April 2020 HSC Section 4 - Plastic and Reconstructive Problems

Annals of Plastic Surgery • Volume 83, Number 2, August 2019

Keloid Excision and Adjuvant Treatments

as funnel plot asymmetry testing is not recommended for meta-analyses with fewer than 10 studies each. 31 Network meta-analysis was conducted to collectively evaluate the role of all the regimens in comparison to the rest by conducting both direct and indirect comparisons. We fitted a Bayesian random-effects network meta-analysis model to account for treatment effect heterogeneity across studies. 32,33 We assumed that the magnitude of the effect heteroge- neity is the same for every treatment (ie, homogeneous REM) and that evidence from direct and indirect comparisons is the same (ie, consis- tency model). We also assessed inconsistency and discrepancy of direct and indirect evidence, using a node-splitting approach. 34 We ranked treatments using the “ Surface under the Cumulative Ranking ” curve (SUCRA) value. 35 SUCRA is the percentage ratio between the surface underneath the cumulative ranking curve and the entire space in each plot. A larger SUCRA value indicates treatments with a higher rank. To obtain the cumulative ranking probabilities, we first calculated the ranking probabilities for each treatment at any possible ranks (eg, prob- ability of being the best, second best, or worst) and then calculated the cumulative sum of these ranking probabilities (eg, probability of being among the top 2 treatments). Markov chain Monte Carlo (MCMC) algo- rithms were used to estimate treatment effects. We also used the “ gemtc ” R package, which recalls JAGS in R for MCMC sampling. 36 We em- ployed 4 parallel chains and obtained 50,000 samples after discarding 20,000 samples in each chain. Convergence of MCMC chains was checked using the Gelman and Rubin diagnostic 37 and trace plots. 38 The initial search after removing duplicates revealed 3187 unique titles that were screened for eligibility by title and abstract. Seventy- four were considered eligible and were selected for evaluation based on the full-text articles. Finally, 14 were included in our meta-analysis. No fur- ther articles were identified by screening reference lists of other included studies. The quality of the 14 selected studies was evaluated, and the re- sults are available in Table 1. Table 1 also contains basic relevant infor- mation regarding the patients included in each study and prior treatments received. Table 2 presents the treatments used and related recurrence rates. RESULTS Literature Search We performed 6 paired meta-analyses, comparing different treatment dyads: (1)No excision versus excision only: Two studies 39,52 were included, resulting in 25 patients in the nonexcision group and 15 in the excision- only group. Odds for recurrence were lower for the nonexcision group at the marginally statistically significant level (OR, 0.25; REM; 95% CI, 0.06 – 1.00; P = 0.05; I 2 = 0%, Fig. 2). (2)Excision only versus excision + 1 adjuvant drug: Five stud- ies 40,41,47,48,52 were included, resulting in 78 patients among the excision-only group and 192 among the excision + 1 adjuvant drug group. Odds for recurrence were greater for the excision-only group but were not statistically significant (OR, 1.41; REM; 95% CI, 0.78 – 2.54; P = 0.26; I 2 = 0%; Fig. 3). (3)Excision only versus excision + radiation: Six studies 39,40,47 – 49,51 were included, resulting in 119 patients among excision-only group and 127 among the excision + radiation group. Odds for recurrence were greater for the excision-only group and were not statistically significant (OR, 3.59; REM; 95% CI, 0.56 – 22.96; P = 0.18; I 2 = 77%; Fig. 4). (4)Excision + 1 adjuvant drug versus excision + 2 adjuvant drugs: Three studies 43 – 45 were included, resulting in 60 patients among the excision + 1 adjuvant drug and 66 among the excision + 2 adjuvant drugs group. Odds for recurrence were greater among the patients

N/S N/S

(24.1%), laser

therapy (17.2%), excision with intralesional

steroid injection (17.2%)

Prior excision

N/S Piercing (38.9%), trauma or

surgery (26.1%), vaccination

(17.7%), acne

(7.9%), burn

(6.9%), idiopathic (2.5%)

7.7 (group A),

7.2 (group B)

Mean height,

(23.4%), lip (18.7%)

Ear (42.4%), extremities (24.1%),

shoulder/deltoid (17.7%), chest

(9.4%), abdomen/

perineum (5.4%),

face/neck (1.0%)

0% (Asian 100%)

(interquartile range, 12 – 18)

Outcomes *MINORS criteria for quality assessment of observation studies was used (range of scores, 0 – 24). † Jadad quality assessment score for clinical trials was used (range of scores, 0 – 5). ‡ Full text not available. N/S, not specified.

Mean, 27 100% Face (59.8%), ear Median, 15

167 females, 36 males

59 males

Korea

Prospective cohort 15* 2005 – 2006 Nigeria 48 Females, Retrospective cohort 16* 2002 – 2012 South

Chang, 52 2015

Aluko-Olokun et al, 51 2014

Park and

www.annalsplasticsurgery.com

© 2019 Wolters Kluwer Health, Inc. All rights reserved.

130