2017 Repair dental restorations

Data extracted was tabulated in a google sheet. Then exported as csv file and imported in R (R Core Team (2017). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.)

Data cleaning and organization was done with package tidyverse (Hadley Wickham (2017). tidyverse: Easily Install and Load the ‘Tidyverse’. R package version 1.2.1. https://CRAN.R-project.org/package=tidyverse).

The package meta (Guido Schwarzer (2007), meta: An R package for meta-analysis, R News, 7(3), 40-45.) was used for the meta-analysis. A funnel plot was used to detect publication bias. The heterogenicity between studies was checked with I2 and visualized with a Baujat plot (Baujat B, Mahé C, Pignon JP, Hill C (2002), A graphical method for exploring heterogeneity in meta-analyses: Application to a meta-analysis of 65 trials. Statistics in Medicine, 30, 2641–2652.). We grouped the studies with same intervention and control and considered any adverse outcome. A random effect meta-analysis using odds-ratio as outcome was performed with a Mantel-Haenzel method. A forest plot was used to visualize the overall effect of the interventions, with a 95% confidence interval.

Paquetes

Dataset

df <- read_csv("https://docs.google.com/spreadsheets/d/e/2PACX-1vRSKuBlcQTVJK2fZyZ4Nvf4SwSqVrxcfAGhNhl6dWHU9W39oppE4Pl4lcPy-0CRGRTqQQ0IHwFgfrY8/pub?gid=0&single=true&output=csv")

Data cleaning

glimpse(df)

Clean dataset from empty rows, strange symbols, etc

df <- df %>% filter(str_detect(df$Comparison, "vs"), 
                      trimws(EvA) != "", 
                    !str_detect(EvA, "\\?"), 
                    !str_detect(EvA, "%") )

df$firstAuthor <- str_trim(df$firstAuthor, "right")

Select only relevant columns

df <- df %>% select(Comparison:Outcome) %>% 
  select(-quality)

create a new column id

df <-  mutate(df, id = paste(firstAuthor, ", ", year, paper))

Dataset clean

Converting factors to numeric variables

glimpse(df)

df$EvA <- as.integer(df$EvA)
df$TotalA <- as.integer(df$TotalA)
df$EvB <- as.integer(df$EvB)
df$TotalB <- as.integer(df$TotalB)

df <- df %>% 
  mutate(groups = paste(Comparison, Outcome))

summary(df)

Create groups for comparisons

Any intervention (Sealant, Refinishing or Repair) vs grouped comparison and grouped outcome

df <- mutate(df, groups = paste( intervention_b, Outcome))
table(df$groups)

1. No-treatment MA (6)

Data selection

data_meta <- df %>% 
  filter(groups == "No-treatment MA")

meta1 <- metabin(EvA, TotalA, 
                 EvB, TotalB, 
                 data = data_meta,  
                 sm = "OR",  method.tau = "DL", 
                 comb.fixed = FALSE, 
                 studlab = paste(firstAuthor, year, intervention_a))

Bias

funnel.meta(meta1, 
            studlab = TRUE)

Heterogeneity

baujat.meta(meta1, 
            yscale = 10, xmin = 3, ymin = 10, 
            cex.studlab = .50)

Meta-analysis and forest plot

summary(meta1)

Number of studies combined: k = 6

                         OR           95%-CI    z  p-value
Random effects model 1.0181 [0.3054; 3.3946] 0.03   0.9767

Quantifying heterogeneity:
 tau^2 = 1.5285; H = 1.95 [1.29; 2.95]; I^2 = 73.8% [40.2%; 88.5%]

Test of heterogeneity:
     Q d.f.  p-value
 19.10    5   0.0018

Details on meta-analytical method:
- Mantel-Haenszel method
- DerSimonian-Laird estimator for tau^2
- Continuity correction of 0.5 in studies with zero cell frequencies

meta1

                             OR            95%-CI %W(random)
Gordan 2009 Sealant      0.6400 [0.0880;  4.6554]       14.8
Gordan 2009 Refinishing  2.8000 [0.5321; 14.7350]       16.8
Moncada 2015 Sealant     0.0345 [0.0017;  0.6894]        9.8
Estay 2017 Refinishing   0.2864 [0.0836;  0.9813]       19.6
Fernandez 2015 Sealant  12.0000 [2.1471; 67.0674]       16.4
Estay 2017 Refinishing   1.4082 [0.6730;  2.9464]       22.6

Number of studies combined: k = 6

                         OR           95%-CI    z  p-value
Random effects model 1.0181 [0.3054; 3.3946] 0.03   0.9767

Quantifying heterogeneity:
 tau^2 = 1.5285; H = 1.95 [1.29; 2.95]; I^2 = 73.8% [40.2%; 88.5%]

Test of heterogeneity:
     Q d.f.  p-value
 19.10    5   0.0018

Details on meta-analytical method:
- Mantel-Haenszel method
- DerSimonian-Laird estimator for tau^2
- Continuity correction of 0.5 in studies with zero cell frequencies

Forest

forest.meta(meta1,  
       comb.fixed = FALSE,
       sortvar = year, 
       # LEFT
       label.left         = "Seal or ref.", 
       col.label.left     = "darkgreen", 
       # RIGHT
       label.right        = "No treatment",
       col.label.right    = "darkred")

Unknown or uninitialised column: '.subset'.

Export

png(filename = "1.png", 
    type = "cairo",
    units = "in", 
    width = 10, 
    height = 5, 
    pointsize = 12, 
    res = 96)
forest.meta(meta1,  
       comb.fixed = FALSE,
       sortvar = year, 
       # LEFT
       label.left         = "Seal or ref.", 
       col.label.left     = "darkgreen", 
       # RIGHT
       label.right        = "No treatment",
       col.label.right    = "darkred")

Unknown or uninitialised column: '.subset'.

dev.off()

null device 
          1 

---

2. No-treatment SC (6)

Data selection

data_meta <- df %>% 
  filter(groups == "No-treatment SC")
data_meta

meta1 <- metabin(EvA, TotalA, 
                 EvB, TotalB, 
                 data = data_meta,  
                 sm = "OR",  method.tau = "DL", 
                 comb.fixed = FALSE, 
                 studlab = paste(firstAuthor, year, intervention_a))

Bias

funnel.meta(meta1, 
            studlab = TRUE, 
            cex.studlab = .55)

Heterogeneity

baujat.meta(meta1, 
            yscale = 10, xmin = 3, ymin = 10, 
            cex.studlab = .50)

Meta-analysis and forest plot

summary(meta1)

Number of studies combined: k = 6

                        OR           95%-CI    z  p-value
Random effects model 1.286 [0.4991; 3.3137] 0.52   0.6024

Quantifying heterogeneity:
 tau^2 = 0; H = 1.00 [1.00; 1.26]; I^2 = 0.0% [0.0%; 36.7%]

Test of heterogeneity:
    Q d.f.  p-value
 2.00    5   0.8487

Details on meta-analytical method:
- Mantel-Haenszel method
- DerSimonian-Laird estimator for tau^2
- Continuity correction of 0.5 in studies with zero cell frequencies

meta1

                            OR             95%-CI %W(random)
Gordan 2009 Sealant     0.5556 [0.0200;  15.4620]        8.1
Gordan 2009 Refinishing 1.2000 [0.0663;  21.7233]       10.7
Moncada 2015 Sealant    0.3103 [0.0116;   8.2917]        8.3
Estay 2017 Refinishing  5.0571 [0.1952; 131.0509]        8.5
Fernandez 2015 Sealant  0.8889 [0.1119;   7.0614]       20.9
Estay 2017 Refinishing  1.8333 [0.4372;   7.6869]       43.6

Number of studies combined: k = 6

                        OR           95%-CI    z  p-value
Random effects model 1.286 [0.4991; 3.3137] 0.52   0.6024

Quantifying heterogeneity:
 tau^2 = 0; H = 1.00 [1.00; 1.26]; I^2 = 0.0% [0.0%; 36.7%]

Test of heterogeneity:
    Q d.f.  p-value
 2.00    5   0.8487

Details on meta-analytical method:
- Mantel-Haenszel method
- DerSimonian-Laird estimator for tau^2
- Continuity correction of 0.5 in studies with zero cell frequencies

Forest

forest.meta(meta1,  
       comb.fixed = FALSE,
       sortvar = year, 
       # LEFT
       label.left         = "Seal or refinishing", 
       col.label.left     = "darkgreen", 
       # RIGHT
       label.right        = "No treatment",
       col.label.right    = "darkred")

Unknown or uninitialised column: '.subset'.

Export

png(filename = "2.png", 
    type = "cairo",
    units = "in", 
    width = 10, 
    height = 5, 
    pointsize = 12, 
    res = 96)
forest.meta(meta1,  
       comb.fixed = FALSE,
       sortvar = year, 
       # LEFT
       label.left         = "Seal or refinishing", 
       col.label.left     = "darkgreen", 
       # RIGHT
       label.right        = "No treatment",
       col.label.right    = "darkred")

Unknown or uninitialised column: '.subset'.

dev.off()

null device 
          1 

---

3. No-treatment TS (3)

Data selection

data_meta <- df %>% 
  filter(groups == "No-treatment TS")
data_meta

meta1 <- metabin(EvA, TotalA, 
                 EvB, TotalB, 
                 data = data_meta,  
                 sm = "OR",  method.tau = "DL", 
                 comb.fixed = FALSE, 
                 studlab = paste(firstAuthor, year, intervention_a))

Bias

funnel.meta(meta1, 
            studlab = TRUE, 
            cex.studlab = .55)

Heterogeneity

baujat.meta(meta1, 
            yscale = 10, xmin = 3, ymin = 10, 
            cex.studlab = .50)

Meta-analysis and forest plot

summary(meta1)

Number of studies combined: k = 3

                         OR            95%-CI    z  p-value
Random effects model 3.1616 [0.6856; 14.5791] 1.48   0.1400

Quantifying heterogeneity:
 tau^2 = 0; H = 1.00 [1.00; 2.27]; I^2 = 0.0% [0.0%; 80.6%]

Test of heterogeneity:
    Q d.f.  p-value
 1.07    2   0.5850

Details on meta-analytical method:
- Mantel-Haenszel method
- DerSimonian-Laird estimator for tau^2
- Continuity correction of 0.5 in studies with zero cell frequencies

meta1

                           OR             95%-CI %W(random)
Moncada 2015 Sealant   1.0000 [0.0563;  17.7510]       28.2
Estay 2017 Refinishing 8.9394 [0.4045; 197.5585]       24.4
Estay 2017 Refinishing 3.6774 [0.3992;  33.8804]       47.4

Number of studies combined: k = 3

                         OR            95%-CI    z  p-value
Random effects model 3.1616 [0.6856; 14.5791] 1.48   0.1400

Quantifying heterogeneity:
 tau^2 = 0; H = 1.00 [1.00; 2.27]; I^2 = 0.0% [0.0%; 80.6%]

Test of heterogeneity:
    Q d.f.  p-value
 1.07    2   0.5850

Details on meta-analytical method:
- Mantel-Haenszel method
- DerSimonian-Laird estimator for tau^2
- Continuity correction of 0.5 in studies with zero cell frequencies

Forest

forest.meta(meta1,  
       comb.fixed = FALSE,
       sortvar = year, 
       # LEFT
       label.left         = "Seal or refinish", 
       col.label.left     = "darkgreen", 
       # RIGHT
       label.right        = "No treatment",
       col.label.right    = "darkred")

Unknown or uninitialised column: '.subset'.

Export

png(filename = "3.png", 
    type = "cairo",
    units = "in", 
    width = 10, 
    height = 5, 
    pointsize = 12, 
    res = 96)
forest.meta(meta1,  
       comb.fixed = FALSE,
       sortvar = year, 
       # LEFT
       label.left         = "Seal or refinish", 
       col.label.left     = "darkgreen", 
       # RIGHT
       label.right        = "No treatment",
       col.label.right    = "darkred")

Unknown or uninitialised column: '.subset'.

dev.off()

null device 
          1 

---

4. Replacement MA (5)

Data selection

data_meta <- df %>% 
  filter(groups == "Replacement MA")
data_meta

meta1 <- metabin(EvA, TotalA, 
                 EvB, TotalB, 
                 data = data_meta,  
                 sm = "OR",  method.tau = "DL", 
                 comb.fixed = FALSE, 
                 studlab = paste(firstAuthor, year, intervention_a))

Bias

funnel.meta(meta1, 
            studlab = TRUE, 
            cex.studlab = .55)

Heterogeneity

baujat.meta(meta1, 
            yscale = 10, xmin = 3, ymin = 10, 
            cex.studlab = .50)

Meta-analysis and forest plot

summary(meta1)

Number of studies combined: k = 4

                         OR           95%-CI     z  p-value
Random effects model 0.7996 [0.2774; 2.3052] -0.41   0.6789

Quantifying heterogeneity:
 tau^2 = 0.5096; H = 1.34 [1.00; 2.31]; I^2 = 44.1% [0.0%; 81.3%]

Test of heterogeneity:
    Q d.f.  p-value
 5.37    3   0.1467

Details on meta-analytical method:
- Mantel-Haenszel method
- DerSimonian-Laird estimator for tau^2

meta1

                         OR            95%-CI %W(random)
Gordan 2011 Repair   0.1319 [0.0141;  1.2354]       16.1
Moncada 2015 Sealant     NA                          0.0
Estay 2017 Repair    0.6222 [0.1554;  2.4920]       28.9
Estay 2017 Repair    2.7045 [0.6384; 11.4576]       27.7
Gordan 2009 Repair   0.8750 [0.2020;  3.7907]       27.3

Number of studies combined: k = 4

                         OR           95%-CI     z  p-value
Random effects model 0.7996 [0.2774; 2.3052] -0.41   0.6789

Quantifying heterogeneity:
 tau^2 = 0.5096; H = 1.34 [1.00; 2.31]; I^2 = 44.1% [0.0%; 81.3%]

Test of heterogeneity:
    Q d.f.  p-value
 5.37    3   0.1467

Details on meta-analytical method:
- Mantel-Haenszel method
- DerSimonian-Laird estimator for tau^2

Forest

forest.meta(meta1,  
       comb.fixed = FALSE,
       sortvar = year, 
       # LEFT
       label.left         = "Seal or repair", 
       col.label.left     = "darkgreen", 
       # RIGHT
       label.right        = "Replacement",
       col.label.right    = "darkred")

Unknown or uninitialised column: '.subset'.

Export

png(filename = "4.png", 
    type = "cairo",
    units = "in", 
    width = 10, 
    height = 5, 
    pointsize = 12, 
    res = 96)
forest.meta(meta1,  
       comb.fixed = FALSE,
       sortvar = year, 
       # LEFT
       label.left         = "Seal or repair", 
       col.label.left     = "darkgreen", 
       # RIGHT
       label.right        = "Replacement",
       col.label.right    = "darkred")

Unknown or uninitialised column: '.subset'.

dev.off()

null device 
          1 

---

5. Replacement SC (5)

Data selection

data_meta <- df %>% 
  filter(groups == "Replacement SC")
data_meta

meta1 <- metabin(EvA, TotalA, 
                 EvB, TotalB, 
                 data = data_meta,  
                 sm = "OR",  method.tau = "DL", 
                 comb.fixed = FALSE, 
                 studlab = paste(firstAuthor, year, intervention_a))

Bias

funnel.meta(meta1, 
            studlab = TRUE, 
            cex.studlab = .55)

Heterogeneity

baujat.meta(meta1, 
            yscale = 10, xmin = 3, ymin = 10, 
            cex.studlab = .50)

Meta-analysis and forest plot

summary(meta1)

Number of studies combined: k = 4

                         OR           95%-CI     z  p-value
Random effects model 0.9363 [0.1065; 8.2309] -0.06   0.9526

Quantifying heterogeneity:
 tau^2 = 2.9848; H = 1.61 [1.00; 2.78]; I^2 = 61.4% [0.0%; 87.1%]

Test of heterogeneity:
    Q d.f.  p-value
 7.77    3   0.0511

Details on meta-analytical method:
- Mantel-Haenszel method
- DerSimonian-Laird estimator for tau^2
- Continuity correction of 0.5 in studies with zero cell frequencies

meta1

                         OR             95%-CI %W(random)
Gordan 2011 Repair   0.1319 [0.0141;   1.2354]       28.7
Moncada 2015 Sealant     NA                           0.0
Estay 2017 Repair    5.7273 [0.5312;  61.7487]       27.6
Estay 2017 Repair    6.5152 [0.2925; 145.1132]       22.4
Gordan 2009 Repair   0.1634 [0.0062;   4.3051]       21.3

Number of studies combined: k = 4

                         OR           95%-CI     z  p-value
Random effects model 0.9363 [0.1065; 8.2309] -0.06   0.9526

Quantifying heterogeneity:
 tau^2 = 2.9848; H = 1.61 [1.00; 2.78]; I^2 = 61.4% [0.0%; 87.1%]

Test of heterogeneity:
    Q d.f.  p-value
 7.77    3   0.0511

Details on meta-analytical method:
- Mantel-Haenszel method
- DerSimonian-Laird estimator for tau^2
- Continuity correction of 0.5 in studies with zero cell frequencies

Forest

forest.meta(meta1,  
       comb.fixed = FALSE,
       sortvar = year, 
       # LEFT
       label.left         = "Seal or repair", 
       col.label.left     = "darkgreen", 
       # RIGHT
       label.right        = "Replacement",
       col.label.right    = "darkred")

Unknown or uninitialised column: '.subset'.

Export

png(filename = "5.png", 
    type = "cairo",
    units = "in", 
    width = 10, 
    height = 5, 
    pointsize = 12, 
    res = 96)
forest.meta(meta1,  
       comb.fixed = FALSE,
       sortvar = year, 
       # LEFT
       label.left         = "Seal or repair", 
       col.label.left     = "darkgreen", 
       # RIGHT
       label.right        = "Replacement",
       col.label.right    = "darkred")

Unknown or uninitialised column: '.subset'.

dev.off()

null device 
          1 

---

6. Replacement TS (5)

Data selection

data_meta <- df %>% 
  filter(groups == "Replacement TS")
data_meta

meta1 <- metabin(EvA, TotalA, 
                 EvB, TotalB, 
                 data = data_meta,  
                 sm = "OR",  method.tau = "DL", 
                 comb.fixed = FALSE, 
                 studlab = paste(firstAuthor, year, intervention_a))

Bias

funnel.meta(meta1, 
            studlab = TRUE, 
            cex.studlab = .55)

Heterogeneity

baujat.meta(meta1, 
            yscale = 10, xmin = 3, ymin = 10, 
            cex.studlab = .50)

Meta-analysis and forest plot

summary(meta1)

Number of studies combined: k = 3

                        OR           95%-CI     z  p-value
Random effects model 0.369 [0.0875; 1.5561] -1.36   0.1745

Quantifying heterogeneity:
 tau^2 = 0; H = 1.00 [1.00; 2.73]; I^2 = 0.0% [0.0%; 86.5%]

Test of heterogeneity:
    Q d.f.  p-value
 1.55    2   0.4617

Details on meta-analytical method:
- Mantel-Haenszel method
- DerSimonian-Laird estimator for tau^2

meta1

                         OR            95%-CI %W(random)
Gordan 2011 Repair   0.1319 [0.0141;  1.2354]       41.4
Moncada 2015 Sealant 1.1538 [0.0654; 20.3419]       25.2
Estay 2017 Repair        NA                          0.0
Estay 2017 Repair    0.5588 [0.0464;  6.7269]       33.5
Gordan 2009 Repair       NA                          0.0

Number of studies combined: k = 3

                        OR           95%-CI     z  p-value
Random effects model 0.369 [0.0875; 1.5561] -1.36   0.1745

Quantifying heterogeneity:
 tau^2 = 0; H = 1.00 [1.00; 2.73]; I^2 = 0.0% [0.0%; 86.5%]

Test of heterogeneity:
    Q d.f.  p-value
 1.55    2   0.4617

Details on meta-analytical method:
- Mantel-Haenszel method
- DerSimonian-Laird estimator for tau^2

Forest

forest.meta(meta1,  
       comb.fixed = FALSE,
       sortvar = year, 
       # LEFT
       label.left         = "Seal or repair", 
       col.label.left     = "darkgreen", 
       # RIGHT
       label.right        = "Replacement",
       col.label.right    = "darkred")

Unknown or uninitialised column: '.subset'.

Export

png(filename = "6.png", 
    type = "cairo",
    units = "in", 
    width = 10, 
    height = 5, 
    pointsize = 12, 
    res = 96)
forest.meta(meta1,  
       comb.fixed = FALSE,
       sortvar = year, 
       # LEFT
       label.left         = "Seal or repair", 
       col.label.left     = "darkgreen", 
       # RIGHT
       label.right        = "Replacement",
       col.label.right    = "darkred")

Unknown or uninitialised column: '.subset'.

dev.off()

null device 
          1 

---

Citations

citation()

To cite R in publications use:

  R Core Team (2017). R: A language and environment for statistical computing. R Foundation for Statistical
  Computing, Vienna, Austria. URL https://www.R-project.org/.

A BibTeX entry for LaTeX users is

  @Manual{,
    title = {R: A Language and Environment for Statistical Computing},
    author = {{R Core Team}},
    organization = {R Foundation for Statistical Computing},
    address = {Vienna, Austria},
    year = {2017},
    url = {https://www.R-project.org/},
  }

We have invested a lot of time and effort in creating R, please cite it when using it for data analysis. See also
‘citation("pkgname")’ for citing R packages.

citation(package = "tidyverse")

To cite package ‘tidyverse’ in publications use:

  Hadley Wickham (2017). tidyverse: Easily Install and Load the 'Tidyverse'. R package version 1.2.1.
  https://CRAN.R-project.org/package=tidyverse

A BibTeX entry for LaTeX users is

  @Manual{,
    title = {tidyverse: Easily Install and Load the 'Tidyverse'},
    author = {Hadley Wickham},
    year = {2017},
    note = {R package version 1.2.1},
    url = {https://CRAN.R-project.org/package=tidyverse},
  }

citation(package = "meta")

To cite package 'meta' in publications use:

  Guido Schwarzer (2007), meta: An R package for meta-analysis, R News, 7(3), 40-45.

A BibTeX entry for LaTeX users is

  @Article{,
    title = {meta: {A}n {R} package for meta-analysis},
    author = {Guido Schwarzer},
    journal = {R News},
    year = {2007},
    volume = {7},
    number = {3},
    pages = {40--45},
  }

URL https://cran.r-project.org/doc/Rnews/Rnews_2007-3.pdf