CAEP7712 - Week 5

Compassion Data Statistical Analysis

CD <- read.csv("~/Northeastern/Fall 2018/CAEP7712/Week 5/compassiondata.csv")

Data Exploration

dat <- CD %>% select(starts_with("com"))
(de.out <- lapply(names(dat), dat = dat, FUN = function(nm, dat) {
    g <- ggplot(data = dat, mapping = aes_string(x = nm)) + geom_density() + theme_light() + 
        xlab(nm) + ylab("Density") + ggtitle(paste("Density Distribution of", ifelse(str_detect(nm, 
        "pre"), "Compassion Pre Test Group", "Compassion Post Test Group"))) + theme(plot.title = element_text(hjust = 0.5), 
        plot.subtitle = element_text(hjust = 0.5))
    return(list(Graph = g, Shapiro = dat[[nm]] %>% shapiro.test()))
}))

## [[1]]
## [[1]]$Graph

## 
## [[1]]$Shapiro
## 
##  Shapiro-Wilk normality test
## 
## data:  .
## W = 0.98367, p-value = 0.2536
## 
## 
## 
## [[2]]
## [[2]]$Graph

## 
## [[2]]$Shapiro
## 
##  Shapiro-Wilk normality test
## 
## data:  .
## W = 0.98977, p-value = 0.6464

The compassion pre-test Shapiro-Wilk test indicates that the distribution does not differ significantly from normality p>.05 and appears to be normally distributed and therefore is amenable to a t-test. The compassion post-test Shapiro-wilk test indicates that the distribution does not differ significantly from normality p>.05 and also appears to be normally distributed and therefore is amenable to a t-test. ## T-Tests ### Single Sample

Is the mean of the pre-test group 1 equal to 50?

Hypothesis
\(H_0: \mu = 50\)
\(H_a: \mu \neq 50\)

t.out <- CD %>% filter(group == 1) %>% select(compassion.pre) %>% t.test(mu = 50)
d.out <- (CD %>% filter(group == 1) %>% select(compassion.pre) %>% unlist %>% mean(na.rm = T) %>% 
    {
        . - 50
    }/CD %>% filter(group == 1) %>% select(compassion.pre) %>% unlist %>% sd(na.rm = T))

t.apa <- t.out %>% apa_t

A two-tailed one-sample test of whether the compassion pre test group 1 has a mean value of 50 showed that the mean is not signficantly different from 50 at the 95% confidence level but is significant at the .1 \(\alpha\) level t (49) = -1.94 , CI[ 44.29,50.1 ], p<.1. The effect size d(50) = -0.2743329. This effect size is considered small.

Two-sample Test

Are the means of the pre-sample groups significantly different?

Hypothesis
\(H_0: \mu_1 - \mu_2 = 0\)
\(H_a: \mu_1 - \mu_2 \neq 0\)

f.out <- CD %>% var.test(compassion.pre ~ group, data = .)
tags$p("Tests of variance indicate that the variance differs signficantly between groups.")

Tests of variance indicate that the variance differs signficantly between groups.

t.out <- CD %>% t.test(compassion.pre ~ group, data = .)
d.out <- CD %>% effsize::cohen.d(compassion.pre ~ group, data = .)

A two-tailed two-sample t-test at the 95% confidence level indicates that the means of the two groups different significantly t (90.14) = -4.12 , CI[ -10.97,-3.83 ], p<.001. The effect size d = -0.82 , CI[ -1.24,-0.41 ], is considered large.

Paired Two-Sample Test

Did the intervention between pre & post test groups have a significant effect per group?

Hypothesis
\[H_0: \mu_{1pre} - \mu_{1post} = 0 | \mu_{2pre} - \mu_{2post} = 0 \\ H_a: \mu_{1pre} - \mu_{1post} \neq 0 | \mu_{2pre} \neq \mu_{2post} = 0\]

prd.out <- lapply(c(1, 2), dat = CD, FUN = function(grp, dat) {
    t.out <- t.test(dat[["compassion.pre"]][dat[["group"]] == grp], dat[["compassion.post"]][dat[["group"]] == 
        grp], paired = T)
    d.out <- effsize::cohen.d(dat[["compassion.pre"]][dat[["group"]] == grp], dat[["compassion.post"]][dat[["group"]] == 
        grp], paired = T)
    return(list(T = t.out, D = d.out))
})

The paired two-sample t-test at the 95% confidence level for group 1 indicated that the intervention had a significant effect t (49) = -6.1 , CI[ -2.22,-1.12 ], p<.001 and a large effect size d = -0.86 , CI[ -1.28,-0.45 ], . Similarly for group 2 t (49) = -6.99 , CI[ -2.51,-1.39 ], p<.001, also with a large effect size d = -0.99 , CI[ -1.41,-0.57 ], .

Study Summary

Arthroscopic partial meniscectomy versus placebo surgery for a degenerative meniscus tear: a 2-year follow-up of the randomised controlled trial

hilghts <- rpdfclown::extractPDF("C:\\Users\\Administrator\\Documents\\Northeastern\\Fall 2018\\CAEP7712\\Week 5\\Sx vs Placebo annrheumdis-2017-211172.full.pdf")[1] %>% 
    gsub("[^[:alnum:][:blank:]?&/\\-\\:\\.\\,\\']", " ", .) %>% gsub("\\s(?=Page.?.?\\d)", 
    "\\\n", ., perl = T) %>% gsub("(?<=Page.\\d)\\:", ": ", ., perl = T) %>% str_split(pattern = "\\n") %>% 
    .[1] %>% lapply(function(.) gsub("\\s{2,}", " ", .)) %>% unlist

Research Questions

Page.1: Research Question: To assess if arthroscopic partial meniscectomy APM is superior to placebo surgery in the treatment of patients with degenerative tear of the medial meniscus.

Page 2: Research question / Hypothesis: Accordingly the aim of this extension of our recently published Finnish Degenerative Meniscal Lesion Study FIDELITY trial 29 was twofold: a. to assess if APM is superior over placebo surgery over the course of 24 month follow up determined using patient relevant outcomes, the frequency of unblinding of te treatment group allocation and clinical examination of the knee and b. to assess whether our data corroborates or refutes common assertions regarding existence of subgroups of patients likely to benefit from APM

Sample

Page.1: Sample: 146 adults, aged 35 65 years, with knee symptoms consistent with degenerative medial meniscus tear and no knee osteoarthritis

Hypothesis

See 'Research Questions'
Page.1: Such recommendations rest on three issues: gener ally favourable clinical experience, some before after studies on patients undergoing APM due to persisting symptoms despite conservative treatment 17 18 and particularly the evidence from three RCTs 19 21 in which one third of participants initially allocated to non surgical treatment opted for crossing over to APM due to persisting knee symptoms or insufficient improvement. After undergoing APM, participants achieved similar outcomes compared with those initially assigned to surgery and those responding favourably to initial non surgical/conservative treat ment. 19 21 These findings have been interpreted as evidence that APM should be performed after failed conservative treatment. 22 Although such hypotheses might well be true, an alternative accounting can explain the number of crossovers and the beneficial treatment effects of surgery after failed conservative treatment: lack of blinding participants knowledge of not having undergone surgery may drive conser vatively treated patients to request surgery and also make them feel more content with the outcome once having undergone surgery.

Design

Page 1: Multicentre randomised participant blinded and outcome assessor blinded placebo surgery controlled trial of 146 adults aged 35-65 years with knee symptoms consistent with degenerative medial meniscus tear and no knee osteoarthritis randomised to APM or placebo surgery.

Method

Page 1: Methods (Measures): The primary outcome was the between group difference in the change from baseline in the Western Ontario Meniscal Evaluation Tool WOMET and Lysholm knee scores and knee pain after exercise at 24 months after surgery.

Page 2: MATERIALS AND METHODS
We conducted a multicentre randomised participant blinded and outcome assessor blinded placebo surgery controlled effi cacy trial involving participants aged 35 65 years with knee symptoms over 3 months consistent with degenerative medial meniscus tear and unresponsive to conventional conservative treatment and no clinical 30 or radiographic Kellgren Lawrence grade 1 31 knee osteoarthritis. The study took place in five orthopaedic centres in Finland during the period from December 2007 through March 2014. All patients had a suspicion of a meniscus tear based on symptoms and clinical tests a tear that was later verified on both MRI and knee arthroscopy. Patients with an obvious trauma induced onset of symptoms or with a recent history of a locked knee were excluded from the trial. On entering the study participants were informed that they would be allowed to consider a reoperation 6 months or later after the procedure if they did not have adequate relief of symptoms. Participants first underwent diagnostic knee arthroscopy and then during the same operation were assigned to APM or placebo surgery. For the randomisation the sequentially numbered opaque sealed envelopes were prepared by a statistician. Randomisation was performed in a 1:1 ratio with a block size of 4 and with stratification according to study site age 35-50 or 51-65 years sex and the absence or presence of minor degenerative changes on a radiograph, Kellgren Law rence grade 0 or 1 respectively. The participants, all caregivers, and those assessing the outcomes were blinded to the treatment assignment. Participants were followed up by questionnaires at 2, 6, 12 and 24 months. At the 24 month follow-up all participants were also clinically examined by an independent orthopaedic surgeon unaware of the treatment allocation. Standardised clinical examination included clinical meniscal tests, McMurray test pain provoked by joint line palpation and pain provoked by forced flexion and varus. Also range of knee motion knee crepitus bony enlargement effusion location of pain at palpation and knee stability was recorded.

Statistical Tests Used

tags$p(hilghts[c(10)])

Page.4: Statistical Analysis: A Student s t test and non parametric test Mann Whitney U test were used to compare continuous variables normally distributed and not normally distributed, respectively between the groups, and Fisher s exact test was used with binomial and categorical variables. All statistical analyses were performed on an intention to treat basis as the frequency of crossover was low, no per protocol analysis was performed. A p value of 0.05 was considered to indicate statistical significance. SPSS Statistics, V.23 IBM , was used for all statistical analyses.

Findings

hilghts[c(11:13)] %>% lapply(tags$p)

[[1]]

Page.4: Findings: Both groups showed a marked improvement in all primary outcomes. However, the difference between the two groups did not reach statistical significance and 95 CIs excluded clinically relevant effect in any of the three primary outcomes over the course of the 24 month follow up table 2 and figure 2 . Five participants 7.1 in the APM group and seven 9.2 in the placebo surgery group complained of symptoms severe enough to result in the unblinding of the treatment group allocation p 0.767 . Most of the participants, in both groups, were satis fied and reported improvement with no statistically significant difference between the two treatment groups. One participant in the APM group had a serious adverse event a knee infection 4 months after the initial operation . No between group differ ence was observed in the participants frequency in returning to normal activity level or in the frequency of mechanical symp toms. No statistically significant difference was found between the two groups in the meniscal tests during clinical examination either table 3 . The outcome of the patients who declined to participate n 17, five lost to follow up were similar with those randomised, excluding the change in WOMET score SD , which was greater for those declined 43.2 22.4 as compared with those randomised 29.5 21.1 with a between group difference 13.7 95 CI 25.6 to 2.9 . In the two subgroup analyses, one assessing the effect of preoperative mechanical symptoms and the other the effect of unstable tear on the treatment outcome, there was no difference in any of the primary or secondary outcomes between the APM and placebo surgery groups tables 4 and 5 .

[[2]]

Page.7: In our blinded trial, the frequency of unblinding of the treatment group allocation due to persisting symptoms was clearly lower than in the previous unblinded studies and we found no difference between our two treatment groups. Our data thus highlight the vital importance of proper blinding of study participants in surgical RCTs. Considering the rationale to carry out APM on those having failed previous conser vative treatment further, a recent study comparing exercise therapy to APM alone with no postoperative rehabilitation showed that although 19 of participants allocated to exercise therapy crossed over to surgery during the 2 year follow up, APM did not result in any additional benefit for them. 58

[[3]]

Page.7: Findings: In conclusion, the results of this randomised, placebo controlled trial with 24 months follow up show that APM provides no signif icant benefit over placebo surgery in patients with a degenerative meniscal tear and no knee osteoarthritis. These results support the evolving consensus that degenerative meniscus tear represents an early sign of knee osteoarthritis, rather than a clinical entity on its own, and accordingly, caution should be exercised in referring patients with knee pain and suspicion of a degenerative meniscal tear to MRI examination or APM, even after a failed attempt of conservative treatment.