reported for the primary outcome, the precision of any group contrast cannot be judged. A p-value alone, even if one were reported, would not communicate the magnitude or clinical importance of a difference; in this record, the primary-outcome p-value is not reported.

How should missing observations be considered?

The registry says the mixed model allowed inclusion of the majority of participants with any missing data. As a general matter, likelihood-based mixed models can use available repeated measurements without requiring every participant to have a complete series, subject to assumptions about the missing-data process and the model. The record does not describe why measurements were missing, how much data were missing, whether missingness differed by group, or whether alternative missing-data analyses were performed. Therefore, the stated modeling approach should not be read as proof that missingness had no effect on the findings.

The analysis population was modified intent-to-treat and required completion of at least 50% of the opioid treatment period, or at least 10 weeks. This population definition excludes participants who did not meet that completion rule from the stated primary analysis population. The record does not explain the effect of this restriction on the results, and no primary-outcome results are reported. The exact model estimates, standard errors, degrees-of-freedom method, and diagnostic checks are also not reported in the registry record.

10. Confidence Intervals for Each Arm

No two-arm comparison is shown for this trial, but each arm's mean can be given with a t-based 95% confidence interval computed from the posted mean, spread and number of participants.

R code and output

arm <- c("Low NA", "Moderate NA", "High NA")
n   <- c(24, 7, 24)
m   <- c(38.6, 30, 20.6)
sp  <- c(35, 24.4, 26.5)
sd  <- sp      # posted standard deviations
for (i in seq_along(n)) {
  se <- sd[i] / sqrt(n[i]); h <- qt(0.975, n[i] - 1) * se
  cat(sprintf("CARM|%s|%d|%.3f|%.3f|%.3f|%.3f\n", arm[i], n[i], m[i], sd[i], m[i] - h, m[i] + h))
}

CARM  Low NA  24  38.600  35.000  23.821  53.379
CARM  Moderate NA  7  30.000  24.400  7.434  52.566
CARM  High NA  24  20.600  26.500  9.410  31.790

11. Multiplicity, Missing Data and Analysis Populations

The record lists one primary outcome measure and one analysis. It does not report a multiplicity adjustment or a plan for handling multiple comparisons. The absence of a reported adjustment is not evidence that no other outcomes, subgroups, or analyses were examined; a complete analysis plan and full protocol are not reported in the registry record. Because the research question concerns response across negative-affect groups and drug conditions, the number and hierarchy of comparisons would be relevant to interpreting any inferential results, but those details are not fully specified in the posted record.

Missing data are addressed in the outcome description: a linear mixed model was used to allow the majority of participants with any missing data to be included. The record does not quantify missingness, identify its reasons, state whether it differed by negative-affect group, or describe sensitivity analyses. In general, mixed models can use observed repeated measurements without requiring complete data for every participant, but interpretation depends on assumptions about the relationship between missingness and the observed and unobserved outcomes. Those assumptions and their evaluation are not reported.

The modified intent-to-treat analysis population was defined as participants completing at least 50% of the opioid treatment period, or at least 10 weeks. The record does not state how many participants met this condition or contributed to each model estimate. It also does not report a separate analysis population definition for safety. No numerical primary-outcome estimates are reported, so the impact of either missing observations or the population restriction on the treatment-effect estimates cannot be assessed.

12. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 0% reporting threshold.

The registry lists serious adverse events and other adverse events separately for the low, moderate, and high negative-affect groups. For each group, the posted figures show zero participants affected by serious adverse events and zero affected by other adverse events. The corresponding at-risk counts are listed as 30 for the low negative-affect group, 10 for the moderate group, and 32 for the high group. These are registry safety entries and should be read as counts within the recorded safety groups, not as evidence that no adverse events could occur.

Deaths are recorded as null for each group, which means death counts are not reported in the registry record; null should not be read as a reported zero. No serious-event terms are listed. The safety entries do not explain the ascertainment procedures, reporting window, event definitions, or completeness of follow-up. The outcome record also does not provide treatment-specific safety contrasts that would support attributing the group counts to placebo, morphine, or oxycodone.

In general, a zero count in a registry entry does not establish absence of risk. It describes what was recorded in that entry, and its interpretation depends on the population observed and the completeness and duration of monitoring. The registry record does not provide enough detail to assess those factors fully.

13. What the Result Does, and Does Not, Mean

The study was designed to examine whether opioid pain responses differed according to negative-affect symptom levels among people with chronic low back pain. It defines a continuous pain outcome, describes weekly measurement over up to 20 weeks, and specifies a mixed-model approach. The registry also provides information about the groups, drug regimens, and the modified intent-to-treat population. These features clarify the intended question and some aspects of how it was addressed.

What the record does not establish is the result of that question. Numerical primary-outcome estimates and inferential statistics are not reported in the registry record. It is therefore not possible to describe whether pain scores improved or worsened on average in a particular group, whether responses differed between groups, or how precisely any contrast was estimated. The study’s purpose is not itself evidence of a finding.

The design also limits causal interpretation of differences across negative-affect groups. Those groups were not randomized, and the registry describes non-randomized overall allocation. Although the arm descriptions say drug periods were received in random order, that does not make the negative-affect group comparison randomized. The model includes some stated covariates, but the record does not establish that all relevant group differences were measured and accounted for.

The safety entries report zero serious and other adverse events in each group, while death counts are not reported. Those entries should be interpreted as registry counts, not as a comprehensive claim about the safety of the drugs. The record does not include enough information about surveillance and follow-up to assess event ascertainment fully.

Overall, the posted record allows a reader to understand the question, endpoint, broad design, and selected analysis features. It does not provide the outcome results needed to judge the size, direction, statistical precision, or clinical relevance of a pain-response difference. No treatment recommendation follows from the information recorded.

14. Why This Trial Matters Statistically

This study raises a clinically relevant statistical question: whether pain responses during opioid treatment vary with depression and anxiety symptoms in people with chronic low back pain. The registry makes clear that the negative-affect categories are central to the question, and it defines those categories using symptom-scale thresholds. It also identifies repeated pain ratings and a continuous change measure, rather than relying only on a single final rating or a binary response category.

Several design features make careful interpretation especially important. The study is recorded as non-randomized, and the categories of negative affect are not randomized treatment groups. Differences among them may reflect factors beyond symptom level. The model adjusts for average baseline pain and opioid use at baseline, but the record does not establish that these are the only relevant differences or that adjustment removes confounding.

The outcome description also illustrates why analysis details matter. Repeated measurements and missing observations motivated a linear mixed model, with specified covariance structures and both fixed and random effects. Those choices define assumptions about the data and influence estimates and uncertainty. The registry describes the approach, but does not report diagnostics, sensitivity analyses, or the numerical results needed to assess the model’s conclusions.

Finally, the record’s main statistical contribution to interpretation is the distinction between the stated research question and evidence answering it. Without primary-outcome estimates and precision measures, the reader cannot evaluate the magnitude or reliability of a difference. This is a reminder that a detailed endpoint and model description are not substitutes for posted results. It is also important to distinguish within-group random ordering of drug periods, as described, from random assignment to the negative-affect groups.

15. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The principal limitation for interpreting the primary question is that numerical pain-outcome results are not reported in the registry record. There are no reported group means, standard deviations, contrasts, confidence intervals, or p-values. As a result, the direction, magnitude, and precision of any association between negative affect and pain response cannot be assessed.

The overall allocation is non-randomized. Negative-affect groups may differ in ways that also relate to pain ratings or treatment response. The model description identifies adjustment for average baseline pain and baseline opioid use, but does not report other candidate covariates, the rationale for their inclusion, or whether residual confounding was considered. The registry cautions that the arms are not a like-for-like control comparison.

The drug descriptions are incomplete. They report immediate-release morphine doses of 15 to 30 mg and oxycodone doses of 5 to 10 mg, used up to three times a day as needed for 1 week each in random order, followed by titration to a maximum allowable daily dose. The full titration details are not reported in the registry record. Nor does the record provide enough information to evaluate period effects or carryover between drug periods.

The analysis population required completion of at least 50% of the opioid treatment period, or at least 10 weeks. The number meeting that definition is not reported, and the record does not describe how excluded participants differed from those analyzed. Missing-data amounts, reasons, patterns, and sensitivity analyses are also not reported. Finally, the full statistical analysis plan, detailed masking procedures, and safety ascertainment methods are not reported in the registry record.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Opioid Treatment for Chronic Low Back Pain and the Impact of Mood Symptoms — NCT01502644.
• Registry record last updated 2017-07-11; status last verified 2017-06.
• Results first posted 2017-06-08.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

19. Record Summary

This completed phase 4 study examined opioid treatment for chronic low back pain in groups defined by low, moderate, or high negative-affect symptoms. The registry records non-randomized allocation, double masking, a parallel model, and enrollment of 81 participants. Its primary outcome was percent change in average daily pain score, based on an 11-point pain rating and weekly averages over up to 20 weeks.

The registry describes a linear mixed model and a modified intent-to-treat population requiring at least 50% of the treatment period. It also records zero serious and other adverse events in each negative-affect group; death counts are not reported. Numerical results for the primary pain outcome are not reported in the registry record, so the central group-response question cannot be quantitatively assessed from the posted results.