In this meta-analysis, we evaluated the diagnostic accuracy of PGFNAB and USGFNAB in diagnosing thyroid nodule malignancy. The results of the USGFNAB index test showed that the values for pooled sensitivity, pooled specificity, DOR, and AUC were 90%, 80%, 40 and 0.92 respectively and had an estimated point on the SROC Curve in the upper left (see Fig. 5). These results indicated that the USGFNAB index test had excellent diagnostic accuracy. PGFNAB index test had lower results for pooled sensitivity, pooled specificity, DOR, and AUC than USGFNAB, namely 76%, 77%, 11, and 0.827, respectively.
The positive Tsens regression coefficient suggests that the USGFNAB sensitivity was better than the PGFNAB, and a p value <0.05 indicated that the result was statistically significant. The regression coefficient for Tfpr was negative, suggesting that the specificity of PGFNAB was better than USGFNAB, yet these results were not statistically significant. Relating to AUC value significance, a difference of 0.093 (p = 0.000023) was found. The comparison of the SROC curve image showed that the estimation points of the two curves were very distinct from the spheres or the CI values that were slightly intersected (see Fig. 5), suggesting a significant difference between the two index tests.
Meta-analyzes assessing the accuracy of the PGFNAB and USGFNAB diagnoses had previously been performed. Two meta-analyses evaluating the accuracy of the diagnosis of FNAB in the thyroid was noted. First, Ospina et al. (2016) conducted a meta-analysis pertaining the accuracy of the diagnosis of USGFNAB in thyroid nodules but did not compare it to PGFNAB [4]. Second, Matz et al. (2014) on this meta-analysis assessed the comparison between USGFNAB and PGFNAB [6]. The fundamental difference conducted by Matz et al. compared the two methods statistically, of which three studies were published before 2014 in that study. The results of this meta-analysis were consistent with the results of the study by Matz et al., where the pooled sensitivity value of USGFNAB was higher than that of PGFNAB [0.91 (CI = 0.82, 1.0) and 0.79 (CI = 0.69, 0.85), respectively]. However, the pooled specificity values were slightly higher for USGFNAB than in PGFNAB [0.77 (CI = 0.69, 0.85) and 0.73 (CI = 0.64, 0.81), respectively]. Matz et al. conducted a comparison between the two tests using the SROC curve [6]. Yet, no comparison using the diagnostic meta-regression and likelihood-ratio test was used, unlike this meta-analysis.
A study conducted by Taha et al. (2020) showed that the sensitivity value of USGFNAB was greater than that of PGFNAB, namely 69% and 52%, respectively. Meanwhile, the PGFNAB specificity value was slightly higher than USGFNAB at 94% and 91%, respectively [5]. However, this study was not included in this meta-analysis because the raw data displayed between the number of tests performed and those described was not suitable.
Studies by Choong et al. (2018) and Guo et al. (2015), were not included in this meta-analysis study due to differences in the primary criteria used in the 2 x 2 table. The results of the studies were different from the majority of previous studies. In these studies, the sensitivity and specificity values of PGFNAB were greater than USGFNAB. In the study by Choong et al., the sensitivity and specificity values were 86% vs 85.5% and 100% vs 99%, respectively [7]. In the study by Guo et al., the sensitivity and specificity values were 93% vs 90% and 96% vs 67%, respectively [12].
The benign criteria was used for indeterminacy/ AUS/ FLUS/ FN, suspicion of malignancy as the criteria for malignancy. It aimed to create a 2 x 2 table and determine true positive, false positive, true negative and false negative values in the index test column. For the gold-standard column, the histo-pathological results of the surgery were divided into benign and malignant. In some studies, indeterminant groups were classified as benign, and some were categorized as malignant. If it is included in the malignant criteria in the independent group, it can increase the false positive number on the result [6, 19].
Some previous studies suggest that USGFNAB is obviously preferable in patients with non-palpable or difficult to palpate nodule, predominantly cystic nodules with a small solid component and non-diagnostic PGFNAB, whether USGFNAB should be preferentially used for all palpable nodules is not clear.1,14 However, in this meta-analysis, the size of the nodules in the PGFNAB method were almost all larger than 1 cm and can be palpated. In the USGFNAB method, there were nodules less than 1 cm, nodules difficult to palpate, nodules not palpable and nodules greater than 1 cm. Therefore, the results of this meta-analysis found that USGFNAB is preferable for all palpable and non-palpable nodules.
The inadequacy number of PGFNAB method and the USGFNAB method were 14.6% and 9%, respectively. From these results, there was a significant difference between the two with a P = <0.0001, suggesting that the USGFNAB method had better results compared to PGFNAB. These results were consistent with the study by Matz et al., in which inadequacy rate of PGFNAB was 14.7% and USGFNAB was 8.4% [6]. Moreover, in a meta-analysis carried out by Gharib et al, in which more than 18000 cases were evaluated, the inadequacy rates of FNAB was 17% [8].
The occurrence of inadequate material after a biopsy may be caused by several factors including: nodule size; number of aspiration times during FNAB; operator factors; and the results' definition, which were inadequate in each study [13, 19]. Some studies have suggested that the adequate rate of biopsy results increased with increasing nodule size [13, 16, 25]. Aspiration during FNAB was recommended 2-4 times aspiration per one nodule [26, 29].
The quality of the main outcome of this meta-analysis was assessed based on the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach. It included the risk of bias, imprecision, inconsistency, indirectness, and publication bias. Each section was assessed, one-point reduction for any significant findings and two-points reduction for very significant findings or no serious findings (not reduced). The results of the quality assessment were divided into high, moderate, low, and very low. The results of the assessment can be seen in Table 5.
Table 5
Summary of Findings for the Diagnostic Accuracy of PGFNAB vs USGFNAB
Outcome
|
Number of subjects
(number of studies)
|
Pooled effect estimates
|
Quality of evidence (GRADE)
|
Summary of evidence quality
|
All ages
|
2382 (11 studies)
|
PGFNAB:
Sn 76 % (95 % CI, 49-89 %)
Sp 77 % (95% CI, 56-95 %)
AUC = 0.827
USGFNAB:
Sn 90 % (95 % CI, 81-95 %)
Sp 80 % (95% CI, 66-89 %)
AUC = 0.92
|
⨁⨁⨁◯
There is heterogeneity
|
Sufficient
|
* Description related to the evidence quality |
⨁⨁⨁⨁: High. The authors are confident that the effect obtained in this meta-analysis is an effect which accurately happened
⨁⨁⨁◯: Sufficient. The authors are reasonably confident that the effect obtained in this meta-analysis is an approximate actual effect, but there is still a possibility that there may be a substantial effect difference happened
⨁⨁◯◯: Low. The authors have limited confidence in the effect obtained in this meta-analysis. The actual effect could be significantly different from the effect obtained in this meta-analysis.
⨁◯◯◯: Very low. The authors are not sure of the effect obtained. This meta-analysis might have the same as the actual effect.
The weaknesses of this study is that several studies did not display the results entirely, so that complete data cannot be obtained to make 2 by 2 contigency tables according to the research criteria. Therefore, no intersection point for measuring the output parameters was included in this meta-analysis. Also, heterogeneity is still present in this meta-analysis.