Development and Validation of a Nomogram for Predicting Drug-induced thrombocytopenia in Hospitalized Patients: A Case‒Control Study Based on Propensity-Score Matching

doi:10.21203/rs.3.rs-2954229/v1

Download PDF

Research Article

Development and Validation of a Nomogram for Predicting Drug-induced thrombocytopenia in Hospitalized Patients: A Case‒Control Study Based on Propensity-Score Matching

https://doi.org/10.21203/rs.3.rs-2954229/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Objective

Drug-induced thrombocytopenia (DITP) is associated with increased mortality. This study aims to establish a nomogram to predict the occurrence of DITP in hospitalized patients in a multidrug environment.

Methods

A single-centre retrospective study among hospitalized adult patients was conducted from January 2021 to December 2021 and was based on the Adverse Drug Events Active Surveillance and Assessment System-Ⅱ (ADE-ASAS-Ⅱ). Three controls were matched for each case according to the propensity score matching algorithm to eliminate confounding bias due to individual baseline variables. Predictors of DITP were obtained by LASSO regression and were used to build the nomogram.

Results

Among 88151 hospitalized patients, 478 were confirmed to have DITP, which is an incidence rate of 0.54%. After matching, 382 pairs and 1146 patients were included in the study, including 1070 cases in the development group and 427 cases in the validation group. Five variables were used to construct the nomogram: hospital stay ≥ 14days, surgery before using suspected drugs, baseline platelet count < 150×10⁹/L, higher baseline BUN and combined with antibacterial. The areas under the curve (AUC) in the development group and validation group were 0.827 (95% CI 0.800-0.854) and 0.785 (95% CI 0.736–0.834), respectively, and the model also showed good calibration (P > 0.05) in the development group and validation group.

Conclusion

The established nomogram can help identify high-risk patients with DITP, assist doctors in decision-making, and effectively prevent DITP in the early stage.

Adverse drug reactions

drug-induced thrombocytopenia

nomogram

predictive models

active surveillance

Thrombocytopenia (TP) is a disease characterized by decreased platelet counts in the peripheral blood that can be caused by both congenital and acquired causes[1]. Drug-induced thrombocytopenia (DITP) is one of the most common types of TP. DITP is an adverse drug reaction (ADR) and a significantly underrecognized clinical problem. In the event of DITP, the platelet count is usually below 100×10⁹/L; therefore, DITP can often lead to abrupt and severe bleeding complications and even death[2, 3]. DITP urgently requires further attention since its cumulative incidence is 10 cases per million population per year, with a 25% prevalence in critically ill patients[4]. Many drugs can cause TP by inhibiting platelet production and/or promoting the elimination or destruction of platelets from the peripheral blood[2]. More than 300 drugs, including antibiotic[5–8], neurogenic[9] and antineoplastic drugs[10, 11], are reported to cause DITP and are associated an increased risk of microtraumatic bleeding and spontaneous bleeding, thus becoming a health hazard.

Although a reduced platelet count is the basis for the diagnosis of DITP, early identification is difficult. Several studies have been conducted to develop methods for detecting DITP signals using electronic health records[12, 13]. Other studies have been conducted to identify the characteristics and risk factors for DITP [14, 15]. However, these studies cannot provide a model to accurately predict the risk of DITP. Therefore, it is particularly important to develop a clinically convenient, accurate, and efficient DITP risk prediction model. Currently, nomograms are used to predict ADRs because of their intuitiveness and easy-to-understand character. Nomograms are rarely used to predict the risk of DITP in a multidrug environment and in patients of the general hospital population.

Therefore, based on the Adverse Drug Events Active Surveillance and Assessment System-Ⅱ (ADE-ASAS-Ⅱ), we conducted a single-centre, case‒control study with propensity score matching (PSM) to explore the clinical characteristics of patients with DITP and developed a nomogram to predict the risk of DITP in hospitalized patients.

2.1 Study Design and Patient Selection

This study is a single-centre, retrospective, PSM case‒control study. Eligible patients included those who developed DITP after treatment with drugs in our hospital from January 1, 2021, to December 31, 2021.

All patient data, including medical records and examination information, were obtained from the Hospital Information System (HIS). The ADE-ASAS-Ⅱ was based on trigger technology and text recognition technology, which can be connected to the HIS to extract patient information. Detailed descriptions of the ADE-ASAS-Ⅱ and its applications have been described in our previous study[16–18]. The monitoring plan of this study was conducted in the full-prescription mode, and the investigator preset the alarm index of the trigger without the monitoring drug, so once the monitoring index triggered the alarm index, the system could actively monitor all drugs that had a temporal relationship with TP as an early warning signal of DITP. Then, two clinical pharmacists blindly evaluated the alarm cases to confirm the results, and the cases with inconsistent evaluation results were submitted to the experts for final judgement to determine whether DITP occurred.

The inclusion criteria were as follows: 1) two consecutive platelet (PLT) counts that were < 100×109/L-1; 2) age ≥ 18 years old; and 3) a full prescription model (temporary and long-term prescriptions).

The exclusion criteria were as follows: 1) baseline PLT < 100×10⁹/L-1; 2) patients with incomplete laboratory indexes before or after medication administration; 3) patients with incomplete clinical records; 4) patients of the haematology department; 5) involved drugs used in transarterial chemoembolization; and 5) a diagnosis that may cause TP, including primary immune thrombocytopenia, pseudothrombocytopenia purpura, haemophagocytic syndrome, thrombotic microvascular disease, disseminated intravascular coagulation, haemolUS syndrome, systemic lupus erythematosus, antiphospholipid syndrome, hyperplenism, myelodysplastic syndrome, leukaemia, aplastic anaemia, Gaucher disease (Gaucher disease), Evans syndrome, Fanconi anaemia, platelet angiohemophilia, hereditary thrombocytopenia (giant platelet syndrome, grey platelet syndrome, Berna And rd-Soulier syndrome, thrombocytopenia with radial deletion syndrome, Wiskott-Aldrich syndrome, MYH 9-associated disease), rheumatoid arthritis, lymphoma, HELLP syndrome, and cirrhosis.

The Naranjo Adverse Drug Reaction Probability Scale (Naranjo Scale) was used as a causality assessment tool[19].The diagnosis of an ADR was assigned to a probability category from the total score as follows: definite ≥ 9, probable 5 to 8, possible 1 to 4, and doubtful ≤ 0. Patients with scores ≥ 1 were defined as having DITP. Controls were defined as patients treated with similar drugs who did not develop DITP. To minimize confounding bias due to demographic characteristics, we performed 1:3 propensity core matching between the DITP and non-DITP groups to make the groups more comparable.

2.2 Data Collection and Definitions

Patient data were extracted from the HIS using the ADE-ASAS-Ⅱ. The patient characteristics included age, sex, weight, smoking history, drinking history, drug and food allergy history, history of autoimmune diseases, length of hospital days, suspected drug duration, and surgery before using suspected drugs. Concomitant drugs included antiplatelet drugs, anticoagulant drugs, antibiotics, antineoplastic drugs, immunosuppressant, and nonsteroidal anti-inflammatory drugs (NASIDs). Laboratory data included haemoglobin (HB), red blood cell (RBC) count, white blood cell (WBC) count, neutrophil (NEU) count, eosinophil (EOS) count, basophil (BASO) count, lymphocyte (LYM) count, mean corpuscular volume (MCV), mean corpuscular haemoglobin (MCH), mean corpuscular haemoglobin (MCHC), red blood cell distribution width (RDW), PLT, mean platelet volume (MPV), direct bilirubin (DBIL), total bilirubin (TBIL), alanine transaminase (ALT), aspartate aminotransferase (AST), gamma glutamyl transferase (GGT), AST/ALT ratio, blood urea nitrogen (BUN), and estimated glomerular filtration rate (eGFR). Secondary data, such as eGFR, were calculated. We used the epidemiological collaborative creatinine equation (CKD-EPI) to calculate eGFR because it is more accuratethan the dietary modification (MDRD) formula recommended by clinical practice guidelines[20, 21]. Baseline PLT was defined as the last laboratory measurement between 7 days before and 2 hours after suspected drug administration. The other laboratory values were the most recent laboratory findings before the suspected drug was administered.

2.3 Statistical methods:

All statistical analyses and model development were performed with the statistical software packages R 4.0.2 (http://www.R-project.org, The R Foundation), STATA software (version 16.0) and Free Statistics software versions 1.7. For baseline characteristics, qualitative data were expressed by quantity and percentage using the χ2 test or Fisher's exact probability test. Quantitative data were presented as the mean ± standard deviation (standard deviation, SD) using the t test; those not meeting the normal distribution are presented as median and interquartile spacing (inter quartile range, IQR) using the Mann‒Whitney U test. All P values were flanked, and P < 0.05 was statistically significant.

To achieve similar baseline characteristics between the two groups, the patients with DITP and DITP and without DITP were matched using PSM. A 1:3 scale matching (calliper value of 0.05) was performed using the "psmatch2 package" of the STATA software. The propensity score was assessed according to age, sex, and BMI.

The cases and controls were matched with percentages of 70% and 30% to the development and validation groups, respectively. To identify the variables to build the DITP prediction model, we used the LASSO cross-validation algorithm to screen the variables [22]. We used logistic regression analysis for the development group. After the univariate analysis, the variables with significant differences (P < 0.05) were included in the LASSO regression algorithm to screen for predictors of DITP, the parameters with a parameter regression coefficient of 0 were deleted for dimension reduction, 5-fold cross validation was conducted, and two dotted lines were drawn at the selected values. Finally, the variables corresponding to the optimal λ were used to draw the nomogram.

The model was validated in terms of discrimination and calibration. Model discrimination was evaluated using the receiver operating characteristic (ROC) curve and the area under the curve (AUC). The calibration degree of the model was evaluated using a calibration curve.

3.1 Clinical characteristics of the patients

A total of 88151 patients were monitored by the ADE-ASAS-II from January 1, 2021 to December 31, 2021, and 1835 (2.45%) indicated early warning signals. Among the patients with early warning signals, 1229 patients were excluded according to the exclusion criteria. After independent evaluation by two clinical pharmacists using the Naranjo Scale, 128 patients with no causal relationship to the drug were excluded. A total of 478 patients with a Naranjo score ≥ 1 were identified as the DITP group. In this study, 478 of the 88151 inpatients (age ≥ 18 years old) developed DITP, representing an overall incidence of 0.54%. Among the hospitalized patients during the current study period, 32,173 patients used the same suspected drug but did not develop DITP. By PSM, the final matching included 382 DITP and 1146 without DITP (N-DITP) patients, and the patient exclusion and inclusion process is detailed in Fig. 1. After PSM, there were no significant differences between age, sex and BMI (P > 0.05), as shown in Table 1. In this study, 1528 patients were finally included in the analysis and randomized to a 7:3 development group (n = 1070) and a validation group (n = 458). The baseline clinical characteristics of the development and validation groups are shown in Table 2.

Table 1

Baseline characteristics of patients before and after propensity score matched analysis
Variables	Before matched(n = 32173)			After matched
Variables	N- DITP (n = 32173)	DITP (n = 478)	P	N-DITP (n = 1146)	DITP (n = 382)	P
Age,years(Mean ± SD)	56.4 ± 16.3	62.1 ± 15.1	< 0.001	62.6 ± 15.1	64.8 ± 14.1	0.013
Males,n(%)	16738(52)	320(66.9)	< 0.001	765(66.8)	239(62.6)	0.135
BMI,kg/m²(Mean ± SD)	24.3 ± 3.8	23.3 ± 3.5	< 0.001	23.6 ± 3.9	23.3 ± 3.6	0.232

2.2 Predictors of drug-induced thrombocytopenia

Of the 1070 patients in the development group, 280 had DITP and 790 did not. After logistic univariate analysis of the development group, the variables with statistically significant differences (P < 0.05) included baseline WBC, RBC, MONO, RDW, MPV, ALB, TBIL, DBIL, AST/ALT, AST ≥ 40U/L, GGT ≥ 50U/L, eGFR ≥ 60ml/min/1.73m², length of hospital stay ≥ 14 d, history of autoimmune disease, suspected drug duration ≥ 5d, combination with antiplatelet drugs, combination with anticoagulant drugs, combination with antibiotics, and surgery before suspected drug, as shown in Table 3.

Table 2

Baseline characteristics in the development and validation groups
Variables	Development group(n = 1070)			Validation group(n = 458)
Variables	N- DITP (n = 790)	DITP (n = 280)	P	N-DITP (n = 356)	DITP (n = 102)	P
Demographicdata
age,years,(Mean ± SD)	62.6 ± 15.0	63.6 ± 13.4	0.317	62.6 ± 15.3	64.5 ± 14.2	0.254
Males,n(%)	516(65.3)	181(64.6)	0.839	249(69.9)	58(57.4)	0.018
BMI.kg/m²,Mean ± SD	23.6 ± 3.8	23.4 ± 3.8	0.391	23.5 ± 4.0	23.0 ± 3.2	0.216
Median hospital stay,days(IQR)	11.0(6.0,19.0)	19.0(11.0,28.2)	< 0.001	10.0(6.0,19.0)	17.0(10.0,32.0)	< 0.001
Median suspected drug duration,days(IQR)	1.0(1.0,5.0)	2.0(1.0,7.0)	< 0.001	1.0(1.0,4.0)	3.0(1.0,8.0)	< 0.001
Autoimmune.disease.history,n(%)	18(2.3)	14(5)	0.022	4(1.1)	1(1)	1
Drug.allergy.history,n(%)	84(10.6)	39(13.9)	0.137	29(8.1)	15(14.9)	0.044
Alcohol.history,n(%)	290(36.7)	96(34.3)	0.468	131(36.8)	35(34.7)	0.692
surgery,n(%)	251(31.8)	66(23.6)	0.01	122(34.3)	23(22.8)	0.028
Concomitantuseofdrugs
Antiplatelet drugs,n(%)	45(5.7)	29(10.4)	0.008	21(5.9)	11(10.9)	0.083
GPⅡb.Ⅲa inhibitors,n(%)	4(0.5)	1(0.4)	1	3(0.8)	1(1)	1
NSAIDs,n(%)	175(22.2)	70(25)	0.33	76(21.3)	26(25.7)	0.349
Antineoplastic drugs,n(%)	353(44.7)	109(38.9)	0.095	145(40.7)	40(39.6)	0.839
Antibacterial drugs,n(%)	265(33.5)	131(46.8)	< 0.001	123(34.6)	49(48.5)	0.011
Antiviral drugs,n(%)	14(1.8)	9(3.2)	0.153	5(1.4)	3(3)	0.383
Immunosuppressive drugs,n(%)	12(1.5)	8(2.9)	0.155	3(0.8)	3(3)	0.125
Anticoagulants,n(%)	98(12.4)	48(17.1)	0.047	41(11.5)	19(18.8)	0.055
Laboratorymeasurements
HB,g/L(Mean ± SD)	118.1 ± 21.3	117.5 ± 21.9	0.688	118.7 ± 22.1	114.3 ± 21.8	0.078
RBC count,×10¹²/L(Mean ± SD)	3.9 ± 0.7	3.8 ± 0.8	0.004	3.9 ± 0.7	3.7 ± 0.7	0.007
Median WBC count,×10⁹/L(IQR)	6.2(4.8,8.0)	5.6(4.3,7.9)	0.004	6.4(5.0,8.6)	5.7(3.8,7.4)	< 0.001
MCV,fl,(Mean ± SD)	89.4 ± 6.1	91.4 ± 6.2	< 0.001	89.3 ± 6.2	90.6 ± 5.9	0.065
RDW,%,(Mean ± SD)	14.3 ± 2.5	14.2 ± 2.5	0.035	14.1 ± 2.2	14.5 ± 2.3	0.167
MCH,pg,(Mean ± SD)	30.3 ± 2.4	31.4 ± 2.4	< 0.001	30.4 ± 2.6	30.9 ± 2.3	0.079
MCHC,g/L,(Mean ± SD)	339.2 ± 12.1	343.5 ± 13.5	< 0.001	340.1 ± 12.0	341.0 ± 12.6	0.475
MPV,fL,(Mean ± SD)	10.3 ± 1.1	10.6 ± 1.1	< 0.001	10.2 ± 1.1	10.6 ± 1.2	< 0.001
Median LYM count,×10⁹/L(IQR)	1.4(1.0,1.8)	1.2(0.8,1.6)	< 0.001	1.4(0.9,1.8)	1.1(0.7,1.5)	< 0.001
Median ESO count,×10⁹/L(IQR)	0.1(0.0,0.2)	0.1(0.0,0.2)	< 0.001	0.1(0.0,0.2)	0.1(0.0,0.2)	0.048
Median MONO count,×10⁹/L(IQR)	0.5(0.4,0.6)	0.4(0.3,0.6)	< 0.001	0.5(0.4,0.6)	0.4(0.3,0.6)	< 0.001
Median neu count,×10⁹/L(IQR)	3.9(2.8,5.8)	3.7(2.5,5.6)	0.136	4.2(3.0,6.3)	3.5(2.2,5.6)	0.005
Median PLT count,×10⁹/L(IQR)	212.5(168.0,258.0)	133.0(112.0,172.2)	< 0.001	211.0(164.0,260.2)	135.0(111.0,187.0)	< 0.001
Median eGFR,ml/min/1.73m²(IQR)	111.7(90.8,139.5)	110.7(82.4,139.8)	0.104	107.4(88.2,140.5)	111.8(89.1,142.2)	0.562
Median BUN,mmol/L(IQR)	5.1(4.1,6.8)	5.6(4.3,8.0)	< 0.001	5.2(4.1,6.6)	5.6(4.2,6.7)	0.32
AST/ALT	1.1(0.8,1.5)	1.2(0.9,1.6)	0.013	1.1(0.8,1.5)	1.1(0.9,1.5)	0.959
ALT ≥ 40,n(%)	92(11.6)	39(13.9)	0.317	40(11.2)	14(13.9)	0.471
AST ≥ 40,n(%)	73(9.2)	41(14.6)	0.012	37(10.4)	15(14.9)	0.213
GGT ≥ 50,n(%)	205(25.9)	105(37.5)	< 0.001	85(23.9)	55(54.5)	< 0.001
eGFR>60,n(%)	731(92.5)	237(84.6)	< 0.001	326(91.6)	93(92.1)	0.871
Median TBIL,µmol/L,(IQR)	8.1(5.5,12.3)	10.4(7.1,15.3)	< 0.001	8.4(6.0,12.2)	11.4(7.1,14.9)	0.003
Median DBIL,µmol/L,(IQR)	3.0(2.1,5.0)	4.8(3.1,7.6)	< 0.001	3.2(2.1,5.2)	4.7(2.9,7.3)	< 0.001
Median BASO count,×10⁹/L(IQR)	0.0(0.0,0.0)	0.0(0.0,0.0)	0.005	0.0(0.0,0.0)	0.0(0.0,0.0)	0.009

Table 3

Results of the univariate logistic regression analysis of the patients in the development group
	Variable	OR_95CI	P
1	Antiplatelet drugs	1.91(1.17 ~ 3.12)	0.009
2	Antibacterial agent	1.74(1.32 ~ 2.3)	< 0.001
3	Anticoagulants	1.46(1 ~ 2.13)	0.048
4	Autoimmune disease history	2.26(1.11 ~ 4.6)	0.025
5	surgery	0.66(0.48 ~ 0.91)	0.01
6	RBC	0.75(0.62 ~ 0.91)	0.004
7	WBC	0.95(0.91 ~ 1)	0.037
8	RDW	1.06(1 ~ 1.11)	0.036
9	MPV	1.42(1.25 ~ 1.62)	< 0.001
10	ALB	0.97(0.95 ~ 1)	0.026
11	TBIL	1.01(1 ~ 1.01)	0.024
12	DBIL	1.01(1 ~ 1.02)	0.011
13	BUN	1.07 (1.04 ~ 1.1)	< 0.001
14	BASO	0(0 ~ 0.77)	0.042
15	LYM	0.53(0.42 ~ 0.69)	< 0.001
16	MONO	0.46(0.25 ~ 0.86)	0.015
17	PLT ≥ 150	0.11(0.08 ~ 0.16)	< 0.001
18	AST.ALT	1.21(1.02 ~ 1.43)	0.028
19	AST ≥ 40	1.68(1.12 ~ 2.54)	0.013
20	GGT ≥ 50	1.71(1.28 ~ 2.29)	< 0.001
21	eGFR ≥ 60	0.44(0.29 ~ 0.68)	< 0.001
22	hospital.days ≥ 14d	3.17(2.38 ~ 4.21)	< 0.001
23	Suspected drug duration ≥ 5d	1.65(1.23 ~ 2.21)	0.001

The variables with statistically significant differences (P < 0.05) in the logistic univariate analysis of the development group were included in the LASSO regression analysis, and the variable screening process is shown in Fig. 2. A coefficient distribution map was obtained according to the log (λ) sequence. Through 5-fold cross-validation, two dashed lines were plotted at the selected values, where the optimal λ was 0.026 (1 standard error of the lowest standard), which corresponded to five non zero variables, namely,hospital stay ≥ 14days, surgery before using suspected drugs, baseline platelet count < 150×10⁹/L, higher baseline BUN and combined with antibiotics.

2.3 Development and validation of the nomogram

A nomogram was constructed to predict DITP based on the independent predictors from the LASSO regression analysis (see Fig. 3). In this nomogram, baseline BUN were continuous variables; the other predictors were categorical variables, including whether the PLT was above or below 150×10⁹/L, whether or not the patient received surgery before medication, whether or not the patient combined with antibiotics and whether the hospital stay was more or less than 14 days. The nomogram has 8 rows; the first line is the point assignment of the predictor, and lines 2 to 6 are the predictors of DITP. The probability of DITP is the probability predicted by matching the total score to the score on the total score table (line 7). To assess the discrimination of the model and reduce overfitting bias, we performed internal validation of the validation group. Figure 4 shows the receiver operating characteristic (ROC) curves during development and validation. In our study, the area under the curve (AUC) of the development group was 0.827 (95% CI 0.800-0.854) and 0.785 (95% CI 0.736–0.834), respectively. The calibration degree was verified, and the nomogram model also showed good calibration, with P values of 0.687 and 0.285 for the development and validation groups, respectively (Fig. 5).

This study found that the overall incidence of DITP was approximately 0.54% in our adult inpatient population. A cross-sectional study from the Netherlands showed an inpatient incidence of DITP of approximately 5% [23]. One possible reason for the lower results of this study than those reported in the literature is that those previous studies only required one PLT < 100×10⁹/L result for a diagnosis of DITP. However, the patients with DITP in the study must have had two consecutive PLT < 100×10⁹/L tests, and the use of the drug for hepatic artery chemoembolization must have been excluded, which also reduces the incidence. In this study, PLT < 100×10⁹/L was used as a cut-off value because this value was closely associated with an increased ICU residence time and increased mortality[9, 24]. Therefore, this study showed that the overall incidence of DITP was lower than that reported in the literature.

In the current study on DITP, a test model based on a Chinese patient population to accurately assess the risk of developing DITP in a multi-drug environment had not been developed[25, 26] .Therefore, a simple and easy-to-use prediction of DITP for adult inpatients was developed to help clinicians more accurately identify patients at potential risk of DITP.

The DITP risk prediction in this study was developed in the development group and validated in the validation group. Previous research experience shows that an AUC value greater than 0.7 indicates a good predictive performance of the model[27]. In our model, the area under the curve (AUC) was 0.724 (95% CI 0.690–0.758) in the development group and 0.733 (95% CI 0.678–0.787) in the validation group. A calibration degree validation was also performed, and the nomogram model also showed good calibration, with P values of 0.687 and 0.285 for the development and validation groups, respectively, which are P values that are > 0.05, which shows good calibration. In clinical practice, clinicians can use the predictive model to assess the risk of developing DITP, as needed, and initially classify newly admitted patients according to this risk. Once high-risk patients are identified, PLT should be closely monitored.

The nomogram model for this study was based on five predictors, including hospital stay ≥ 14days, surgery before using suspected drugs, baseline platelet count < 150×109/L, higher baseline BUN and combined with antibiotics. Previous literature suggested that the baseline PLT counts were significantly lower in patients with TP than in patients without TP[28]. Another analysis of DITP risk factors in elderly patients found that a baseline platelet count < 150×10⁹/L was significantly associated with the occurrence of TP in elderly patients[29]. The results of this study are consistent with those reported in the literature.

This study is the first to show that a hospital stay of 14 days was associated with an increased risk of DITP. Although there was no similar report in previous studies, previous studies on risk factors for DITP caused by the administration of linezolid showed that a long duration of medication increased the risk of developing DITP [29]. Because this study included patients in a multidrug setting, the incubation time of DITP varies between drug varieties, but patients hospitalized for 14 days will have an increased DITP leakage rate.

Previous studies have shown BUN > 6.4 mmol/L as a risk factor for increased mortality in patients with severe fever with thrombocytopenia[30]. There are still no studies on the association between DITP and BUN, and this study found that increased BUN was a predictor of the increased risk of DITP

This study is a retrospective single-centre study, and therefore, potential bias is inevitable. These shortcomings will be addressed in our future studies.

In conclusion, by combining information means and manual evaluation, we efficiently obtained study data and developed and validated a risk prediction model for thrombocytopenia in hospitalized patients under whole-drug conditions. After validation, the model has good differentiation and calibration, and it can predict the risk of granulocyte deficiency and effectively prevent and control adverse drug reactions.

Statements and Declarations

Funding This work was supported by the Fund by “the Project of Monitoring and Evaluation of the Use of Key Clinical Drugs” commissioned by the Chinese Association of Research Hospitals (grant number Y2022FH-YWPJ01-101).

Conflict of interest:All authors certify that they have no affiliations with or involvement in any organization or entity with any financial interest or non-financial interest in the subject matter or materials discussed in this manuscript.

Ethical approval This study was approved by the Ethics Committee of Chinese PLA General Hospital (S2022-369-01). All patient data were kept strictly confidential.

Consent for publications All authors agree to publish in the journal “drug safety”.

Author contributions ZT and MZ conception and design.ZT data collection, analysis of data, drafting, and revision of manuscript. MZ interpretation of results and critical revision of manuscript.JL,KG,RZ,JN,FZ,YL ,TXADR ev,lutation.All authors read and approved the final version.

Kenney B, Stack G. Drug-induced thrombocytopenia. Arch Pathol Lab Med. 2009;133:309–14.
Vayne C, Guéry E-A, Rollin J, Baglo T, Petermann R, Gruel Y. Pathophysiology and Diagnosis of Drug-Induced Immune Thrombocytopenia. J Clin Med. 2020;9:2212.
Doodnauth AV, Sivakumar S, Mulatu Y, Alicea E, McFarlane SI. Acute Severe Thrombocytopenia Event Associated with Trimethoprim/Sulfamethoxazole Use. Am J Med Case Rep. 2021;9:155–7.
Danese E, Montagnana M, Favaloro EJ, Lippi G. Drug-Induced Thrombocytopenia: Mechanisms and Laboratory Diagnostics. Semin Thromb Hemost. 2020;46:264–74.
Butt MU, Jabri A, Elayi SC. Azithromycin-Induced Thrombocytopenia: A Rare Etiology of Drug-Induced Immune Thrombocytopenia. Case Rep Med. 2019;2019:6109831.
Mohammadi M, Jahangard-Rafsanjani Z, Sarayani A, Hadjibabaei M, Taghizadeh-Ghehi M. Vancomycin-Induced Thrombocytopenia: A Narrative Review. Drug Saf. 2017;40:49–59.
Savage-Elliott I, Wu VJ, Sanchez FL. Drug-Induced Thrombocytopenia Secondary to Commonly Used Antibiotics in Total Joint Arthroplasty. Arthroplasty Today. 2020;6:137–40.
Cachia J, Torpiano P, Pace D. Meropenem-induced thrombocytopenia: a paediatric case. BMJ Case Rep. 2021;14.
Kim DW, Kim W, Lee C-H, Chun YI. Thrombocytopenia during intravenous valproic acid therapy in the neurological intensive care unit. J Clin Pharm Ther. 2020;45:1014–20.
Stack A, Khanal R, Denlinger CS. Oxaliplatin-induced Immune Thrombocytopenia: A Case Report and Literature Review. Clin Colorectal Cancer. 2021;20:e1–4.
Tam EL, Draksharam PL, Park JA, Sidhu GS. Acute Immune-Mediated Thrombocytopenia due to Oxaliplatin and Irinotecan Therapy. Case Rep Oncol Med. 2019;2019:4314797.
Moore C, Li J, Hung C-C, Downs J, Nebeker JR. Predictive value of alert triggers for identification of developing adverse drug events. J Patient Saf. 2009;5:223–8.
Lee YK, Yoon HS, Lee EH, Cho SY. Can we Predict the Clinical Course of Immune Thrombocytopenia in Children by The Mean Platelet Volume? A Preliminary Study. Clin Lab. 2021;67.
Ong SY, Tan CW, Ramya V, Malik AA, Lee XH, Hwang JCC, et al. Risk factors and predictors of treatment responses and complications in immune thrombocytopenia. Ann Hematol. 2021;100:645–51.
Matsumoto K, Shigemi A, Ikawa K, Kanazawa N, Fujisaki Y, Morikawa N, et al. Risk Factors for Ganciclovir-Induced Thrombocytopenia and Leukopenia. Biol Pharm Bull. 2015;38:235–8.
Chen C, Jia W, Guo D, Zhu M, Xu Y, Wang X, et al. Development of a Computer-Assisted Adverse Drug Events Alarm and Assessment System for Hospital Inpatients in China. Ther Innov Regul Sci. 2020;54:32–41.
Yu C, Guo D, Yao C, Zhu Y, Liu S, Kong X. Development and Validation of a Nomogram for Predicting Drug-Induced Acute Kidney Injury in Hospitalized Patients: A Case-Control Study Based on Propensity-Score Matching. Front Pharmacol. 2021;12:657853.
Yu C, Guo D, Yao C, Yang H, Liu S, Zhu Y, et al. Clinical Characteristics of Hospitalized Patients with Drug-Induced Acute Kidney Injury and Associated Risk Factors: A Case-Control Study. BioMed Res Int. 2020;2020:1–8.
Naranjo CA, Busto U, Sellers EM, Sandor P, Ruiz I, Roberts EA, et al. A method for estimating the probability of adverse drug reactions. Clin Pharmacol Ther. 1981;30:239–45.
Kellum JA, Lameire N. Diagnosis, evaluation, and management of acute kidney injury: a KDIGO summary (Part 1). Crit Care Lond Engl. 2013;17:204.
Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF 3rd, Feldman HI, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med. 2009;150:604–12.
Wu W, Deng Z, Alafate W, Wang Y, Xiang J, Zhu L, et al. Preoperative Prediction Nomogram Based on Integrated Profiling for Glioblastoma Multiforme in Glioma Patients. Front Oncol. 2020;10:1750.
ten Berg MJ, van Solinge WW, van den Bemt PMLA, Huisman A, Schobben AFAM, Egberts TCG. Platelet measurements versus discharge diagnoses for identification of patients with potential drug-induced thrombocytopenia: a cross-sectional study in the Netherlands. Drug Saf. 2009;32:69–76.
Vandijck DM, Blot SI, De Waele JJ, Hoste EA, Vandewoude KH, Decruyenaere JM. Thrombocytopenia and outcome in critically ill patients with bloodstream infection. Heart Lung J Crit Care. 2010;39:21–6.
Song B, Zhou S, Li C, Zheng H, Zhang X, Jin X, et al. A Prediction Model for Chemotherapy-Induced Thrombocytopenia Based on Real-World Data and a Close Relationship Between AST/ALT Ratio and Platelet Count in Patients with Solid Tumors. Int J Gen Med. 2022;15:8003–15.
Inoue Y, Takekuma Y, Miyai T, Kashiwagi H, Sato Y, Sugawara M, et al. Use of Japanese big data from electronic medical records to investigate risk factors and identify their high-risk combinations for linezolid-induced thrombocytopenia. Eur J Clin Pharmacol. 2023;79:415–25.
Wang Q, Tang Y, Zhou J, Qin W. A prospective study of acute kidney injury in the intensive care unit: development and validation of a risk prediction model. J Transl Med. 2019;17:359.
Choi GW, Lee J-Y, Chang MJ, Kim YK, Cho Y, Yu YM, et al. Risk factors for linezolid-induced thrombocytopenia in patients without haemato-oncologic diseases. Basic Clin Pharmacol Toxicol. 2019;124:228–34.
Qin Y, Chen Z, Gao S, Shen Y, Ye Y. Development and validation of a risk prediction model for linezolid-induced thrombocytopenia in elderly patients. Eur J Hosp Pharm Sci Pract. 2022;ejhpharm-2022-003258.
Gong L, Zhang L, Wu J, Lu S, Lyu Y, Zhu M, et al. Clinical Progress and Risk Factors for Death from Severe Fever with Thrombocytopenia Syndrome: A Multihospital Retrospective Investigation in Anhui, China. Am J Trop Med Hyg. 2021;104:1425–31.

No competing interests reported.

Download PDF

Editorial decision: Revision requested
21 May, 2024
Reviewers agreed at journal
06 Mar, 2024
Reviewers agreed at journal
20 Nov, 2023
Reviewers invited by journal
23 Jun, 2023
Editor assigned by journal
22 May, 2023
Submission checks completed at journal
22 May, 2023
First submitted to journal
18 May, 2023

You are reading this latest preprint version

Development and Validation of a Nomogram for Predicting Drug-induced thrombocytopenia in Hospitalized Patients: A Case‒Control Study Based on Propensity-Score Matching

Status:

Version 1

Abstract

Objective

Methods

Results

Conclusion

Figures

1. Introduction

2. Methods

2.1 Study Design and Patient Selection

2.2 Data Collection and Definitions

2.3 Statistical methods:

3. Results

3.1 Clinical characteristics of the patients

2.2 Predictors of drug-induced thrombocytopenia

2.3 Development and validation of the nomogram

4. Discussion

5. Limitations and deficiencies

6. Conclusion

Declarations

References

Additional Declarations

Status:

Version 1