Simple clinical predictors for making directive decisions in osteoporosis screening for women: a cross-sectional study Authors: Jirapong Leeyaphan, Karn Rojjananukulpong, Piyapong Intarasompun, Yuthasak Peerakul Published: 2024 ABSTRACT AbstractBackgroundOsteoporosis screening strategies vary according to a country’s policies. Simple and highly accurate diagnostic predictors should be developed and utilized. This cross-sectional study aimed to determine the diagnostic performances of weight, body mass index (BMI), and age for osteoporosis. Moreover, this study proposes and validates new cutoff values for these indicators.MethodsWomen aged 50 years or older who underwent Dual-energy x-ray absorptiometry (DXA) and did not receive any treatment for osteoporosis were enrolled. Age, weight, and BMI were used to analyze the diagnostic models. New cutoff values were proposed using the maximum Youden index value.ResultsA total of 1598 women were classified into normal bone mineral density, osteopenia, and osteoporosis groups. The mean age, weight, and BMI were 73.2 years, 53.6 kg, and 23.2 kg/m2, respectively, in the osteoporosis group, which significantly differed from those of the other groups. The new cutoff values for diagnosing osteoporosis at any site were 57.4 kg for weight, 23.8 kg/m2 for BMI, and 72 years for age. The area under the receiver operating characteristic curve (AUC) for weight < 57.4 kg, BMI < 23.8 kg/m2, and age ≥ 72 years cutoff values were 0.664, 0.633, and 0.558, respectively. The weight cutoff value yielded a significantly higher AUC than the BMI and age cutoff values (P = 0.002 and P < 0.001, respectively).ConclusionsThis study proposes new cutoff values for simple clinical predictors to facilitate directive decision-making in osteoporosis screening in women. Weight, which are easily obtained in clinical practice, are the most effective predictors of osteoporosis screening.Keywords: Osteoporosis, Bone mineral density, Body weight, Body mass index, Age, Women Introduction The age structure of the global population has undergone significant changes, particularly in certain countries in the Asia–Pacific region. This demographic shift underscores the need for active revision of clinical practice guidelines (CPGs) to enhance healthcare management and alleviate the economic burden of age-related diseases the older population [1, 2]. The prevalence of aging societies is increasing worldwide, resulting in increased economic strain due to age-related illnesses [2–4]. Osteoporosis is an age- and sex-specific disease [5, 6]. The prevalence of osteoporosis in Asia–Pacific developed countries ranges from 10 to 30% for women aged 40 and older. Consequently, the incidence of osteoporotic fractures in adults aged 50 and older ranges between 500 and 1000 per 100,000 person-years [7]. According to the current CPGs for osteoporosis focus on three main aspects: screening, diagnosis, and management [8]. Dual-energy x-ray absorptiometry (DXA) is the primary tool used to assess bone mineral density (BMD) in several CPGs. A BMD T-score of ≤ − 2.5 standard deviations (SDs) is indicative of osteoporosis and warrants treatment intervention [9, 10]. In fact, DXA screening strategies vary depending on individual country-specific policies. In addition, many fracture risk assessment tools or osteoporosis risk assessment tools have been used in the screening part of the Asia–Pacific region osteoporosis CPGs [8]. However, various demographic and characteristic data recognized as risk factors for osteoporosis or osteoporotic fractures serve as indicators for DXA screening. Across Asia–Pacific region countries, low body mass index (BMI) or low body weight has been incorporated as an indicator in 16 of 18 guidelines for undergoing DXA [11–14]. Regarding age, different cutoff values have been employed as DXA screening criteria [12–16]. The cutoff values for making directive decisions in osteoporosis screening vary based on the diagnostic ability of the predictors. A broad range of cutoff values has been suggested for body weight. For instance, using 70 kg is recommended because of its high sensitivity (89% sensitivity and 38% specificity for lumbar spine osteoporosis and 94% sensitivity and 36% specificity for femoral neck osteoporosis) [17]. Alternatively, employing 51 kg is advocated because of its high specificity (22% sensitivity and 97% specificity for osteoporosis at any site) [18] whereas 57 kg is suggested because of its high discriminative ability for osteoporosis [19]. In addition, BMI, age, and weight have been chosen as osteoporosis screening criteria because of their association with a high prevalence of osteoporotic fracture [20–22]. However, individual guidelines have diverse criteria for decision-making regarding DXA screening. It is essential to prioritize either osteoporosis risk screening or fracture risk screening before undergoing DXA, with consideration given to the country-specific osteoporosis assessment tools [8, 23]. In Thailand, the osteoporosis CPG was initially established in 2010. This guideline utilized either women aged ≥ 65 years or those aged < 65 years with a BMI < 20 kg/m2 as cutoff values for making directive decisions in osteoporosis screening. However, body weight was not considered in the DXA screening criteria of the 2010 CPG [13]. Despite the CPG being updated in 2021, these simple clinical predictors retained the same cutoff points. The median BMI of patients with osteoporosis has been reported to be approximately 25 kg/m2 [24–26], which is slightly higher than the typical cutoff values used for classifying individuals at risk for osteoporosis [27, 28]. It is recommended that the BMI cutoff point be raised to better align with the characteristics of the osteoporosis population. We hypothesized that societal changes toward aging and urbanization may influence the development of osteoporosis. Consequently, re-evaluation of the new cutoff values of simple clinical indicators is warranted to enhance diagnostic performance in osteoporosis. This study aimed to evaluate the effectiveness of simple clinical indicators such as body weight, BMI, and age as screening tools for osteoporosis. Additionally, it introduces new cutoff values for these indicators to guide decision-making in osteoporosis screening for women. This study further compared the performance of these new cutoff values with those outlined in the 2021 CPG. Methods Data collection and study population Demographic and clinical data were collected and reviewed from electronic medical records and the DXA machine database of women aged 50 and older who underwent DXA examination between November 2017 and September 2023 at a tertiary care public hospital. Participants who had received osteoporosis treatment or had a history of lumbar spine or hip surgery were excluded from the study. Additionally, participants with incomplete lumbar spine and hip BMD data were excluded. The study protocol was approved by the institutional review board (S025h/66_ExPD). Bone mineral density measurement All the participants underwent BMD measurements using the same DXA machine (Discovery W; Hologic Inc., Bedford, MA, USA). The DXA machine followed the World Health Organization criteria for T-score interpretation derived from two sites: the lumbar spine (L1-4) and the hip (femoral neck and total hip). Osteoporosis was defined as T-scores ≤ − 2.5 SD, osteopenia as T-scores between − 1.0 and − 2.5 SD, and normal bone density as T-scores ≥ − 1.0 SD. Weight and height measurement Each participant’s body weight and height were measured on the day of BMD assessment. All the participants wore lightweight clothing and removed their shoes before undergoing weight and height measurements. A weighing scale with a precision of 0.1 kg and a height scale with a precision of 1 cm were used in this study. Height and weight were measured using an electronic stadiometer (NAGATA BW-1116MH; Nagata Scale Co. Ltd., Tainan, Taiwan). Statistical analysis The participants were categorized into three groups based on their highest baseline BMD T-score among the three BMD sites: normal, osteopenic, and osteoporotic. Participants were also divided into two groups: an osteoporosis group for participants with a T-score ≤ − 2.5 and a non-osteoporosis group for participants with a T-score > − 2.5. Descriptive statistics, including mean values with SDs or median values with the 25th–75th percentile ranges, were presented according to the data distribution. BMI was computed by dividing weight (in kilograms) by the square of body height (in meters). To compare clinical characteristics, the chi-square test, Student’s t-test, analysis of variance, and Mann–Whitney U test were used based on the type and distribution of the data. Fractional polynomial regression analysis was used to examine the relationships among weight, BMI, age, and BMD. Binary logistic regression analysis was used to assess the predictive ability of weight, BMI, and age for osteoporosis. The selected cutoff values derived from BMD at any BMD site were evaluated for overall model fit performance, discrimination, and calibration using the bootstrapping resampling technique with 1000 replications [29]. The osteoporosis discriminative abilities of weight, BMI, and age were evaluated using the area under the receiver operating characteristic curve (AUC) with a 95% confidence interval (CI). The new optimal cutoff values were determined based on the maximum value of the Youden index (sensitivity + specificity − 1) [30]. Sensitivity, specificity, positive predictive values, and negative predictive values were analyzed and reported. Finally, the AUC of each cutoff value was compared. According to the sample size calculations, to achieve the target sensitivity or specificity of the model with an identified osteoporosis prevalence of 38.3% [28] and a margin of error of 4% [31], targets of 61% sensitivity and 66% specificity for weight, as well as 57% sensitivity and 64% specificity for BMI, were set [32] according to the 95% CIs. Consequently, the sample sizes required to achieve the desired sensitivity or specificity targets were determined to be 1,492, 874, 1,537, and 897 participants, respectively. Furthermore, to compare the AUC of the two methods on the same subjects, assuming a 10% difference from the AUC of 0.6, with an alpha level of 0.05 and a power of 0.8, a minimum of 198 participants with osteoporosis were deemed necessary to detect a significant difference in AUC effects [33]. Therefore, the sample size of this study was 1,537 participants to achieve all objectives. Statistical analyses were conducted using STATA version 15 (StataCorp LLC, College Station, TX, USA), and a P-value < 0.05 was considered statistically significant. Results In this study, a total of 2179 women aged 50 years or older were initially included. Out of these, 440 participants were excluded due to prior osteoporosis treatment, 110 due to previous spinal or hip surgery, and 31 due to incomplete BMD data. Among 1739 women over the age of 50 without a previous diagnosis of osteoporosis, 141 cases were excluded due to lumbar spine surgery, hip surgery, or incomplete DXA data. We reviewed these cases and found that while lumbar spine and hip surgeries were not directly associated with fractures, prosthesis insertion did affect DXA results, leading to their exclusion. As a result, the final analysis comprised 1598 women (Fig. 1). The mean age (SD) of the participants was 71.8 (7.5) years. The most prevalent age group was 70–79 years, comprising 40.2% of the cases, while the least represented age group was 50–59 years, comprising 4.1%. Study flow diagram of participant enrollment. DXA, dual energy x-ray absorptiometry Prevalence of osteoporosis The prevalence of osteoporosis, osteopenia, and normal bone density at any site were 35.6%, 50.1%, and 14.3%, respectively. In the osteoporosis group, the mean age, weight, and BMI were 73.2 years, 53.6 kg, and 23.2 kg/m2, respectively. The clinical characteristics of the participants categorized by BMD results are summarized in Table 1, whereas Table 2 presents the characteristics of participants according to whether they are in the osteoporosis or non-osteoporosis group. Previous fracture history was shown in Table 1 and this study found a significant correlation between each individual factor and fracture occurrence. The prevalence of osteoporosis in the femoral neck, total hip, and lumbar spine was 28.1%, 8.9%, and 20.0%, respectively. Age, weight, height, and BMI were significantly different between the groups (P < 0.001) (Table 2). Fractional polynomial regression analysis demonstrated a positive relationship among BMD, weight, and BMI, whereas a negative relationship was observed between age and all BMD sites (Fig. 2). Table 1Demographic data and clinical characteristics of study participantsVariablesTotal (n = 1598)Osteoporosis (n = 569)Osteopenia (n = 800)Normal (n = 229)P-valueAge (years)71.8 ± 7.573.2 ± 8.171.2 ± 7.270.5 ± 6.8 < 0.001Weight (kg)58.4 ± 10.253.6 ± 9.059.4 ± 9.166.3 ± 10.9 < 0.001Height (cm)153.4 ± 5.7152.1 ± 6.0153.9 ± 5.4154.6 ± 5.7 < 0.001BMI (kg/m2)24.8 ± 4.123.2 ± 3.825.1 ± 3.727.7 ± 4.4 < 0.001BMD (g/cm2)Lumbar spine0.850 ± 0.1530.732 ± 0.1160.879 ± 0.1081.043 ± 0.117 < 0.001Femoral neck0.607 ± 0.1130.505 ± 0.0710.631 ± 0.0560.782 ± 0.083 < 0.001Total hip0.743 ± 0.1260.635 ± 0.0930.770 ± 0.0730.918 ± 0.096 < 0.001BMD (T-score)Lumbar spine − 1.4 ± 1.3 − 2.4 ± 1.0 − 1.1 ± 0.90.4 ± 1.0 < 0.001Femoral neck − 1.8 ± 1.1 − 2.8 ± 0.7 − 1.6 ± 0.5 − 0.2 ± 0.8 < 0.001Total hip − 1.0 ± 1.1 − 1.9 ± 0.8 − 0.7 ± 0.60.6 ± 0.8 < 0.001Any site − 1.9 ± 1.0 − 2.8 ± 0.6 − 1.6 ± 0.5 − 0.3 ± 0.7 < 0.001Previous fracture269 (16.8)137 (50.9)108 (40.2)24 (8.9) < 0.001Values are expressed as mean ± standard deviation or number (percentage)BMD bone mineral density, BMI body mass index Demographic data and clinical characteristics of study participants Values are expressed as mean ± standard deviation or number (percentage) BMD bone mineral density, BMI body mass index Table 2Demographic data and characteristics compared between study participants with and without osteoporosisVariablesFemoral neckTotal hipLumbar spineOsteoporosis (n = 449, 28.1%)Non-osteoporosis (n = 1149)P-valueOsteoporosis (n = 142, 8.9%)Non-osteoporosis (n = 1456)P-valueOsteoporosis (n = 320, 20.0%)Non-osteoporosis (n = 1278)P-valueAge (years)74.3 ± 8.070.8 ± 7.1 < 0.00176.0 ± 8.771.4 ± 7.3 < 0.00171.9 ± 8.171.8 ± 7.40.773Weight (kg)52.9 ± 9.160.5 ± 9.9 < 0.00150.0 ± 9.859.2 ± 9.9 < 0.00152.5 ± 9.159.8 ± 10.0 < 0.001BMI (kg/m2)23.0 ± 3.825.5 ± 4.0 < 0.00121.8 ± 4.125.1 ± 4.0 < 0.00122.6 ± 3.725.4 ± 4.0 < 0.001Values are expressed as mean ± standard deviationBMI body mass index Demographic data and characteristics compared between study participants with and without osteoporosis Values are expressed as mean ± standard deviation BMI body mass index The association between weight, BMI, age and BMD. BMD, bone mineral density; BMI, body mass index; CI, confidence interval Development of cutoff values Logistic regression models for osteoporosis status and individual predictors at various BMD sites are presented in Table 3. The pseudo-R2 value for the weighted model at the total hip was higher than those of the other models across all BMD sites (Table 3). Internal validation using the bootstrapping resampling technique demonstrated that weight had a Brier scale score of 12.4%, whereas age had a Brier scale score of 1.9%. The highest AUC was 0.710 (95% CI 0.685–0.737), which was derived from the weight predictor. Meanwhile, age yielded the lowest AUC of 0.579 (95% CI 0.550–0.611) (Table 4) (Fig. 3). Table 3Logistic regression model of the individual predictors at BMD sitesBMD sites and predictorsInterceptCoefficient (SE)Odds ratio (95% CI)P-valuePseudo-R2Femoral neckWeight4.078 − 0.089 (0.007)0.915 (0.902–0.928) < 0.0010.104BMI3.293 − 0.175 (0.017)0.839 (0.813–0.867) < 0.0010.069Age − 5.5620.064 (0.008)1.066 (1.050–1.082) < 0.0010.038Total hipWeight3.782 − 0.113 (0.011)0.894 (0.874–0.914) < 0.0010.126BMI3.440 − 0.247 (0.028)0.781 (0.739–0.825) < 0.0010.098Age − 8.2450.080 (0.012)1.084 (1.059–1.109) < 0.0010.051Lumbar spineWeight3.396 − 0.085 (0.008)0.918 (0.904–0.932) < 0.0010.092BMI3.430 − 0.202 (0.019)0.817 (0.787–0.849) < 0.0010.082Age − 1.5560.002 (0.008)1.066 (1.050–1.082) < 0.0010.000Any siteWeight4.211 − 0.084 (0.007)0.919 (0.908–0.931) < 0.0010.099BMI3.624 − 0.173 (0.015)0.841 (0.816–0.867) < 0.0010.071Age − 3.4350.039 (0.007)1.040 (1.026–1.055) < 0.0010.015BMD bone mineral density, BMI body mass index, CI confidence interval, SE standard error Logistic regression model of the individual predictors at BMD sites BMD bone mineral density, BMI body mass index, CI confidence interval, SE standard error Table 4Internal validation of the individual predictors at any BMD site via bootstrap procedures with 1000 replicatesVariablesOptimist AUC (95% CI)Brier scaled (%)E:O ratioCITL (95% CI)Slope (95% CI)Weight0.710 (0.685–0.737)12.41.0000.001 (− 0.111 to 0.114)1.005 (0.858–1.168)BMI0.683 (0.657–0.710)9.21.0000.001 (− 0.105 to 0.110)1.007 (0.838–1.213)Age0.579 (0.550–0.611)1.91.0000.001 (− 0.100 to 0.104)1.034 (0.733–1.598)AUC area under the receiver operating characteristic curve, BMD bone mineral density, BMI body mass index, CI confidence interval, CITL calibration-in-the-large, E:O ratio expected/observed ratioBrier scaled = 1 − (Brier score/Brier max) Internal validation of the individual predictors at any BMD site via bootstrap procedures with 1000 replicates AUC area under the receiver operating characteristic curve, BMD bone mineral density, BMI body mass index, CI confidence interval, CITL calibration-in-the-large, E:O ratio expected/observed ratio Brier scaled = 1 − (Brier score/Brier max) Model calibration plot of observed against expected probabilities by deciles of predicted risk with confidence intervals for the groupings that are displayed as well as a lowess line. CITL, calibration-in-the-large; E:O ratio, expected/observed ratio. (A) Weight. (B) Body mass index. (C) Age According to the Youden index, the cutoff values for osteoporosis prediction at any BMD site for weight, BMI, and age were suggested as 57.4 kg, 23.8 kg/m2, and 72 years, respectively. A cutoff weight value of < 57.4 kg at any BMD site yielded a Youden index of 0.3289, with a sensitivity of 62.6%, specificity of 70.3%, and positive predictive value of 79.2%. A cutoff age value of ≥ 65 years at any BMD site provided a Youden index of 0.0073, with a sensitivity of 89.5%, specificity of 11.3%, and positive predictive value of 35.8% (Table 5). The AUC of weight < 57.4 kg was significantly higher than the AUC of BMI < 23.8 kg/m2 and the AUC of age ≥ 72 years cutoff value (P = 0.002 and P < 0.001, respectively) (Table 6) (Fig. 4). Table 5Performance of selected cutoff values for osteoporosis prediction and different BMD sitesVariablesCutoff valueAUC (95% CI)Sensitivity (%)Specificity (%)PPVNPVLR + Youden indexFemoral neckWeight (kg) < 57.40.666 (0.641–0.691)60.273.185.141.82.230.3328BMI (kg/m2) < 23.80.631 (0.605–0.658)62.164.181.639.81.730.2628 < 20.00.562 (0.542–0.581)93.718.774.753.51.100.1236Age (years) ≥ 720.593 (0.567–0.620)60.857.936.179.11.310.1868 ≥ 650.516 (0.500–0.532)91.311.928.877.81.000.0323Total hipWeight (kg) < 52.50.689 (0.648–0.730)73.764.195.519.22.050.3778BMI (kg/m2) < 22.80.662 (0.621–0.704)69.163.495.116.71.890.3247 < 20.00.635 (0.596–0.675)92.634.593.531.21.410.2709Age (years) ≥ 760.633 (0.590–0.675)54.272.316.094.21.650.2655 ≥ 650.510 (0.485–0.535)90.911.29.192.60.970.0205Lumbar spineWeight (kg) < 56.60.653 (0.625–0.682)60.370.389.030.72.030.3064BMI (kg/m2) < 23.10.642 (0.613–0.672)69.159.487.232.51.480.2847 < 20.00.581 (0.557–0.605)93.422.882.946.51.140.1624Age (years) ≥ 720.493 (0.462–0.523)46.352.419.679.50.85 − 0.0014 ≥ 650.493 (0.473–0.513)87.810.719.877.80.94 − 0.0147Any siteWeight (kg) < 57.40.664 (0.641–0.688)62.670.379.251.02.110.3289BMI (kg/m2) < 23.80.633 (0.608–0.658)64.262.475.549.11.520.2663 < 20.00.555 (0.538–0.572)94.116.967.261.11.090.1094Age (years) ≥ 720.558 (0.532–0.583)54.856.841.269.41.140.1158 ≥ 650.504 (0.488–0.520)89.511.335.865.91.010.0073AUC area under the receiver operating characteristic curve, BMD bone mineral density, BMI body mass index, CI confidence interval, LR + positive likelihood ratio, NPV negative predictive value, PPV positive predictive value Performance of selected cutoff values for osteoporosis prediction and different BMD sites AUC area under the receiver operating characteristic curve, BMD bone mineral density, BMI body mass index, CI confidence interval, LR + positive likelihood ratio, NPV negative predictive value, PPV positive predictive value Table 6AUC difference comparison of cutoff values at any BMD siteVariablesAUCSEP-valueWeight < 57.4 kg (reference)0.6640.012BMI < 23.8 kg/m20.6330.0130.002Age ≥ 72 years0.5580.013 < 0.001BMI < 23.8 kg/m20.6330.013BMI < 20 kg/m20.4450.009 < 0.001Age ≥ 72 years0.5580.013Age ≥ 65 years0.5040.008 < 0.001AUC area under the receiver operating characteristic curve, BMI body mass index, SE standard error, BMD bone mineral density AUC difference comparison of cutoff values at any BMD site AUC area under the receiver operating characteristic curve, BMI body mass index, SE standard error, BMD bone mineral density Comparative ROC curves for age, weight, or BMI at the cutoff value model at any bone mineral density site for osteoporosis prediction. BMI, body mass index; ROC, receiver operating characteristic Discussion Simple clinical indicators can be potential predictors to guide physicians in making decisions regarding osteoporosis screening in women. These indicators are commonly used in osteoporosis risk assessment tools or as part of osteoporosis screening guidelines. The present study suggests an optimal cutoff value for body weight in predicting osteoporosis risk in Thai women. Additionally, this study re-evaluated the diagnostic performance of the cutoff BMI and age values for osteoporosis prediction. The present study demonstrated that weight is a more effective predictor of osteoporosis than BMI or age. Changing these cutoff values could improve the performance of osteoporosis prediction compared to previous cutoff values in CPGs [13]. However, cost-effectiveness studies are required to determine the efficacy of these screening strategies. As the prevalence of osteoporosis and osteopenia rises with age, this study has identified a higher prevalence of osteoporosis compared to previous research from other Asia–Pacific countries [7]. Specifically, the prevalence of osteoporosis among women in this study was 35.6%, which is notably higher than the 23.0% reported in a previous study conducted in Thailand. This discrepancy may be attributed to the older mean age of participants in our study relative to those in prior studies [7, 34]. This study demonstrated a negative correlation between BMD of the femoral neck or hip and age. The mean age of patients with osteoporosis at the femoral neck or hip was also significantly higher than that of patients without osteoporosis. In contrast to those with osteoporosis of the lumbar spine, a negative trend in the relationship between lumbar BMD and age was not observed. This may be due to degeneration of the lumbar spine in elderly women, which may alter lumbar BMD results and increase with age [35]. The diagnostic performances of weight and BMI for osteoporosis have been debated in several studies. Prior studies have reported the discriminative ability of weight over BMI and age for predicting osteoporosis [19, 36]. Similarly, this study showed that weight had significantly greater osteoporosis discriminative ability than BMI and age (AUC: 0.664 vs. 0.633 and 0.558). In addition, the cutoff weight of 57.4 kg identified in this study is similar to the 57 kg cutoff weight reported in a previous study, which demonstrated a high discriminative ability for femoral neck osteoporosis in women aged 40–59 years [19]. The present study demonstrated that weight and BMI may be more important than age in predicting osteoporosis in women aged ≥ 50 years. However, the sensitivity of the cutoff BMI of < 23.8 kg/m2 and the cutoff age of ≥ 72 years was lower than that of the cutoff BMI of < 20.0 kg/m2 and the cutoff age of ≥ 65 years (BMI: 64.2% vs. 94.1%; age: 54.8% vs. 89.5%). In the present study, the Youden index, which is a possible cutoff value, was optimized between the true and false positive rates. As a result, the specificity of the cutoff BMI of < 23.8 kg/m2 and the cutoff age of ≥ 72 years was higher than that of the cutoff BMI of < 20.0 kg/m2 and the cutoff age of ≥ 65 years (BMI: 62.4% vs. 16.9%; age: 56.8% vs. 11.3%). A previous study proposed that body weight < 58 kg could be a predictor of hip fractures in postmenopausal women aged ≥ 65 years [22]. The present study documented a sensitivity of 62.6% and a specificity of 70.3% at a weight threshold of 57.4 kg for predicting osteoporosis at any site T-score in women aged ≥ 50 years. Because the new cutoff value was derived from the highest Youden index rather than from a high prevalence fracture point value, and from a different age group and ethnicity, the new cutoff value of weight in this study was slightly lower than the former cutoff value (57.4 vs. 58 kg). According to the DXA screening indications of the Thailand osteoporosis CPG, age ≥ 65 years in women is the primary indication, and age < 65 years with a BMI of < 20 kg/m2 is the secondary indication in making DXA screening decisions. However, the present study demonstrated that weight and BMI exhibited greater diagnostic performance than age when used alone for the prediction of osteoporosis. The new cutoff values of weight or BMI may be more appropriate as primary indications than age for the DXA screening criteria in the CPG. This study has several limitations. First, the cutoff values were evaluated using a single statistical technique. Many diagnostic indices are useful for identifying the optimal cutoff point in a diagnostic model [30]. Secondly, the objective of this study is to evaluate the performance of individual predictive factors for determining the need for osteoporosis screening. However, it does not address fracture risk or treatment decision-making based on these clinical factors. Further studies evaluating the performance of FRAX in combination with these factors will be necessary to develop a more predictive model for osteoporosis treatment. Thirdly, while the cutoff points for each clinical predictor are determined by the highest AUC, none achieve 100% positive or negative predictive values. Although the AUC for individual clinical factors may be lower than that of combined factors, using individual factors may be simpler and more practical for healthcare providers. Finally, this study evaluated only a single population in a hospital setting. Before generalizing and implementing these new cutoff points, they should be validated in different population contexts and settings. Moreover, the absence of data on menopausal status that may be related to osteoporosis was another limitation of this study. Conclusions The present study demonstrated that simple clinical indicators, including BMI and age, can guide decision-making regarding DXA screening criteria for osteoporosis. This study suggests an optimal cutoff value for body weight in predicting osteoporosis risk in Thai women. All newly suggested cutoff values provided better diagnostic performance for osteoporosis prediction than the former cutoff values. The weight cutoff value was found to be more appropriate than the BMI and age cutoff values for osteoporosis prediction. Acknowledgements Not applicable. Abbreviations Body mass index Dual-energy x-ray absorptiometry Area under the receiver operating characteristic curve Clinical practice guidelines Bone mineral density Standard deviations Confidence interval Author contributions J.L. performed the investigation. J.L. performed data analysis and wrote the document. K.R. performed the investigation and reviewed the document. P.I. reviewed the document. Y.P. reviewed the document. All authors read and approved the final manuscript. Funding The author(s) declare that no financial support was received for the research, authorship, and/or publication of this article. Data availability No datasets were generated or analysed during the current study. Declarations Ethics approval and consent to participate This study was approved by the institutional review board (No. S025h/66_ExPD) and was conducted following the Declaration of Helsinki guidelines. The committee waived the requirement of informed consent because of the retrospective nature of the study. Consent for publication I certify that this manuscript is a unique submission and is not being considered for publication, in part or in full, with any other source in any medium. Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement No datasets were generated or analysed during the current study.