Postoperative decrease in serum albumin as predictor of early acute periprosthetic infection after total knee arthroplasty

Abstract

Purpose

Patients with hypoalbuminemia, (serum albumin (SA) < 3.5 g/dL) are at greater risk for complications after surgery, including increased postoperative infection rates after total knee arthroplasty (TKA). This study aimed to analyze both preoperative and postoperative SA in patients who experienced acute periprosthetic infection within the first 4 weeks after surgery.

Methods

We retrospectively analyzed data from 490 consecutive TKAs (314 patients). Five patients developed early acute infection requiring surgical treatment. SA data were collected preoperatively (SA₀) and 1 week postoperatively (SA_1W) to evaluate SA dynamics. Multiple patient and operative parameters that could influence SA were also analyzed.

Results

No statistical differences were found in parameters expected to influence SA values between the surgically treated (STG) and non-treated groups (non-STG). None of the patients in STG had SA₀ and SA_1W below 3.5 g/dL. However, the amount and rate of SA reduction before and after surgery were significantly greater in STG than in non-STG.

Conclusion

SA dynamics revealed a greater reduction in both the amount and rate in STG before and after surgery than in non-STG. No correlation was found between early acute periprosthetic infection after TKA and each SA₀ and SA_1W time point. Further evaluation of the SA value of 3.5 g/dL as a threshold for acute early acute infection is warranted.

Introduction

Serum albumin (SA) is a validated screening tool for predicting postoperative complications after total joint arthroplasty (TJA) [1–5]. Patients with hypoalbuminemia (< 3.5 g/dL) develop higher postoperative complications compared with normal subjects [1–8]. Mobagwu et al. [9] reported a meta-analysis of seven large-scale articles detailing complications associated with hypoalbuminemia, concluding that hypoalbuminemia after elective TJA led to an all-cause relative risk increase of 1.93.

Increased postoperative wound infection rates are one of the complications in patients with hypoalbuminemia [1, 2, 4–8]. However, most reports are based on preoperative SA values alone. Including postoperative SA and its dynamics in addition to preoperative SA for postoperative infections could better assess their association.

Therefore, this study aimed to compare the SA values and their dynamics in the preoperative and first postoperative week of the early acute periprosthetic infection, defined as occurring within the first 4 weeks postoperatively [10] and requiring additional surgery after the initial TKA procedure. Data from infected patients were compared with those of non-infected patients to clarify differentiating characteristics.

Materials and methods

Patients and study design

Our institutional review board approved this retrospective evaluation of medical records from patients treated with primary TKA from January 2001 to September 2023. Informed consent was obtained from all patients. A total of 490 consecutive primary TKAs (314 patients) were retrospectively investigated. Patients who had undergone revision arthroplasties or previous tibial osteotomies, and patients with rheumatoid arthritis were excluded. Clinical characteristics of the patients are summarized in Table 1.

Table 1.

Patient clinical characteristics

Characteristic	Value
Number of patients	387 patients (490 knees)
Median age	74 (69, 78) (range; 34–89) years
Gender	Male; 65 pts, 77 TKAs, Female 322 pts, 413 TKA s
Preoperative diagnosis	Osteoarthritis, 490 TKAs
Body height cm	151 (147, 155) (range; 135,184)
Body weight kg	58 (53,66) (range; 38–130)
BMI (kg/m2)	26 (23,28) (range; 18–42)

BMI, body mass index; pts, patients; TKA, total knee arthroplasty

Patients were stratified into those who developed an early acute periprosthetic infection that required subsequent surgical treatment (surgical treatment group, STG; n = 5 TKAs) and those who did not (non-STG; n = 485 TKAs). During post-TKA weeks 2–4, five knees in five patients developed early acute infection as indicated by massive discharge or joint swelling with elevated C-reactive protein (median, 10.5 mg/L; range, 1.3–23.1). These patients required an additional surgical procedure on median postoperative day 21 (range, 14–74). Five knees (primary causative organisms: Staphylococcus aureus, 2 cases; methicillin-resistant Staphylococcus aureus (MRSA), 2 cases; and Candida and Enterococcus spp., 1 case) required massive irrigation, meticulous joint debridement, or synovectomy, and one knee infected with MRSA required the removal of components.

SA values preoperatively (SA₀) and 1 week postoperatively (SA_1W) were investigated. Differences between timepoints of SA were calculated and expressed as ΔSA. ΔSA values relative to SA₀ were converted to a percentage as (SA_1w − SA₀)/SA₀ and defined as reduction rate (RR). In addition, we examined the following factors Factors potentially influencing SA₀ were also examined: sex [1, 3, 4, 11, 12], age [1, 3, 4, 11], body mass index [4, 13], American Society of Anesthesiologists [14] grade [3, 5, 11], preoperative complications (hypertension, hyperlipidemia, diabetes) [1, 4, 5, 11, 12, 15], and smoking history [12], clinical score (Hospital for Special Surgery knee-rating score) [16], operative time [12], tourniquet time, and blood loss [5, 12].

Data analyses

Differences between the STG and non-STG were analyzed using Wilcoxon’s rank sum test and Fisher’s exact test. Receiver operating characteristic (ROC) analysis was performed for each variable to determine the optimal cut-off value. We also calculated the area under the curve (AUC) to determine which variables were accurate in predicting deep infection after TKA. A higher AUC represents a better performance and an AUC of more than 0.7 is considered good efficiency [17]. To predict significant factors that associate with periprosthetic infection after TKA, we performed univariate and multivariate logistic analysis. Significant factors were selected by using the stepwise selection method. The study power was calculated based on the sample size. Statistical analyses were performed using IBM SPSS Statistics for Windows, version 20.0 (IBM Japan, Tokyo, Japan). All variables are expressed as medians (25th and 75th percentiles). In all tests, a p-value < 0.05 was considered to indicate statistical significance.

Results

There were no statistical differences in any factors that were reported to influence or might influence SA values in the perioperative period between the STG and non-STG (Table 2).

Table 2.

Comparison of factors reported to influence or potentially influence serum albumin values in the perioperative period between the surgical and non-surgical treatment groups

Factor	Non-STG (N = 485)	STG (N = 5)	p value
Patients agea	74 (69,78)	75 (75,81)	0.135
Gender (male/female)b	75/410	2/3	0.177
BMIa	26 (23,28)	24 (23/26)	0.653
ASA (I/II) Grade b [14]	143/342	2/3	0.635
HSS　scorea [16]	45 (37,51)	47 (36, 50)	0.830
Hypertension: yes/ no b	303/182	3/2	> 0.999
Hyperlipidemia: yes/ no b	96/389	0/5	0.588
Diabetes Mellites: yes/ no b	59/426	1/4	0.481
Smoking history: yes/ no b	24/461	1/4	0.231
Operation timea	55 (50, 61)	53 (48, 58)	0.627
Tourniquet timea	59 (53, 65)	55 (51, 69)	0.438
Measured blood loss (ml)	420 (200, 720)	700 (500, 760)	0.346

^a Wilcoxon’s rank sum test; ^b Fisher’s exact test. Values are presented as average (SD) or number

ASA; American Society of Anesthesiologists: BMI, body mass index; HSS, Hospital for Special Surgery; non-STG, non-surgical treatment group; STG, surgical treatment group

Median values for SA₀ were 4.4 (4.2, 4.6) g/dL in the non-STG and 4.4 (4.3, 4.5) g/dL in the STG, (P = 0.925), and those for SA_1W were 3.8 (3.6, 3.9) g/dL and 4.4 (4.3, 4.5) g/dL (P = 0.072), respectively (Table 3). ΔSA in the non-STG was − 0.6 (− 0.8, − 0.4) and − 0.8 (− 0.9, − 0.8) in the STG, with a significant difference (P = 0.048). The RR was − 13.6% (− 17.7%, − 13.4%) in the non-STG and − 19.6% (− 20.1%, − 18.1%) in the STG (P = 0.038) (Table 3). Both univariate and multivariate logistic regression analyses were performed to identify variables to predict STG risk. However, no significant variables including SA₀, SA_1W, ΔSA, and RR were identified.

Table 3.

Comparison of serum albumin data between the surgical treatment group (n = 5) and non-surgical treatment group (n = 485) during 1 week after surgery

Median (percentile), range	Non-STG	STG	P
SA0	4.4 (4.2,4.6), 2.3–5.1	4.4 (4.3, 4.5), 4.2–4.5	0.925
SA1w	3.8 (3.6, 3.9), 2.6–4.5	3.6 (3.5, 3.6), 3.5–3.7	0.072
ΔSA	−0.6 (− 0.8, − 0.4), − 1.3–0.4	−0.8 (− 0.9, − 0.8), − 0.9 − 0.5	0.048
Reduction rate; ΔSA /SA0 × 100 (%)	−13.6% (− 17.7%, − 13.4%) −32.7% − 18.5%	−19.6% (− 20.1%, − 18.1%) −21.0% - −11.7%	0.038

Values in bold type are statistically significant

Non-STG, non-surgical treatment group; SA, serum albumin; SA₀, preoperative SA; SA_1w, postoperative SA at 1 week; ΔSA, SA_1W − SA₀; STG, surgical treatment group

Four patients (0.8%) in the non-STG and no one (0%) in the STG showed a SA₀ level of less than 3.5 g/dL, and 63 (13%) and 0 (0%) in SA_1w, respectively (Table 4).

Table 4.

Distribution of patients between serum albumin concentration ranges at each time point in the surgical treatment group and non-surgical treatment group

Serum Albumin Level (g/dL)	Number of patient (%)
Group	Non-STG		STG
Time point	SA0	SA1w	SA0	SA1w
< 3.0	2 (0.4)	3 (0.6)	0	0
3.0-3.4	2 (0.4)	60 (12.4)	0	0
3.5–3.9	38 (7.9)	301 (62.3)	0	5 (100%)
4.0-4.4	254 (52.0)	119 (24.2)	3 (60%)	0
≥ 4.5	189 (39.3)	2 (0.4)	2 (40%)	0

Non-STG, non-surgical treatment group; SA, serum albumin; SA₀, preoperative SA; SA_1w, postoperative SA at 1 week; STG, surgical treatment group

In the ROC analysis, the AUC values [95% confidence interval] for SA₀, SA_1W, ΔSA, and RR were 0.512 (0.368, 0.657), 0.734 (0.632, 0.837), 0.757(0.550, 0.963), and 0.770 (0.568, 0.972) respectively. The cut-off value of RR, the best performing the AUC values of the above, showed − 18.4%. ROC curve analysis revealed that the cut-off value for RR was − 18.4% (AUC 0.770; sensitivity 80.0%; specificity 80.2%), and that the negative predictive value was 99.7%, but the positive predictive value was 4.0%.

The estimated power of Wilcoxon’s rank sum test with the given sample size was calculated as 28.8% with the effect size of 0.5 and alpha-error of 0.05.

Discussion

This study yielded several key findings. Firstly, the STG never had SA₀ or SA_1W below 3.5 g/dL, the conventional cut-off for a higher likelihood of postoperative complications. Secondly, the reduction in SA amount and rate was significantly greater in the STG compared to the non-STG. Lastly, the optimal cut-off value for predicting early acute periprosthetic infection was an RR of − 18.4%.

Defining a SA₀ value of 3.5 g/dL as a threshold risk factor for infection [1, 2, 4–8], the incidence of the present study showed only less than 1% in the non-STG and 0% in the STG. The median SA₀ for both groups was 4.4 g/dL, with all SA₀s above 4.0 g/dL, especially in the STG. Previous studies reported hypoalbuminemia (< 3.5 g/dL) incidences of 2.7% (Man et al. [8]), 3.6% (Black et al. [18]), 8.5% (Huang et al. [7]). Nelson et al. [5] found a rate of approximately 4.4% (1,570/35,573) for low SA (< 3.5 g/dL) in the U.S. National Surgical Quality Improvement Program (NSQIP) cohort undergoing elective TKAs. The analysis of 217 cases by Greene et al. [6] found only 6 cases (3%) in their analysis. They also reported SA₀ by the occurrence of wound complications at 10 days postoperatively, with a mean (1 SD) of 4.30 (0.33). In the no-complication group, this was 4.22 (0.31) in the group with persistent serous drainage from their wounds, and 4.13 (0.54) in the major wound complication group. Furthermore, there have been various reports on SA₀ cut-off values. Kishawi et al. [4] also found an albumin level–dependent effect on the odds of infectious adverse outcomes. For the primary arthroplasty cohorts, severely low albumin levels were classified as < 3.0 g/dL by using the U.S. NSQIP database, as the large sample size results from the multicenter patient population. In addition, in seven large multicenter studies, Mobagwu et al. [9] reported the same trend in postoperative complications whether the cut-off value was 3.0 g/dL, 3.5 g/dL, or 3.9 g/dL. Furthermore, Morey et al. [19] reported no association between postoperative wound-related complications and malnutrition as defined using various SA cut-off values. These reports, including the present results, suggest that the need for further validation of a 3.5 g/dL SA value as a threshold for early acute infection.

SA_1w values were below 3.5 g/dL in 13% of the non-STG cases but in none of the STG cases. Therefore, This indicates that a SA_1w value of 3.5 g/dL may not be reliable cut-off value for acute periprosthetic infection. Most previous reports [1, 2, 4, 5, 7, 8] concluded that hypoalbuminemia affects postoperative infection based solely on SA₀. Based on the present results, to maintain rationality with respect to the threshold value of 3.5 g/dL, it is necessary to interpret the patients that showed a SA₀ of less than 3.5 g/dL as “hypoalbuminemia” did not improve after surgery and the postoperative SA might remain below 3.5 g/dL or worsened, resulting in the development of complications. Greene et al. [6] also reported that none of the 38 of 217 patients analyzed (17%) returned to preoperative values by 10 days postoperatively. Therefore, it is suggested that appropriate nutritional management to improve SA, including the administration of essential amino acids (EAA) might be necessary. It is expected to produce a value above 3.5 g/dL in patients with preoperative values below 3.5 g/dL which may be a preventive measure for the development of complications in the early stages of surgery and thereafter. To date, a couple of studies [4, 9] also suggested that reaching optimal albumin levels in such a period be encouraged. Conservative management (i.e., physical therapy and corticosteroid injections) should be undertaken before considering surgery to minimize the risk of perioperative complications and to improve the outcome of the surgical procedure.

Few reports compare SA reduction amounts and rates in TJA. Greene et al. [6] reported a 72% postoperative SA drop relative to preoperative values. There have been a couple of reports on ΔSA in determining the efficacy of EAA administration after TKA surgery. Dreyer et al. [20] and Ueyama et al. [21] reported similar trends in ΔSA after TKA surgery. In the present study, both the reduction of amount and rates were significantly higher in the STG than the non-STG. Overall, the trends reported in these studies are similar to our results, indicating that perioperative SA values are highly dynamic (range; approximately 10–20%). Indeed, a systematic review also reported ΔSA, an interesting biomarker that shows promising results and may more accurately predict complications after gastrointestinal surgery [22]. ΔSA demonstrates superior capability compared to SA₀ in identifying patients with postoperative complications following liver surgery [23]. The present results may indicate that, at least for early infections after TKA, the trend of SA values based on their changes were better predictive variables than those of each time point such as SA₀ and SA_1W, similar to the reports described earlier.

Lastly, based on the results of the cut-off value, when RR is less than 18.4%, there is a 99.7% probability that infection will not occur, but when RR is greater than 18.4%, only 4.0% of these patients may develop infection. Therefore, judging from statistical results, this cut-off value indicated “infection is unlikely unless RR exceeds 18.4%,” rather than “infection is suspected when RR exceeds 18.4%.” The clinical implication is that for patients whose RR exceeds 18.4%, correction of their cut-off values by supplementation with EAAs may result in ever lower rates of early acute infection. For example, an early benefit of perioperative EAA supplementation (in the first 4 weeks after TKA) is a reduction in atrophy of the rectus femoris muscle in patients undergoing TKA [21].

This study has three limitations. First and foremost, the results in this study had limitations owing to the study’s retrospective nature based on using the medical record. Therefore, prospective studies are needed to confirm the validity of this study. The second limitation is that the present study focused on the development of infection in the acute postoperative period and cannot address the association between delayed infection and serum albumin level. Third, with respect to the comparison of variables in the STG and non-STG in this study, the sample size and the results of the power analysis must be considered and interpreted in light of the possibility that Type I or II error may have occurred. However, considering that the incidence rate of periprosthetic joint infection was 0.5–1.3% even in the recent reports [24–26], it is not surprising and is reasonable that there are only 5 cases in the STG out of 490 TKAs. Finally, this was a single-center, single-race study; a multicenter, multiracial setting is needed to confirm the universality of the results. However, the main strength of this study is that it is the first to identify SA trends and their correlation with early acute infection during the perioperative period, specifically, before and 1 week after TKA.

Conclusions

The STG showed a greater reduction in both the amount and rate in terms of SA dynamics before and after surgery than the non-STG. No correlation was found between early acute periprosthetic infection after TKA and each time point such as SA₀ and SA_1W. The cut-off value for RR was also identified, suggesting that the validation of SA as a screening tool for predicting postoperative complications may be enhanced by taking their changes into account, in addition to the conventional evaluation of SA₀ alone.

Abbreviations

SA

Serum Albumin

TKA

Total Knee Arthroplasty

SA₀

SA Preoperatively

SA_1W

1 week postoperatively

STG

Surgically Treated Group

non-STG

non-surgically-treated group (non-STG)

TJA

Total Joint Arthroplasty

MRSA

Methicillin-Resistant Staphylococcus aureus

ΔSA

Differences between timepoints of SA

RR

ΔSA values relative to SA₀ were converted to a percentage as (SA_1w − SA₀)/SA₀ and defined as Reduction Rate

ROC

Receiver Operating Characteristic

AUC

The Area Under the Curve

Author contributions

Y.I. contributed to the study conception and design, drafted the article, and ensured the accuracy of the data and analysis. H.N., J.S., and I.T. contributed to the study conception and design and to the analysis and interpretation of the data. H.I., R.I., Ke.I. and Ka.I. contributed to the data collection. S.T. provided statistical expertise and contributed to ensuring the accuracy of the data and analysis. All authors approved the final manuscript.

Funding

None declared.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

The local institutional review board approved this study. All patients provided informed consent. Informed consent was waived by the institutional review board because of the retrospective study design.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.