Establishment of a nomogram prediction model for severe primary lower limb lymphedema

doi:10.21203/rs.3.rs-4903908/v1

Download PDF

Article

Establishment of a nomogram prediction model for severe primary lower limb lymphedema

https://doi.org/10.21203/rs.3.rs-4903908/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background

The International Society of Lymphology (ISL) guidelines have established grading criteria for primary lower limb lymphedema (PLEL), but there is a lack of model on a unified standard for assessing the severity of the disease.

Purpose

The aim of this study was to establish and validate a predictive model for evaluating severe PLEL.

Methods and Materials:

This retrospective study included 226 patients with unilateral PLEL from 2018 to 2023, who were divided into non-severe (143 cases) and severe (83 cases) groups according to the ISL grading criteria. The two groups of patients had a total of 26 MRI and 15 clinical features recorded. One-way ANOVA was performed first, followed by multi-factor ANOVA, and logistic regression was used to construct a nomogram prediction model. The model’s performance was evaluated via the area under the receiver operating characteristic (ROC) curve (AUC), decision curve analysis, and internal validation.

Results

The predictive model identified six independent risk factors associated with the severity of PLEL, including the parallel line sign, crescent sign, longitudinal range, band sign thickness, fat area, and fat diameter. The nomogram model established based on the above six factors predicts a training set AUC of 0.908 (95% CI: 0.868–0.947) for severe PLEL, with a sensitivity of 0.868, specificity of 0.832, accuracy of 0.845, precision of 0.75. The AUC of the validation set was 0.891 (95% CI: 0.847 ~ 0.935), the sensitivity was 0.831, the specificity was 0.825, the accuracy was 0.827, the precision was 0.734. In decision curve analysis, more net benefit can be achieved when the threshold probability is between 1% and 90%.

Conclusions

The severity risk prediction model based on MRI and clinical practice has good discriminatory power and accuracy in evaluating the severity of PLEL which can provide a reference for individualized clinical prediction of PLEL.

Health sciences/Risk factors

Health sciences/Health care/Medical imaging/Magnetic resonance imaging

lymphedema

MRI

grade

prediction model

nomogram

Lymphedema is a chronic disease caused by the disproportionate production and entry of lymphatic fluid into the circulatory system. Owing to the abnormal function of lymphatic vessels/lymph nodes, lymphatic fluid accumulates in the interstitium and stimulates abnormal proliferation of adipose tissue and fibrous connective tissue, leading to swelling of subcutaneous soft tissue and abnormal thickening of the limbs[1]. Depending on the aetiology, the disease can be divided into primary lymphedema (PLE) and secondary lymphedema. PLE is relatively rare and is caused by congenital underdevelopment of lymphatic vessels/nodes, whereas secondary lymphedema is common and results from damage or blockage of lymphatic vessels due to trauma, infection, tumours, and other reasons. The most common cause is postoperative malignant tumours[2].

The International Society of Lymphology (ISL) guidelines state that the grading of lymphedema is based on the difference in volume between the affected limb and the healthy side/healthy side, with 5% -20% being mild, 20% -40% being moderate, and > 40% being severe[2]. The commonly used calculation methods in practical clinical work include the water displacement method and the limb circumference measurement method. However, the above two methods for measuring volume are time-consuming and laborious and have certain errors in clinical applications.

Currently, most prediction models of lymphedema at home and abroad are used to predict the risk of lymphedema in cancer patients after surgery[3–8]. To the best of our knowledge, no scholars have yet used predictive models to evaluate the severity of primary lower limb lymphedema (PLEL). The aim of this study was to establish and validate a predictive model for the severity of PLEL using a nomogram-based on clinical and imaging data.

Patient

A retrospective analysis of 817 patients with PLEL in A Hospital from January 2018 to December 2023 was performed. The patients were diagnosed with PLEL through clinical and lymphoscintigraphy and oedema without any cause or inducement. A total of 365 patients with bilateral oedema, 220 patients with clinical or incomplete MRI data, and six patients with Klippel-Trenaunay syndrome were excluded from the study. Finally, 226 patients were enrolled. The flowchart for the inclusion and exclusion criteria is shown in Fig. 1. Six planes were selected: the ankle, lower third of the calf, upper third of the calf, knee joint, lower third of the thigh, and upper third of the thigh. The following formula was used for calculating limb volume[9]: \(\:\frac{1}{3}\)πh(R₁² + r₁² + R₁×r₁)+\(\:\frac{1}{3}\)πh༈R₂²+r₂²+R₂×r₂)+…+\(\:\frac{1}{3}\)πh༈R_n²+r_n²+R_n×r_n) (R = half of the measured segment's lower landmark point corresponding to the limb circumference/2π; r = half of the measured segment's upper landmark point corresponding to the limb circumference/2π; h = the distance between the two landmark points of the measured segment) for limb volume calculation. Grading was based on the formula [(affected limb volume—healthy limb volume)/healthy limb volume] ×100%. In accordance with the 2020 ISL grading criteria, the enrolled patients were graded as follows: 5% to < 20% as mild, > 20% to < 40% as moderate, and > 40% as severe. The enrolled patients were divided into a non-severe group (mild and moderate) and a severe group. Clinical features include sex, onset at birth, age of onset, disease duration, admission age, nondepressed swelling, skin pigmentation, high skin temperature, skin redness, ulcers, poor skin elasticity, significant skin thickening, verrucous hyperplasia, pain, and clinical stage.

This retrospective study was approved by the Ethics Committee of Beijing Shijitan Hospital affiliated with Capital Medical University. All methods were carried out in accordance with the tenets set by the declaration of Helsinki. Because it is a retrospectivestudy, the requirement for informed consent to review medical records and images has been waived.

MRI

All MRI examinations in this study were performed using a 1.5T MR scanner (Ingenia, Philips, Netherlands) with a 16-channel body coil. The patient is lying on their back with advanced feet. The scanning range extends from the level of the femoral head to the entire foot. The transverse and coronal planes were scanned via a short-term reversal recovery sequence with the following scanning parameters: TR, 5298 ms/5649 ms; TE, 80 ms; FOV, 400 mm × 220 mm/360 mm × 400 mm; layer thickness, 6 mm/7 mm; interlayer spacing, 0.6 mm; voxels, 1.5 mm3; IR, 165 ms/160 ms; and NEX, once.

Image analysis

Two radiologists with more than five years of experience in lymphedema diagnosis, blinded to the clinical indicators, diagnosis and grading, retrospectively reviewed the MRI findings separately. In cases of disagreement, a consensus was reached via joint discussion with a radiologist with 30 years of experience.

The imaging signs included the following: (1) lymphedema range (longitudinal range: only thigh, only calf and ankle, the whole lower extremity; transverse range: <75%, ≥ 75%). (2) MRI lymphedema signs (parallel lines sign, grid sign, honeycomb sign, band sign, lymph lake sign, crescent sign, and nebula sign); (3) lymphedema measurements: total diameter; total circumference; total area; diameter, circumference and area of bone; diameter, circumference and area of muscle; diameter and area of fat; diameter and area of soft tissue; skin thickness; band sign thickness; and crescent sign thickness. The various signs of lymphedema on MRI are defined as follows[10] :i) parallel lines sign (Fig. 2) is manifested as many thin line shadows with a width of 1-2mm, which are parallel to the superficial fascia and do not interweave with each other; ii) grid sign is a fine mesh structure formed by the intersection of multiple lines whose width are less than 3mm, with the width-to-length ratio of each mesh being less than 2/3; iii) honeycomb sign is a honeycomb structure formed by multiple lines with thickness > 3mm in more than two directions, and the width-length ratio of each partition is greater than 2/3; iv) band sign, namely suprafascial effusion, is the effusion surface of the superficial fascia observed in the axial position, which is usually in the form of a wide strip of ≤ 10 mm and shows high signal in the STIR sequence; v) crescent sign(Fig. 3) is subfascia effusion, which is the effusion between superficial fascia and muscle observed on axial plane. It is like crescent structure and shows high signal in STIR sequence; vi) lymph lake sign(Fig. 3) is a large unstructured edema area located in subcutaneous soft tissue. vii) nebula sign is manifested as a large number of small tubular shadows of varying sizes, diameters < 5mm and fuzzy boundaries distributed on the medial side of limbs. Positive signs were scored as 1, otherwise 0.

Statistical analysis:

The Deepwise Medical Multimodal Research Platform version 2.2.1 (https://keyan.deepwise.com. Beijing China Deepwise Bolian Technology Co., Ltd.) was employed for this study. Measurement data that conform to a normal distribution are described as the mean ± standard deviation (x ± s). In contrast, measurement data that do not conform to a normal distribution are described as the median and interquartile range. Measurement data with a normal distribution were compared via the independent sample t-test, and measurement data with a non-normal distribution were compared via the Wilcoxon rank sum test. Count data are described via examples and percentages (%) and were compared via the chi-square test or Fisher's exact probability method. First, univariate and multivariate logistic regression were performed to screen for risk factors for severe PLEL. Factors with P < 0.1 in the univariate analysis were included in the multivariate logistic regression analysis, whereas the factors with P < 0.05 in the multivariate analysis were considered independent risk factors. We conducted a feature correlation analysis and removed the features with correlation coefficients less than 0.9. The features analysed through feature correlation were used for further model training. After completing the feature correlation analysis, we chose L1 as the feature selection method. Based on L1 regularisation, feature selection (linear models penalised with the L1 norm) establishes a linear model on the training data and obtains a sparse coefficient matrix. Features with larger absolute values of coefficients are more likely to be preserved. Parameter C is used to control the strictness of feature selection, with fewer features retained as C decreases. To construct an effective and robust model based on the data included in this study, five-fold cross-validation was used with a random seed of 1. The cross-validation method divides the dataset into K mutually exclusive and similarly sized subsets while maintaining consistency in the data distribution during the partitioning process. The union of K-1 subsets is used as the training set each time, and the remaining subsets are used as the testing set for k training sessions, and the average of k results is calculated. In this study, we used a machine learning model called logistic regression to evaluate the clinical efficacy of the prediction model based on the area and clinical net benefit under the receiver operating characteristic (ROC) curve.

Clinical and imaging baseline data

A total of 226 patients were included in this study, including 129 females and 97 males. The training set and validation set both consisted of 226 samples. There were 143 patients in the non-severe and 83 patients in the severe group. The two groups had statistically significant differences (P < 0.05) in five clinical indicators, including disease course, high skin temperature, red skin, poor skin elasticity, and clinical stage. The eight imaging indicators, including longitudinal involvement range, axial involvement ≥ 75%, parallel line sign, grid sign, honeycomb sign, band sign, lymph lake sign, and crescent sign, were significantly different (P < 0.05) between the two groups. In addition, the ten imaging measurement indicators, including mesothelial fat transverse diameter, subcutaneous soft tissue transverse diameter, total circumference, total transverse diameter, subcutaneous fat area, subcutaneous soft tissue area, skin thickness, superficial fascia fluid thickness, superficial fascia fluid thickness, and total area, were significantly different (P < 0.05) between the two groups (Table 1). After univariate and multivariate logistic regression analyses, the results revealed that the parallel line sign, crescent sign, longitudinal involvement range, band sign width, subcutaneous fat area, and subcutaneous fat transverse diameter were independent risk factors for distinguishing severe PLEL, and a clinical model was constructed based on these findings as shown in Table 2.

Table 1

Baseline Clinical and Imaging Data
	All (226)	Non-severity(n = 143)	Severity(n = 83)	P
Clinical characteristic
Gender				0.464
Male	97(42.9%)	64(44.8%)	33(39.8%)
Female	129(57.1%)	79(55.2%)	50(60.2%)
Onset at birth	36(15.9%)	19(13.3%)	17(20.5%)	0.154
Age of onset, years	16(6–27)	16(7.5–26.5)	13(1.5–30)	0.682
Course of disease, years	7(2–16)	5(2–15)	10(3.5–21)	0.002^*
Age of admission	27(16–40)	25(16.5–36)	32(16–46)	0.106
Non-squamous oedema	42(18.6%)	24(16.8%)	18(21.7%)	0.361
Hyperpigmentation	22(9.7%)	12(8.4%)	10(12.0%)	0.371
High skin temperature	45(19.9)	22(15.4%)	23(27.7%)	0.025^*
Skin redness	65(28.8%)	30(21.0%)	35(42.2%)	0.001^*
Ulcer	18(8.0%)	11(7.7%)	7(8.4%)	0.843
Poor skin elasticity	156(69.0%)	90(62.9%)	66(79.5%)	0.009^*
Skin thickening	175(77.4%)	106(74.1%)	69(83.1%)	0.118
Warty hyperplasia	6(2.7%)	2(1.4%)	4(4.8%)	0.266
Pain	13(5.8%)	7(4.9%)	6(7.2%)	0.667
Staging				0.003^*
Stage I	81(35.8%)	63(44.1%)	18(21.7%)
Stage II	99(43.8%)	55(38.5%)	44(53.0%)
Stage III	46(20.4%)	25(17.5%)	21(25.3%)
Imaging manifestations
Vertical range				0.000^*
Whole lower limb	180(79.6%)	100(69.9%)	80(96.4%)
Only calf and ankle	42(18.6%)	39(27.3%)	3(3.6%)
Only thigh	4(1.8%)	4(2.8%)	0(0)
Parallel lines sign	85(37.6%)	76(53.1%)	9(10.8%)	0.000^*
Grid sign	147(65%)	103(72.0%)	44(53.0%)	0.004^*
Honeycomb sign	169(74.8%)	89(62.2%)	80(96.4%)	0.000^*
Band sign	185(81.9%)	107(74.8%)	78(94.0%)	0.000^*
Lymph lake sign	39(17.3%)	6(4.2%)	33(39.8%)	0.000^*
Crescent sign	65(28.8%)	18(12.6%)	47(56.6%)	0.000^*
Nebula sign	13(5.8%)	10(7.0%)	3(3.6%)	0.450
Expansion tube sign	14(6.2%)	10(7.0%)	4(4.8%)	0.513
Transversal range ≥ 75%	167(73.9%)	89(62.2%)	78(94.0%)	0.000^*
Lymphedema measurements
Total Diameter	13(11.0-15.5)	12(10.7–14.2)	14.9(12.7–16.9)	0.000^*
Bone Diameter	2.4(2-2.9)	2.4(2-2.9)	2.4(2.0–3)	0.739
Muscle diameter	8.8 ± 2.1	8.7 ± 2.1	8.9 ± 2.2	0.262
Fat diameter	4.25(3.2–5.7)	3.5(2.8–4.4)	5.7(4.6–7.1)	0.000^*
Soft Tissue Diameter	10.4(8.6–12.8)	9.5(8.2–11.7)	12.3(10.6–14.4)	0.000^*
Total circumference	40.2(34-47.6)	37.2(33-42.8)	44.9(39.2–51.4)	0.000^*
Bone circumference	7.2(6.4–8.2)	7.2(6.4–8.2)	7.2(6.4–8.3)	0.927
Muscle circumference	27.9 ± 7.4	27.9 ± 7.2	27.9 ± 7.9	0.602
Total area	122.4(90.2-166.5)	107.5(84.8-133.9)	157.1(120.6-201.7)	0.000^*
Bone area	4.2(3.3–5.3)	4.5(3.3–5.5)	4(3.3–5.2)	0.358
Muscle area	57.9(38.3–77.5)	55.1(37.5–78.4)	62.2(40.1–75.4)	0.588
Fat area	64(43.5–97.8)	52.6(38.3–70.6)	93.2(68-131.9)	0.000^*
Soft tissue area	118.6(86.4-161.2)	102.8(81.7-128.4)	151.9(117.4-196.1)	0.000^*
Skin thickness	3(2–5)	3(2–4)	5(3–6)	0.000^*
Band sign thickness	5(3–7)	4(2-5.5)	8(6–10)	0.000^*
Crescent sign thickness	0(0–2)	0(0–0)	2(0–6)	0.000^*
^{*The difference between the two groups is statistically significant}

Table 2

Multivariate logistic regression analysis of clinical and imaging indicators
	OR (95% CI)	P
Vertical range	9.462(2.226–40.218)	0.029
Parallel lines sign	0.318(0.125–0.810)3	0.012
Crescent sign	1.82(0.764–4.338)	0.039
Fat diameter	2.264(1.412–3.630)	0.002
Fat area	9.462(0.967–1.005)	0.008
Band sign thickness	1.25(1.100–1.420)	0.000

The establishment of the nomogram

The parallel lines sign, crescent sign, longitudinal range, band sign width, fat area, and fat diameter are independent risk factors for severe PLEL. A clinical model composed of these factors was used to create a nomogram (Fig. 4). The model demonstrated excellent performance in distinguishing between severe and non-severe PLELs in both the training and validation groups. The AUCs of the training set and validation set are shown in Fig. 5. The AUC of the training set was 0.908 (95% CI: 0.868 ~ 0.947), the sensitivity was 0.868, the specificity was 0.832, the accuracy was 0.845, the precision was 0.75, and the F1 score was 0.804. The AUC of the validation set was 0.891 (95% CI: 0.847 ~ 0.935), the sensitivity was 0.831, the specificity was 0.825, the accuracy was 0.827, the precision was 0.734, and the F1 score was 0.780, as shown in Table 3. In decision curve analysis, more net benefit can be achieved when the threshold probability is between 2% and 90%(as shown in Fig. 6). The waterfall plot shows the visualization results of the training and validation sets, demonstrating the strong discriminative ability of the model in distinguishing between severe and non severe patients(Fig. 7).

Table 3

Performance of different models in the training and validation cohorts
	AUC	sensitivity	specificity	accuracy	precision	F1 score
training set	0.908	0.868	0.832	0.845	0.75	0.804
validation set	0.891	0.831	0.825	0.827	0.734	0.780

Currently, prediction models for lymphedema at home and abroad mostly predict the risk of lymphedema in cancer patients after surgery. Different models have been established based on clinical data, laboratory tests and imaging indicators to predict the risk of early secondary lymphedema in breast cancer or gynaecological cancer patients[3–8, 11, 12]. Xuan Tung Rinh[8] successfully developed a machine-learning model for predicting lymphedema using blood and treatment data and proposed that future research focus on predicting the stages of lymphedema. This study was based on previous research and proposed a PLEL model based on clinical and imaging data that can effectively distinguish between non-severe and severe PLEL, thereby improving the diagnostic performance of clinical models. Therefore, we constructed a nomogram prediction model that includes clinical features, imaging features, and lower limb oedema measurements. This non-invasive method can be used to evaluate the severity of PLE preoperatively, providing a reliable approach for the next treatment plan.

PLEL has two different progression directions, namely, from top to bottom and bottom to top, with the former being more common; the former is caused by the structural or functional insufficiency of peripheral superficial lymphatic vessels, leading to the leakage of lymphatic fluid along the diseased ducts, whereas the latter is caused by the structural or functional insufficiency of central lymphatic vessels or lymphatic trunks (such as the thoracic duct, cisterna chyli, lumbar and iliac lymphatic trunks), leading to the downwards reflux of lymphatic fluid, causing the lymphatic chyle to reflux to the lower limbs [13, 14]. The results of this study revealed that 96.4% of severe patients experienced complete lower limb involvement(Fig. 8), and this indicator is an independent risk factor for severe PLEL. It is speculated that severe patients may have both peripheral and central lymphatic vessel structural and functional deficiencies.

Previous studies, both domestic and international, as well as our research, have found that parallel line signs, grid signs, honeycomb signs, band signs, lymph lake signs, and crescent signs on MR images of patients with lymphedema are closely related to the severity of lymphedema[10, 15, 16]. The results of this study confirmed that these signs significantly differed in terms of imaging indicators between the two groups. This study revealed that the parallel line sign, crescent sign, and band sign thickness are independent risk factors for distinguishing non-severe and severe PLEL. The parallel line sign may be mild dilation of subcutaneous lymphatic capillaries and a small amount of lymphatic exudate, which accumulates and stimulates the proliferation of adipose tissue linearly. Some researchers believe that the parallel line sign is a characteristic manifestation of mild lymphedema[10], which is consistent with the results of our study. Aström KG et al.[15] suggested that subfascial effusion (i.e., the crescent sign in this study) can be used as an indicator to determine the severity of lymphedema. The crescent sign is a vicious cycle formed by the continuous leakage of subcutaneous lymphatic fluid and increased local inflammatory factors that damage lymphatic vessels in severe patients. The accumulated lymphatic fluid gradually spreads to the fascia and muscles. Cellina et al.[16] found that suprafascial effusion (i.e., the band sign in this study) can be used as an indicator to determine the clinical stage of lymphedema. This study revealed a statistically significant difference in the band sign between the two groups, which can be used as an indicator to determine the severity of lymphedema. However, band sign is not an independent risk factor for severe PLEL, and band sign thickness is an independent risk factor for severe PLEL. We speculate that the band sign thickness and the presence of the crescent sign are independent risk factors because of the different anatomical locations of fluid accumulation in each sign. Hauck G[17] confirmed the existence of a "low resistance pathway" for fluid transport from capillaries to lymphatic vessels along connective tissue fibres. The distribution of collagen and elastic fibres in the superficial fascia can guide lymphatic fluid to flow in the correct direction. Fascia is a dense connective tissue with a strong impedance that blocks fluid diffusion. If the superficial fascia changes, lymphatic drainage will be affected. The fluid accumulates on the superficial fascia and enters below the subfacia upon reaching certain levels, forming a crescent sign. The quantitative change in the band sign causes a qualitative change in the crescent sign, so the former measurement can more objectively reflect the severity of lymphedema.

In addition, we also included measurements of four components (skin, bone, fat and soft tissue) of the lower limbs. The total diameter, fat diameter, soft tissue diameter, total circumference, total area, fat area, soft tissue area, skin thickness, band sign thickness, and crescent sign thickness were closely related to the grade, and there was a statistically significant difference in the imaging indicators between the two groups. Fat area and fat diameter were independent risk factors for severe PLEL, which confirms that the disease does not involve bones or muscles. This result is similar to the findings of Liu et al.[18]. Thickened subcutaneous adipose tissue plays a major role in increasing limb volume in patients with primary lower limb lymphedema. This feature is also the primary difference between lymphedema and venous oedema. These findings also indicate that the lesions in the fat layer of lymphedema are the most important areas for determining the severity of lymphedema. Patients with primary lymphedema experience slow lymphatic flow due to congenital lymphangiodysplasia and functional decline [19]. The exudate of lymphatic fluid is usually confined to the surface membrane spaces of the skin and subcutaneous tissue without involving deeper muscles; this leads to a gradual increase in subcutaneous fat layer oedema. The lymphatic flow of patients with severe lymphedema is further slowed, stagnant in the subcutaneous fat layer, and stimulates fat production and deposition, leading to fibroblast activation and excessive growth of connective tissue. Enlarged adipose tissue and fibrous tissue accumulate around the subcutaneous soft tissue, leading to further swelling of the subcutaneous tissue in the limbs[20].

Deficiencies

Staging reflects the pathological and physiological conditions of lymphedema and the process of disease occurrence and development. Grading also reflects the degree of swelling in the affected limbs of patients with lymphedema and is also the main basis for the clinical selection of surgical procedures. Therefore, this paper only discusses lymphedema grade.

This paper only discusses the establishment of a nomogram prediction model for severe primary lower limb lymphedema preliminarily. Our further research will establish a three-class prediction model for mild, moderate, and severe lymphedema.

This study focused on PLEL, which is rarer than secondary lymphedema. PLEL is a type of congenital disease and is less affected by factors than secondary lymphedema. Therefore, this study takes PLEL as the starting point, and subsequent studies will include patients with secondary lymphedema.

In previous studies, we discussed surgical treatment for patients with primary lower limb lymphedema. In this study, we discussed only predictive models and did not address treatment.

The PLEL severity risk prediction model, which is based on clinical and MR images, has good discriminative power and accuracy in evaluating conditions. This model can provide a reference for individualized clinical prediction of severe and non-severe PLEL patients and is of significance for individualized precision treatment.

This study was supported by National Natural Science Foundation of China (No. 61876216)

No conflict of interest exits in the submission of this manuscript, and manuscript is approved by all authors for publication.

Author Contribution

Author contribution:1. Guarantor of integrity of the entire study: X.L., Y.Z., Y.Y., W.S. and R.W.2. Study concepts and design: X.L., M.L., Y.Y. and R.W.3. Literature research: X.L., Y.Z., M.L. and J.G. 4. Clinical studies: X.L., Y.Z., M.L., J.G. and W.S.5. Experimental studies / data analysis: X.L., J.W., J.R. and M.L.6. Statistical analysis: X.L., J.W., J.R. and M.L.7. Manuscript preparation: X.L., M.L., J.G, and R.W.8. Manuscript editing: X.L., Y.Y., W.S. and R.W.

Acknowledgement

We sincerely thank the Department of LymphSurgery at Beijing Century Altar Hospital affiliated with Capital Medical University. They have made significant contributions to the diagnosis, treatment, and care of patients with lymphedema and lymphangitis.

Data Availability

The datasets generated and/or analyzed during the current study are not publicly available due patient privacy regulations but are available from the corresponding author on reasonable request.

Brunelle, C. L. & Taghian, A. G. Lymphoedema screening: setting the standard. Br. J. Cancer. 123 (1), 1–2. https://doi.org/10.1038/s41416-020-0848–0 (2020).
Executive Committee of the International Society of Lymphology. The diagnosis and treatment of peripheral lymphedema: 2020 Consensus Document of the International Society of Lymphology. Lymphology. 53 (1), 3–19 (2020).
Wei, X. et al. Developing and validating a prediction model for lymphedema detection in breast cancer survivors. Eur. J. Oncol. Nurs. 54, 102023. https://doi.org/10.1016/j.ejon.2021.102023 (2021).
Fu, M. R. et al. Machine learning for detection of lymphedema among breast cancer survivors. Mhealth. 4 https://doi.org/10.21037/mhealth.2018.04.02 (2018). :17. Published 2018 May 29.
Armer, J. M., Radina, M. E., Porock, D. & Culbertson, S. D. Predicting breast cancer-related lymphedema using self-reported symptoms. Nurs. Res. 52 (6), 370–379. https://doi.org/10.1097/00006199-200311000-00004 (2003).
Wang, L. et al. A Scoring System to Predict Arm Lymphedema Risk for Individual Chinese Breast Cancer Patients. Breast Care (Basel). 11 (1), 52–56. https://doi.org/10.1159/000443491 (2016).
Penn, I. W. et al. Risk factors and prediction model for persistent breast-cancer-related lymphedema: a 5-year cohort study. Support Care Cancer. 27 (3), 991–1000. https://doi.org/10.1007/s00520-018-4388–6 (2019).
Trinh, X. T. et al. Development of predictive models for lymphedema by using blood tests and therapy data. Sci. Rep. 13 (1), 19720. https://doi.org/10.1038/s41598-023-46567–1 (2023). Published 2023 Nov 13.
Brorson, H. & Höijer, P. Standardised measurements used to order compression garments can be used to calculate arm volumes to evaluate lymphoedema treatment. J. Plast. Surg. Hand Surg. 46 (6), 410–415. https://doi.org/10.3109/2000656X.2012.714785 (2012).
Li, X. et al. Value of the short time inversion recovery sequence of magnetic resonance imaging in the staging of Klippel-Trenaunay syndrome complicated with lymphedema. J. Vasc Surg. Venous Lymphat Disord. 12, 101746. https://doi.org/10.1016/j.jvsv.2023.101746 (2024).
Nomura, Y. et al. Computer-aided diagnosis for screening of lower extremity lymphedema in pelvic computed tomography images using deep learning. Sci. Rep. 13 (1), 16214. https://doi.org/10.1038/s41598-023-43503–1 (2023). Published 2023 Sep 27.
Yaghoobi Notash, A., Yaghoobi Notash, A., Omidi, Z. & Haghighat, S. Prediction of lymphedema occurrence in patients with breast cancer using the optimized combination of ensemble learning algorithm and feature selection. BMC Med. Inf. Decis. Mak. 22 (1), 195. https://doi.org/10.1186/s12911-022-01937-z (2022). Published 2022 Jul 25.
Kazemzadeh, G. H., Sadeghi, R., Ebrahimi, E. & Rad, M. A. A successful experience in managing a chylous reflux: importance of lymphoscintigraphy. Clin. Nucl. Med. 39 (5), 485–487. https://doi.org/10.1097/RLU.0000000000000418 (2014).
Liu, N. F., Yan, Z. X., Wu, X. F. & Luo, Y. Magnetic resonance lymphography demonstrates spontaneous lymphatic disruption and regeneration in obstructive lymphedema. Lymphology. 46 (2), 56–63 (2013).
Aström, K. G., Abdsaleh, S., Brenning, G. C. & Ahlström, K. H. MR imaging of primary, secondary, and mixed forms of lymphedema. Acta Radiol. 42, 409e416. https://doi.org/10.1080/028418501127346909 (2001).
Cellina, M. et al. Noncontrast MR lymphography in secondary lower limb lymphedema. J. Magn. Reson. Imaging. 53, 458e466. https://doi.org/10.1002/jmri.27328 (2021).
Hauck, G. & Castenholz, A. Beitrag prälymphatischer Strukturen zur Lymphdrainage [Contribution of prelymphatic structures to lymph drainage]. Z. Lymphol. 16 (1), 6–9 (1992).
Liu, M. et al. Quantitative evaluation of primary lower extremity lymphedema staging using MRI: a preliminary study. Quant. Imaging Med. Surg. 13 (8), 4839–4851. https://doi.org/10.21037/qims–22–795 (2023).
Grada, A. A., Phillips, T. J. & Lymphedema Pathophysiology and clinical manifestations. J. Am. Acad. Dermatol. 77 (6), 1009–1020. https://doi.org/10.1016/j.jaad.2017.03.022 (2017).
Dayan, J. H. et al. Regional Patterns of Fluid and Fat Accumulation in Patients with Lower Extremity Lymphedema Using Magnetic Resonance Angiography. Plast. Reconstr. Surg. 145 (2), 555–563. https://doi.org/10.1097/PRS.0000000000006520 (2020).

No competing interests reported.

Download PDF

Editor invited by journal
27 Aug, 2024
Submission checks completed at journal
27 Aug, 2024
First submitted to journal
12 Aug, 2024

You are reading this latest preprint version

Establishment of a nomogram prediction model for severe primary lower limb lymphedema

Status:

Version 1

Abstract

Background

Purpose

Methods and Materials:

Results

Conclusions

Figures

Introduction

Materials and Methods

Patient

MRI

Image analysis

Statistical analysis:

Results

Clinical and imaging baseline data

The establishment of the nomogram

Discussion

Conclusion

Declarations

Author Contribution

Acknowledgement

Data Availability

References

Additional Declarations

Status:

Version 1