Insulin Dependence Increases the Risk of Complications and Death in Total Joint Arthroplasty: A Systematic Review and Meta‐(Regression) Analysis

Objectives To investigate the proportion of insulin‐dependent diabetes mellitus (IDDM) patients among diabetic patients undergoing total joint arthroplasty (TJA) and whether insulin dependence is associated with postoperative complications. Methods A systematic literature search was performed in EMBASE, PubMed, Ovid, Medline, the Cochrane Library, Web of Science, the China Science and Technology Journal Database, and China National Knowledge Infrastructure from the inception dates to 10 September 2019. Observational studies reporting adverse events with IDDM following TJA were included. Primary outcomes were cardiovascular complications, pulmonary complications, kidney complications, wound complications, infection, and other complications within 30 days of surgery. Secondary outcomes were the proportion of IDDM patients among diabetic patients undergoing TJA and its time trend. Results A total of 19 studies involving 85,689 participants were included. Among patients undergoing TJA, 26% of diabetic patients had IDDM. Compared with non‐insulin‐dependent diabetes (NIDDM), the incidences of cardiac arrest (risk ratio [RR], 2.346; 95% confidence interval [CI], 1.553 to 3.546), renal failure (relative risk [RR], 2.758; 95% CI, 1.830 to 4.156), deep incisional surgical site infection (RR, 1.968; 95% CI, 1.107 to 3.533), wound dehiscence (RR, 2.209; 95% CI, 1.830 to 4.156), and death (RR, 2.292; 95% CI, 1.568 to 3.349) were all significantly increased in IDDM. A significant time trend was witnessed for the prevalence of IDDM (P = 0.014). There was no statistical significance for organ/space surgical site infection, thrombotic events (deep venous thrombosis/ pulmonary embolism), and revision rates. Conclusion Insulin‐dependent diabetes is an independent high‐risk factor for increased adverse outcomes relative to NIDDM, suggesting that hierarchical and optimal blood glucose management may contribute to reducing the adverse complications after surgery for these patients. In addition, because the risk of sepsis, deep wound infection, organ/space surgical site infection, urinary tract infection, renal insufficiency, and renal failure significantly increase after TJA in IDDM patients, more active postoperative antimicrobial prophylaxis may be needed on the premise of protecting renal function.


Introduction
T otal joint replacement (TJA) is a common and effective method to relieve the pain from osteoarthritis (OA). Globally, the rate of TJA is projected to increase accordingly with the rising prevalence of OA [1][2][3] . Meanwhile, diabetes mellitus (DM) is a pressing public health issue, and its prevalence is expected to increase by 69% and 20%, respectively, in developing and developed countries between 2010 and 2030 4 . There is evidence from clinical and animal models to suggest an underlyingly independent link between DM and severity of OA 5 . Therefore, the number of TJA procedures is likely to increase in diabetic patients. Furthermore, an association between DM and adverse events, hospital readmission, and increased death rates has been witnessed in lower extremity arthroplasty and shoulder arthroplasty 6, 7 . However, these aforementioned studies generally classify patients as DM and non-DM patients, without further considering whether they are relying on insulin to achieve glycemic control. Studies have found that patients dependent on insulin have a higher risk of perioperative adverse events, especially cardiac complications after elective non-cardiac surgery 8,9 . Although this concept of stratification has recently been reported in TJA patients 10 , the relationship between insulin dependence and risk of adverse events has not been well recognized. This information will help in managing patients' expectations, in preoperative risk stratification, in implementing appropriate prevention, and with monitoring measures.
This meta-analysis aims to answer the following questions. First, what is the current proportion of insulindependent DM (IDDM) in diabetic patients undergoing TJA? Second, has there been an increase in the proportion of IDDM in TJA over the past decade? Finally, is IDDM associated with an increased risk of adverse events after TJA.

Method
T his meta-analysis was performed in accordance with the Cochrane Handbook for Systematic Reviews of Interventions and the PRISMA Checklist guidelines 11,12 (Appendix 2).

Search Strategy
The electronic databases of EMBASE, PubMed, Ovid, Medline, the Cochrane Library, Web of Science, China Science and Technology Journal Database, and China National Knowledge Infrastructure were searched from the inception dates to 10 September 2019. The search terms were as follows: (Human Isophane Insulin OR Protophane OR Protophan OR Insulin OR Humulin OR Novolin OR Insularaed) AND (arthroplasty OR knee replacement OR shoulder replacement OR hip replacement). We also identify other possible original studies by searching the Google search engine.

Inclusion Criteria
Studies were included based on the following criteria: (i) cohort or case-control studies focused on the influence of IDDM on the postoperative complication rate of total joint arthroplasty (TJA); (ii) sufficient sample size in each study; (iii) availability of reported outcomes, including proportion of IDDM patients, cardiac arrest, stroke, sepsis, myocardial infarction, extended length of stay (>5 days), on ventilator >48 h, renal failure, superficial incisional surgical site infection (SSI), deep incisional SSI, organ/space SSI, thrombotic event (vein thrombosis embolism/pulmonary embolism), reoperation, readmission, wound dehiscence, urinary tract infection, renal insufficiency, revision, unplanned intubation, pneumonia, and death. The exclusion criteria were: (i) studies that used treatments involving glucose control or combining other drugs; (ii) studies enrolling patients undergoing unicompartmental joint replacement or acute joint arthroplasty due to fracture; and (iii) review articles, expert opinions, and trials that do not consider complications.

Quality Assessment
Two authors (WLM and LMY) independently assessed the methodological quality of observational studies based on the nine-star system of the Newcastle-Ottawa scale (NOS). The NOS evaluates study population representativeness, comparability of IDDM and non-insulin dependent DM (NIDDM), assessment of outcomes, follow-up length, and adequacy of follow up ( Table 1).
The evidence quality of outcomes was systematically assessed by the two reviewers (WLM and YP) based on the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach (https://gdt.gradepro.org/ app/) 31 (Appendix 1). Although the GRADE approach considers the results of observational studies as evidence of low quality, there are a series of criteria to improve the assessment of quality levels, including a substantial effect (risk increased or decreased at least twofold), a dose-response gradient, no confounding factors, and no bias. The evidence quality of each outcome is shown in Table 4.

Data Extraction
Data were independently extracted by two authors (WLM and LMY), including: author, year, area, design, surgery, sample size, age, gender, and rate of cardiac arrest, stroke, sepsis, myocardial infarction, extended length of stay (>5 days), on ventilator >48 h, renal failure, superficial incisional SSI, deep incisional SSI, organ/space SSI, thrombotic event (vein thrombosis embolism/pulmonary embolism), reoperation, readmission, wound dehiscence, urinary tract infection, renal insufficiency, revision, unplanned intubation, pneumonia, and death. Disagreements were resolved by discussion. We also attempted to contact study authors by email for additional information if needed.

Definition of Outcomes
The primary outcomes were cardiovascular complications, pulmonary complications, kidney complications, wound complications, infection, and other complications within 30 days of surgery (unless specifically mentioned).
Understanding these outcomes can help in managing postoperative care and monitoring, improving the quality and safety of TJA. Secondary outcomes were the proportion of IDDM patients among diabetic patients undergoing TJA and its time trend.

Cardiovascular Complications
Cardiac arrest (requiring external or open cardiopulmonary resuscitation), stroke (resulting in residual neurologic deficit), and myocardial infarction (defined on the presence of the ICD-9-CM code 410.xx) were considered cardiovascular complications.

Pulmonary Complications
On ventilator >48 h, unplanned intubation, and pneumonia (hospitalized or radiologically confirmed) were considered pulmonary complications.

Kidney Complications
Renal insufficiency (an increase in serum creatinine level ≥0.3 mg/dL within 48 h) and renal failure (deterioration in renal function sufficient to require dialysis) were considered kidney complications.

Wound Complications
Wound dehiscence, superficial incisional SSI (infections with purulent drainage that occurred at the incision sites), and deep incisional SSI (clinically diagnosed infections below the fascia or joint capsule with persistent wound discharge or joint pain) were considered wound complications.

Infection
Urinary tract infections (an infection in the kidneys, ureters, bladder, or urethra), organ/space SS (infection involves organs or spaces), and sepsis (severe infection resulting in multiple organ affection) were considered postoperative infections.

Other Complications
Extended length of stay (>5 days), reoperation (unplanned return to the operating room or procedure requiring a second anesthetic event), thrombotic events (VTE and PE within 90 days), readmission (≥1 night in hospital and potentially surgically related), prosthetics revision (removal or replacement of at least one prosthetic component), and death were considered other complications.

Data Analysis
The relative risk (RR) and 95% confidence interval (CI) were calculated in Stata/SE (15.1 for Mac 64-bit Intel) for discontinuous data; P < 0.05 was considered statistical significance. The I 2 statistic was used for calculating the heterogeneity of eligible studies and its results were interpreted as follows: 25%, low heterogeneity; 50%, moderate heterogeneity; and 75%, high heterogeneity. To address heterogeneity and publication bias, multiple sensitivity analyses, including outlier removal, subgroup analysis according to type of surgery and areas, Duval and Tweedie's trim, and fill methods, were performed. If I 2 still existed (>50%), we used the randomeffect model. Egger and Harbord's test based on outcomes was used to assess publication bias for continuous and discontinuous data, respectively. Moment mixed-effect meta-regression was applied to assess time trends, and due to the varied duration of study enrollment, the beginning, middle, and end year of patient enrollment were used for regression.

Search Results
There were 1772 initial studies obtained after systematic searching. A total of 949 duplicates and 780 articles were excluded by screening abstracts. We ended up with 43 eligible studies: 19 met the inclusion criteria, 10 studies were not on TJA, 8 studies did not contain available data, 4 articles did not contain a control group, and 2 studies were non-clinical trials and were excluded. The flow chart in Fig. 1 shows the selection process of eligible studies.

Proportion
All 19 studies that were included reported the specific numbers of NIDDM and IDDM patients. Two articles 27,29 were excluded because there were too few participants and there was high heterogeneity in the sensitivity analysis. Two studies reported the proportion of IDDM patients in among TSA patients 10,13,21 , whereas 2 articles reported the proportion of IDDM patients among revision knee arthroplasty patients 14,15 , 4 reported the proportion of IDDM patients among THA patients 16,17,26,30 , 6 studies reported the proportion of IDDM patients among TKA patients 16,18,20,24,26,28 and 1 study separately reported the proportion of IDDM patients among THA and TKA patients 16 . In addition, 4 studies separately reported the proportion of IDDM patients among THA and TKA patients 19,22,23,25 . The pooled analysis showed that the proportion of IDDM patients in total joint replacement group accounted for 26% (95% CI, 24% to 28%) of DM patients (Egger's test, P = 0.518) (Fig. 2). Because of the high heterogeneity of TKA, THA, and TKA and THA groups, we conducted the subgroup analysis according to the research areas. The results showed that the proportion of IDDM patients in the Denmark group was obviously lower than that in the United States (Fig. 3), which potentially explains the high heterogeneity.
Time Trend in the Proportion of Patients with Insulin Dependence Significant time trends in the proportion of patients with insulin dependence were not observed in the first and middle enrollment year for studies published after 2000 in the metaregression (P = 0.509 and 0.149, respectively). However, the meta-regression showed a significant trend in the end enrollment year for studies published after 2000, with an outlier 16 removed (P = 0.014) (Fig. 4).

Postoperative Complications
The results of the meta-analyses are listed in Table 4.

Cardiovascular Complications
The occurrence of cardiac arrest was reported in 5 studies including 61,817 patients; no heterogeneity was found among these studies (I 2 = 0%), and cardiac arrest occurred more often in IDDM patients with an RR of 2.346 (95% CI 1.553 to 3.546). Stroke after TJA was documented in 6 studies, including 62,803 patients with I 2 of 0%. Stroke occurred more often in IDDM patients with an RR of 2.182 (95% CI 1.432 to 3.325). A total of 7 studies examined the risk of myocardial infarction. After pooled analysis, the result showed that an increased risk of myocardial infarction was associated with IDDM (RR, 1.874; 95% CI 1.461 to 2.402; I 2 = 0%).

Pulmonary Complications
Four studies were able to be pooled for the occurrence of being on a ventilator >48 h with no heterogeneity (I 2 = 0%), and the results indicated an almost threefold greater rate in IDDM patients (RR, 2.868; 95% CI, 1.829 to 4.496). Although there were 4 and 5 articles reporting on the occurrence of unplanned intubation and pneumonia, respectively, they could not be pooled in the fixed-effect model because of high heterogeneity (I 2 = 58% and 65%, respectively). Outlier removal was performed in these indictors; however, the heterogeneity remained at a high level (I 2 > 50%). In an effort to elucidate the unexplainable high heterogeneity, a predefined subgroup analysis according to type of TJA was performed on unplanned intubation and pneumonia (

Wound Complications
Five studies involving 55,092 patients reported on the rate of wound dehiscence, and the meta-analysis showed that IDDM

730
ORTHOPAEDIC SURGERY VOLUME 13 • NUMBER 3 • MAY, 2021 significant increase in the rate of readmission was observed (RR, 1.494; 95% CI, 1.381 to 1.615; I 2 = 15.8%). It is worth mentioning that there were 6 studies involving 62,803 participants reporting on the risk of death in IDDM patients after TJA with no heterogeneity (I 2 = 0%) and the pooled analysis showed a greater than twofold risk of death in IDDM patients (RR, 2.292; 95% CI, 1.568 to 3.349). No significant differences were found in the occurrence of thrombotic events (VTE/PE) and revision.

Discussion
T he global prevalence of diabetes is expected to rise from 6.4% to 7.7% between 2010 and 2030 4 ; therefore, its prevalence in the TJA population might increase accordingly. DM leads to a chronic low-level inflammatory state, accompanied by metabolic disturbance, immune decline, and other negative states. Existing published studies show that diabetic patients are more likely to develop perioperative complications than non-diabetic patients, including wound infection 25 , deep prosthesis infection 25,26 , and prothesis revision 32 .
Lovecchio et al. 22 further stratified 42,339 THA and TKA diabetic patients included in the National Surgical Quality Improvement Program into NIDDM and IDDM groups. The results showed that the risk of 30-day mortality and readmission were significantly increased in IDDM patients, which was consistent with our pooled analysis. A retrospective study by Lee et al. 14 indicated that insulin dependence was associated with septic shock after rTKA, leading to renal insufficiency and even renal failure. Therefore, the authors recommend strengthening blood sugar control in IDDM patients to reduce the risk of renal failure after rTKA. In addition, IDDM also increases the risk of blood transfusion after rTKA compared with non-diabetes mellitus and NIDDM.
Therefore, DM is a heterogeneous disease with varying severity in the TJA population. This situation must be taken into consideration when comparing the outcomes of patients with diabetes. Oliva et al. 33 showed that applying personalized postoperative management based on perioperative risk stratification could reduce the incidence of postoperative complications. With the development of medical economics, further stratification of diabetes based on insulin dependence is critical to take into consideration the rapid growth of the diabetic population.
This analysis involved 81,697 patients undergoing TJA. The results showed that IDDM patients had significantly higher risk of adverse events than NIDDM patients. Although NIDDM patients had higher risk of superficial wound infection, IDDM patients' risk of stroke, sepsis, myocardial infarction, extended length of stay (>5 days), being on a ventilator >48 h, renal failure, superficial incisional SSI, deep incisional SSI, reoperation, readmission, wound dehiscence, urinary tract infection, renal insufficiency, unplanned intubation, pneumonia, and death increased significantly.
The incidence of deep incisional SSI infection in IDDM patients was almost double that in NIDDM patients. Therefore, the risk of deep incision SSI in patients with more severe or long-term chronic diabetes after TJA might increase. In contrast, the incidence of superficial SSI in IDDM is significantly lower than that in NIDDM, which may be because insulin effectively controls blood sugar levels in peripheral blood. This is the first review, to our knowledge, to consider the effect of insulin on the complication rate of TJA, and this pooled analysis supports that clinicians should use the diabetic control method as a variable for risks in the preoperative evaluation of total joint replacement in diabetic patients. Our findings can help clinicians to stratify perioperative risks more accurately when questioning patients and provide a theoretical basis for adjusting perioperative management and setting postoperative expectations.
This meta-analysis has several limitations. First, IDDM patients might be in the late stage of DM, during which pancreatic β cells are no longer functioning due to excessive cell damage. NIDDM patients may be in the early and middle stages of DM when pancreatic β cells are still functioning 34 . Second, we cannot further stratify the patients according to the level of HbA1c; that is, to evaluate insulin's control of HbA1c. The increased risk of complications in IDDM patients may be the result of an increased level of HbA1c. Finally, although the results of this study suggest that IDDM is a risk factor for increased incidence of complications after surgery, the study was unable to establish a causal link between IDDM and complications after surgery.

Conclusion
I nsulin dependence is a high-risk factor for increased postoperative complications of TJA, and hierarchical and optimal blood glucose management may contribute to reducing the adverse complications after surgery. In addition, because the risk of sepsis, deep wound infection, organ/space SSI, urinary tract infection, renal insufficiency, and renal failure increased significantly after TJA in IDDM patients, more active postoperative antimicrobial prophylaxis may be needed on the premise of protecting renal function.

ACKNOWLEDGMENTS
T he authors wish to thank Mr Ke Qiu, Mr Dian Fan, and Mr Yuen Teng from West China Hospital for help with the methodology of this study.

Supporting Information
Additional Supporting Information may be found in the online version of this article on the publisher's web-site: