Rate of postoperative recipient site complication according to study size. The number inside each circle is the reference number for that study. The arrow indicates a statistically significant outlier that was excluded from the final sample.
Rate of postoperative recipient site skin loss or necrosis according to study size. The number inside each circle is the reference number for that study. The arrow indicates a statistically significant outlier that was excluded from the final sample.
Szudek J, Taylor SM. Systematic Review of the Platysma Myocutaneous Flap for Head and Neck Reconstruction. Arch Otolaryngol Head Neck Surg. 2007;133(7):655-661. doi:10.1001/archotol.133.7.655
To systematically review and quantify complication rates and to identify preoperative factors among patients who underwent platysma myocutaneous flap reconstruction for head and neck cancer.
This study analyzed 190 patients in 16 case series published between 1982 and 2002. Funnel plots, contingency tables, and χ2 analyses were used to minimize bias and heterogeneity among the studies. Logistic regression models were used to quantify the associations between preoperative factors (age, sex, T stage, prior radiation therapy, and recipient site) and complications (skin loss or necrosis, fistula, dehiscence, hematoma, and infection) at different recipient sites (floor of mouth, alveolar ridge, pharyngeal wall, buccal mucosa, tongue or tongue base, and tonsil).
Patients described in the literature with head and neck surgery who underwent platysma flap reconstruction.
Main Outcome Measures
Results of logistic regression analyses.
Seventy-one patients (37%) developed a complication, ranging from 20% at the buccal mucosa to 55% at the tonsil and at the alveolar ridge. Major complications (ie, those requiring further surgery) occurred in 5% of patients. The most common complication was skin loss or necrosis, occurring in 25% of patients. Postoperative complications were not associated with age or sex but were associated with recipient site and tumor stage. Overall, complications were 0.3 (95% confidence interval [CI], 0.1-1.1) times less common at the buccal mucosa than at other recipient sites. Hematomas were 18.8 (95% CI, 1.6-217) times more common at the buccal mucosa. Infections were 20.0 (95% CI, 1.1-350) times more common at the pharyngeal wall. Major complications were 4.6 (95% CI, 0.9-23.5) times more likely, and fistulas were 9.2 (95% CI, 2.0-43.1) times more likely in patients with stage T3 or T4 oral cancer than in patients with lesser grades.
Postoperative complications were not associated with age, sex, or preoperative radiation therapy, but they were associated with recipient site and tumor stage. These results may guide surgeons considering the platysma flap to reconstruct head and neck cancer.
The platysma myocutaneous flap was introduced for intraoral reconstruction in 1978.1 It is easy to harvest, thin, and pliable; is large enough to close defects up to 70 cm2; and has low donor site morbidity.1- 8 Despite these and other attributes, the platysma flap has been eclipsed in head and neck reconstruction by pectoralis and free flaps.9 Its main limitations include lack of bulk and at least partial dependence on the facial artery,8,10 which preclude its use in large neck node resections.9 Another reason why the platysma flap has been avoided is because it is thought to be unreliable.9 Rates of complications between 10% and 40% have been reported2,3,5- 8 and have been linked to surgical technique, the surgeon's experience, and preoperative factors such as chemotherapy or radiation therapy, significant lymph node involvement, and risk of carotid artery exposure.6 Other studies5,7 have refuted the link between these preoperative factors and postoperative complications. The preponderance of small studies and the failure to consistently define the term complication have added confusion to this debate.
The objectives of our study were to systematically review the literature to quantify complication rates for platysma flaps and to identify preoperative factors associated with their occurrence. To that end, the following questions were posed: (1) In subjects with head and neck cancer who undergo platysma flap reconstruction, what are the rates of postoperative recipient site complications (skin loss or necrosis, fistula, dehiscence, hematoma, and infection)? (2) Do these rates differ among recipient sites (floor of mouth, alveolar ridge, pharyngeal wall, buccal mucosa, tongue or tongue base, and tonsil)? (3) Are any preoperative factors (age, sex, T stage, prior radiation therapy, and recipient site) associated with postoperative recipient site complications?
The assembly group of studies was identified by a PubMed search. The criteria surgical flaps AND head and neck neoplasms (Medical Subject Headings of the US National Library of Medicine) AND platysma were limited to English-language articles with abstracts and human subjects. This resulted in 43 studies. The eligible group of studies included those that described subjects with head and neck cancer who underwent platysma flap reconstruction. This included 28 of 43 articles from the assembly group. Manual cross-checking of their references yielded 3 more articles for a total of 31. The entered sample of studies included those that described preoperative factors (age, sex, T stage, prior radiation therapy, and recipient site) and postoperative complications (skin loss or necrosis, fistula, dehiscence, hematoma, and infection) on each patient. Fourteen of 31 studies in the eligible group lacked sufficient data on individual subjects or specific complications. The remaining 17 studies constituted the entered sample and include 207 patients. After heterogeneity analysis excluded 1 additional study (see the “Results” section), the final sample included 190 patients in 16 studies.
Prepublication bias was minimized by including in the entered sample only studies that reported data on individual patients. The data could then be interpreted directly rather than relying on the authors' interpretations and summary statistics. Publication bias was minimized by a method analogous to funnel plots in meta-analysis. The complication rate for each study was plotted against study size. The plots were inspected for symmetry and outliers. It was assumed that if the complication rate was not close to 0% or 100%, an unbiased distribution would be grossly symmetrical with larger studies centered around the mode. Smaller studies may be more spread out and include outliers. Selection bias was minimized by choosing studies based on methods of reporting data (ie, individual patients) rather than on results.
Contingency tables, χ2 analysis, and analysis of variance were used to test heterogeneity of complication rates among the eligible studies. This was performed by tabulating the presence or absence of complications across all 17 studies. The variable “any complication” was defined as the presence or absence of skin loss or necrosis, fistula, dehiscence, hematoma, or infection. A major complication was one that required further surgery. This generated a 2 × 17 table for each variable, which was then analyzed using Pearson product moment correlation χ2.11 Analysis of variance was used to test for the contribution to heterogeneity made by each study. Studies that contributed to heterogeneity were excluded from subsequent steps in the analysis. Studies not excluded on the basis of heterogeneity were used to calculate the mean complication rates and were included in further factor analysis described herein.
Contingency tables and χ2 analysis were also used to identify potential factors associated with complications. This was done by testing for the following: (1) differences in complications rates among the 6 recipient sites (floor of mouth, alveolar ridge, pharyngeal wall, buccal mucosa, tongue or tongue base, and tonsil) and (2) pairwise associations between recipient site complications (skin loss or necrosis, fistula, dehiscence, hematoma, and infection) and potential factors (age, sex, T stage, prior radiation therapy, and recipient site). Because the objective was to identify factors that should be scrutinized further in logistic regression models, a liberal critical P < .25 was chosen.
Most of the studies did not provide substantial detail regarding their surgical technique. Therefore, surgical technique was not included as a variable in this analysis. Most authors who described the technique used a superiorly based flap, elevated the skin flap to fully expose the platysma, and ensured a minimum muscle pedicle base width more than 3 cm.
The aforementioned significant factors were included in logistic regression models. This enabled the quantification of several associations simultaneously. In patients with multiple complications, only the primary complication was considered in this step. Logistic regression is a technique that attempts to quantify the relationship between a preoperative factor and the log of the odds of the response variable being “present.” For example, the association between the occurrence of a particular complication and the sex of the patient was expressed by the following equation: log(p[1|sex]/1 − p[1|sex]) = α + β(sex), where p[1|sex] is the probability that a complication has occurred given the sex of the patient. Maximum likelihood techniques were used to generate variable estimates. The critical P value for including a covariate in the final, most parsimonious model was P < .1, as is the accepted practice for this method.12 Logistic regression models have a straightforward interpretation for odds ratios. The strength of association between the complication and a factor such as sex is measured by β level. Continuing the example, male patients were exp(β) times more likely to have a complication than female patients. Confidence intervals (CIs) were calculated as ± 1.96 times the SE, where SE is the standard error for β.
After these data were modeled, the Hosmer and Lemeshow goodness-of-fit statistic was used to evaluate the degree of correspondence between the estimated probabilities of a complication produced by a model and the actual complications experienced by patients.12 The goodness-of-fit statistic first grouped patients into deciles of risk by using the logistic model to calculate each patient's predicted probability of complication and by ranking the patients according to this risk probability. The patients were then divided into 10 groups, with each group containing approximately 10% of the total number of patients. A simple χ2 test was then used to test for significant differences between the observed and expected frequencies of complications.
The studies in the entered sample are listed in the references.2- 4,7,8,13- 22 The patients are summarized in Table 1.
Figure 1 shows the proportion of patients with any postoperative recipient site complication according to study size. An apparent outlier is marked with an arrow. The distribution appears grossly symmetrical. Figure 2 shows the proportion of patients with postoperative recipient site skin loss or necrosis according to study size. Again, an outlier is marked with an arrow. The distribution appears to have a right skew. The other complications (major complication, fistula, dehiscence, hematoma, and infection) were not plotted because they occurred in fewer than 10% of patients.
Heterogeneity among different studies was found for several complications. Contingency table analysis comparing complication rates among the different studies demonstrated a statistically significant Pearson product moment correlation χ2 for the occurrence of any complication (P < .001), skin loss or necrosis (P = .005), fistula (P < .001), dehiscence (P < .001), hematoma (P = .12), and infection (P < .001). Major complication (defined as any complication that required further surgery) (P = .09) and hematoma (P = .12) were not significantly heterogeneous. Inspection of the contingency tables, analysis of variance results, and Figure 1 and Figure 2 suggest that 1 outlying study16 with 17 subjects contributed to most of the observed heterogeneity. Analysis of variance showed no heterogeneity (P < .1) when 1 particular study16 was excluded. Heterogeneity persisted (P = .03 and P = .007, respectively) when 2 other potential outlying studies8,17 were excluded. Therefore, only 1 study16 was excluded from the final sample. The remaining studies are listed in the references,2- 4,7,8,13- 22 and the patients are summarized in Table 1.
Overall, 71 patients (37%) developed a complication, ranging from 20% at the buccal mucosa to 55% at the tonsil and at the alveolar ridge (Table 2). Seven patients developed multiple complications as follows: 1 patient developed a hematoma and 2 patients developed fistulas in addition to skin loss or necrosis involving the floor of the mouth; 1 patient developed skin loss or necrosis, a fistula, and an infection involving the pharyngeal wall; 2 patients developed hematomas and skin loss or necrosis involving the buccal mucosa; and 1 patient developed a fistula and skin loss or necrosis involving the tongue or tongue base. Major complications (ie, those requiring further surgery) occurred in 5% of patients and ranged from none at the buccal mucosa and tongue or tongue base to 18% at the tonsil. The most common complication was skin loss or necrosis, occurring in 25% (47 of 190) of patients and ranging from 15% at the buccal mucosa to 36% at the tonsil. Skin loss or necrosis was the underlying postoperative problem in 11 of 13 patients with major complications requiring further surgery. None of the differences among recipient sites were statistically significant.
Table 3 gives the pairwise associations between the presence or absence of recipient site complications and potential preoperative factors. The strengths of association are summarized as odds ratios with 95% CIs. Several associations were found to have P < .25. These are indicated in boldface type in Table 3 and were considered further in a logistic regression analysis. For example, patients with T3 or T4 oral cancer developed postoperative fistulas 9.2 (95% CI, 2.0-43.2) times more often than patients who had a lower T stage. However, this does not consider the association found also between fistulas and male sex or recipient site. Logistic regression models were used to sort out which of these associations persisted once the others were taken into account.
Table 4 shows the associations considered in logistic regression models. For example, the model for the occurrence of postoperative fistulas at the recipient site considered male sex, T3 or T4 oral cancer, and floor of mouth site simultaneously. All of these factors were significant in the previous step, the pairwise analysis (Table 3). In the simultaneous logistic regression model, only the presence of T3 or T4 oral cancer was significantly associated with recipient site fistula: patients with T3 or T4 oral cancer developed postoperative fistulas 9.2 (95% CI, 2.0-43.1) times more often than patients with lower T stage (Table 4). In other models, patients with stage T3 or T4 tumor had major complications 4.6 (95% CI, 0.9-23.5) times more often than patients with lower T stage. As already noted, 11 of 13 major complications were caused by skin loss or necrosis. Patients with a buccal mucosa recipient site developed postoperative hematomas 18.8 (95% CI, 1.6-217) times more often than patients with other recipient sites. However, hematomas were rare, and patients with a buccal mucosa recipient site developed fewer postoperative recipient site complications overall (0.3 [95% CI, 0.12-1.1] times as many as patients with other recipient sites). Patients with a pharyngeal wall recipient site developed postoperative infections 20.0 (95% CI, 1.1-350) times as often as patients with other recipient sites. No significant associations were found for dehiscence or skin loss or necrosis at the recipient site. The Hosmer and Lemeshow statistic showed an adequate goodness-of-fit for all of the models.12
In subjects with head and neck cancer who underwent platysma flap reconstruction, the rate of any postoperative recipient site complication (including skin loss or necrosis, fistula, dehiscence, hematoma, and infection) was 37%. Previous rates have ranged between 10% and 40%.2,3,5- 8 Our study has improved the accuracy and precision of this estimate. We have done this by excluding outliers and by weighing complication rates among the remaining studies according to sample size. The rate of major complication (ie, requiring surgical revision) was 5%. Previous studies have not had a consistent definition of what constitutes a major complication. We defined it as any recipient site complication that requires surgical revision. Although this is a subjective definition that depends on the opinions of the surgeon and the patient, it is an important indicator of success. A measure of postoperative function would be more subjective but perhaps also more informative. However, even if all of the included studies had reported functional outcome, it would have been difficult to pool the data without a standard examination and questionnaire for all subjects. This is a potential area for further research.
The rates of complication (37%) and revision (5%) reported herein are similar to those in the largest single series of pectoralis myocutaneous flaps.23 Other published complication rates for the pectoralis flap range from 14% to 64%.23 Revision rates among free flaps used for head and neck reconstruction range from 4% to 9%.24,25 However, these flaps likely involve different defect and patient characteristics, making a direct comparison difficult.
The likelihood of a postoperative complication was not associated with age or sex (Table 3), but it depended partly on the recipient site (Table 4). Overall, complications were less likely to occur at the buccal mucosa than at any other recipient site. However, if a complication occurred at the buccal mucosa, it was far more likely to be a hematoma. In fact, 2 of 3 reported hematomas involved the buccal mucosa. This effect may be due to venous outflow obstruction, because the flap has to be folded more than 360° for proper placement.
Infections were more likely at the pharyngeal wall than at any other site. Koch9 has posited that pharyngeal recipient sites have higher rates of postoperative fistulas because they lack the support afforded to other oral sites by the mandible and tongue held in close approximation. Although our study found no association between the pharyngeal site and fistulas, the same mechanism may be responsible for the disproportionately high rate of infections at the pharyngeal recipient site.
Major complications (ie, requiring surgical revision) and fistulas were more likely in patients with stage T3 or T4 oral cancer than in patients with lesser grades. These findings are consistent with the idea that a more extensive defect is more difficult to reconstruct using this flap.
Some investigators report that preoperative radiation therapy compromises flap viability,6 while others refute this claim.4,7,9 It has been suggested that preoperative radiation therapy may cause muscle atrophy and may compromise the vascular pedicle to support the flap. It may also cause undesirable fibrosis of the involved skin.6 In our study, tissue loss or necrosis was 2.3 (95% CI, 0.6-8.4) times more likely in patients who had received preoperative radiation therapy. Because the lower 95% CI is not greater than 1.0, this association is not statistically significant and could be due to chance alone. However, only 17 (9%) of our sample had received preoperative radiation therapy, so it is possible that our test lacks the necessary power to refute the null hypothesis.
Our data are based on the observations of multiple examiners in 17 different centers. Multicenter studies often have problems with data consistency and reliability. However, our final sample was not heterogeneous for any recipient site complication.
Although our study incorporated all available data in the literature on platysma flaps, it includes only 190 patients. This was reflected in the large CIs in Table 3. Given that we excluded outliers via heterogeneity analysis, it is unlikely that the wide CIs represent widespread complication rates among the various studies. It is widely known that CIs that do not overlap 1.0 are statistically significant. Nevertheless, wide CIs, like some of those reported in this study, must be treated with caution because the certainty of the true value of the odds ratio is inversely related to the width of its CI. Perhaps as more studies are reported in the literature, these associations can be refined. Also, it would be useful for these future studies to report patients' smoking status, a known influence on perfusion and flap viability.16
Ten studies were excluded from our analysis because of insufficient data on individual patients. This is unlikely to bias our results given that the prevalences of postoperative recipient site complications among these excluded patients were similar to those included in our analysis: 34% of the excluded patients had any recipient site complication, 1% had major complications requiring further surgery, 19% had skin loss or necrosis, 10% had fistulas, 4% had dehiscence, and none had hematoma or infection.
As a final caveat, our findings of associations between preoperative factors and postoperative complications are implications only. They do not confirm a causative relationship.
This study was a systematic review of the published cases describing intraoral reconstruction using the platysma myocutaneous flap. To date, the use of this flap has been hindered at least partly by uncertainty regarding contraindications and complication rates. By critically pooling published cases, this study calculated the most accurate estimate to date of postoperative complication rates for this procedure. Although these rates are similar to those in pectoralis and free flaps, several factors must be considered when choosing a flap, including the site and complexity of the defect, the expertise of the surgeon, and the need for coverage of the great vessels. We also quantified the risk of complication associated with several preoperative factors. Our observations are consistent with previous anecdotal observations that the platysma flap is unaffected by age or sex and that it is best suited to repair small defects. Furthermore, our statistical models suggest that the buccal mucosa may be the recipient site with the lowest risk of postoperative complication, while the pharyngeal wall may be prone to postoperative infection. We anticipate that this information will be useful to surgeons and their patients as they make informed decisions regarding the use of the platysma flap.
Correspondence: Jacek Szudek, MD, PhD, c/o S. Mark Taylor, MD, Division of Otolaryngology, Queen Elizabeth II Health Sciences Centre, 5820 University Ave, Halifax, NS B3H 2Y9, Canada (email@example.com).
Submitted for Publication: March 4, 2006; final revision received January 2, 2007; accepted February 12, 2007.
Author Contributions: Drs Szudek and Taylor had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Taylor. Acquisition of data: Szudek. Analysis and interpretation of data: Szudek and Taylor. Drafting of the manuscript: Szudek and Taylor. Critical revision of the manuscript for important intellectual content: Szudek and Taylor. Statistical analysis: Szudek. Administrative, technical, and material support: Taylor. Study supervision: Taylor.
Financial Disclosure: None reported.